Simplify and harden Serializer and SerialIter:

Reduce both classes to a small, header-only interface built on Slice. Serializer becomes a strictly append-only buffer and SerialIter a strictly forward-only reader; all deprecated and duplicated paths are removed and data copying is entirely eliminated. But the wire format, you'll be happy to know, remains unchanged. Serializer: - Make the buffer append-only. Remove modData, resize, chop, mutable begin/end and getDataPtr, the (void const*, size_t) constructor, and the read-side helpers get8, getInteger and getBitString. - Collapse the accessors (peekData, getData, getDataPtr, getDataLength, getLength, getString, begin/end, operator==) into slice(), data() and size(). - Add takeData(), which moves the buffer out. Use it wherever callers copied via getData/peekData or reached in with std::move(modData()). - addRaw and addVL take a Slice only; the Blob, (void const*, int), Serializer and iterator-range overloads are removed. Both throw if the input aliases the serializer's own storage. - Funnel every append through one private write(), which reserves for the combined size of its parts before inserting. An add that throws leaves the serializer unchanged. - Constrain add8/add16/add32/add64 on the exact argument type: no implicit conversions, no bool or character types, and the width must match the function name. Call sites that relied on silent narrowing now cast explicitly. - The add functions return void rather than the offset written to. - Replace getSHA512Half() with a hash_append hidden friend; callers use sha512Half(s). - Move the VL length-prefix constants into xrpl::detail as std::size_t and fold header encoding into addVL. Remove addEncoded, encodeLengthLength, decodeLengthLength and decodeVLLength; decoding now lives in SerialIter::getVL. SerialIter: - Construct only, and explicitly, from a Slice. Remove the (void const*, size_t) and array-reference constructors. A null pointer with a non-zero size throws. - Remove reset() and skip(), and with them the used_ counter. A reader cannot be rewound; copy it to retain a position. - getVL() returns a Slice into the underlying buffer instead of a Blob copy. Remove getVLDataLength, getSlice, getRaw, getVLBuffer and getRawHelper. - Route every read through a single bounds-checked consume(). Integers are decoded with std::bit_cast and std::byteswap, which drops the boost::endian dependency. - Replace getBytesLeft() with size(), returning std::size_t, and add slice() for the unread remainder. - getBitString<T> takes the BaseUInt type rather than <Bits, Tag>, so tagged types (AccountID, Currency, MPTID) are read directly. - Add geti16() and mark the accessors [[nodiscard]]. Error reporting: - VL length errors, on both the encode and the decode side, now throw std::runtime_error; they previously threw std::overflow_error. - Short reads all throw std::runtime_error("short buffer") in place of the per-function messages. Other changes: - Slice: add overlap() global utility function, used for the aliasing checks above. - sqlBlobLiteral takes a Slice. - LedgerHeader: remove deserializePrefixedHeader. Its caller in InboundLedger skips the prefix locally; the prefix is still not validated, which is now marked with a FIXME. - STAccount: deserialize from the getVL slice via AccountID::fromRaw and remove the (SField, Buffer) constructor. - STVector256: serialize the whole vector with a single addVL, relying on contiguous storage and the absence of padding in UInt256. - STInteger: static_assert that the type is serializable and dispatch add() on sizeof(Integer). Read STInteger<std::int32_t> with geti32(). Log an unknown transaction result code as a number instead of a raw character. - STTx: fold the four-argument getMetaSQL overload into its only caller. - SHAMapTreeNode: the two leaf deserializers share a splitTag helper that slices off the trailing tag, replacing the copy into a Serializer followed by getBitString-at-offset and chop. The size check is done once, up front. - SHAMap::getNodeFat: serialize each node into its own Serializer and move the buffer into the reply instead of copying out of a shared one. - LedgerMaster::getCloseTimeByHash: parse the stored header with deserializeHeader instead of skipping hard-coded offsets. - ConsensusTransSetSF::gotNode and LedgerReplayMsgHandler parse directly from the incoming bytes without an intermediate Serializer copy. - TransactionSign: use sterilize() for the serialization round trip. Tests: - Port to the new interfaces; byte payloads become std::array so they can go through makeSlice. - STAccount: the bad-size case now covers both a short and a long field, and fails if no exception is thrown. Previously it passed silently in that case.
Restructure the slab allocator
2026-10-21 19:18:22 +00:00 · 2026-10-05 19:50:58 -07:00 · 2026-10-05 19:42:53 -07:00 · 2026-10-05 19:42:52 -07:00 · 2026-10-05 19:42:50 -07:00 · 2026-10-05 19:42:50 -07:00
871 changed files with 27715 additions and 44419 deletions
@@ -5,10 +5,7 @@ Checks: "-*,
  -bugprone-exception-escape,
  -bugprone-implicit-widening-of-multiplication-result,
  -bugprone-narrowing-conversions,
-  -bugprone-signed-bitwise,
-  -bugprone-std-exception-baseclass,
  -bugprone-throwing-static-initialization,
-  -bugprone-unhandled-code-paths,

  cppcoreguidelines-*,
  -cppcoreguidelines-avoid-c-arrays,
@@ -17,7 +14,6 @@ Checks: "-*,
  -cppcoreguidelines-avoid-magic-numbers,
  -cppcoreguidelines-avoid-non-const-global-variables,
  -cppcoreguidelines-c-copy-assignment-signature,
-  -cppcoreguidelines-explicit-constructor,
  -cppcoreguidelines-interfaces-global-init,
  -cppcoreguidelines-macro-usage,
  -cppcoreguidelines-missing-std-forward,
@@ -36,7 +32,6 @@ Checks: "-*,
  llvm-namespace-comment,

  misc-*,
-  -misc-explicit-constructor,
  -misc-multiple-inheritance,
  -misc-no-recursion,
  -misc-non-private-member-variables-in-classes,
@@ -50,8 +45,6 @@ Checks: "-*,
  -modernize-avoid-c-style-cast,
  -modernize-return-braced-init-list,
  -modernize-use-integer-sign-comparison,
-  -modernize-use-string-view,
-  -modernize-use-structured-binding,
  -modernize-use-trailing-return-type,

  performance-*,
@@ -60,7 +53,6 @@ Checks: "-*,
  -performance-noexcept-move-constructor,
  -performance-unnecessary-copy-initialization,
  -performance-unnecessary-value-param,
-  -performance-use-std-move,

  readability-*,
  -readability-avoid-const-params-in-decls,
@@ -73,10 +65,7 @@ Checks: "-*,
  -readability-named-parameter,
  -readability-qualified-auto,
  -readability-redundant-access-specifiers,
-  -readability-redundant-nested-if,
-  -readability-redundant-qualified-alias,
  -readability-static-accessed-through-instance,
-  -readability-trivial-switch,
  -readability-uppercase-literal-suffix
  "
 # ---
@@ -108,75 +108,3 @@ endfunction()

 function(patch_nix_binary target)
 endfunction()
-
-function(rpcspec_generate_instantiations)
-    set(options)
-    set(oneValueArgs OUT_VAR VALUE_TYPE VIEW_HEADER INCLUDE_DIR)
-    set(multiValueArgs HANDLERS)
-    cmake_parse_arguments(
-        THIS_FUNCTION_PREFIX
-        "${options}"
-        "${oneValueArgs}"
-        "${multiValueArgs}"
-        ${ARGN}
-    )
-endfunction()
-
-function(corrosion_import_crate)
-    set(options
-        ALL_FEATURES
-        NO_DEFAULT_FEATURES
-        NO_STD
-        NO_LINKER_OVERRIDE
-        NO_USES_TERMINAL
-        LOCKED
-        FROZEN
-    )
-    set(oneValueArgs MANIFEST_PATH PROFILE IMPORTED_CRATES)
-    set(multiValueArgs
-        CRATE_TYPES
-        CRATES
-        FEATURES
-        FLAGS
-        OVERRIDE_CRATE_TYPE
-    )
-    cmake_parse_arguments(
-        THIS_FUNCTION_PREFIX
-        "${options}"
-        "${oneValueArgs}"
-        "${multiValueArgs}"
-        ${ARGN}
-    )
-endfunction()
-
-function(corrosion_set_env_vars target_name env_var)
-endfunction()
-
-function(corrosion_add_cxxbridge cxx_target)
-    set(options)
-    set(oneValueArgs CRATE)
-    set(multiValueArgs FILES)
-    cmake_parse_arguments(
-        THIS_FUNCTION_PREFIX
-        "${options}"
-        "${oneValueArgs}"
-        "${multiValueArgs}"
-        ${ARGN}
-    )
-endfunction()
-
-function(_unlink_libgcc_s crate)
-endfunction()
-
-function(add_xrpl_crate name)
-    set(options)
-    set(oneValueArgs CRATE)
-    set(multiValueArgs FILES)
-    cmake_parse_arguments(
-        THIS_FUNCTION_PREFIX
-        "${options}"
-        "${oneValueArgs}"
-        "${multiValueArgs}"
-        ${ARGN}
-    )
-endfunction()
@@ -15,9 +15,9 @@ inputs:
    required: false
    default: "false"
  log_verbosity:
-    description: 'The logging verbosity ("quiet", "verbose"), or empty to use the Conan defaults.'
+    description: "The logging verbosity."
    required: false
-    default: ""
+    default: "verbose"
  sanitizers:
    description: "The sanitizers to enable."
    required: false
@@ -35,16 +35,6 @@ runs:
        LOG_VERBOSITY: ${{ inputs.log_verbosity }}
        SANITIZERS: ${{ inputs.sanitizers }}
      run: |
-        # By default, leave the verbosity unset, so CMake configure output is
-        # shown, but compile commands and Boost's b2 debug output (~85k lines
-        # when "verbose") are not.
-        VERBOSITY_ARGS=()
-        if [[ -n "${LOG_VERBOSITY}" ]]; then
-            VERBOSITY_ARGS=(
-                --conf:all tools.build:verbosity="${LOG_VERBOSITY}"
-                --conf:all tools.compilation:verbosity="${LOG_VERBOSITY}"
-            )
-        fi
        conan install \
            --profile:all ci \
            --build="${BUILD_OPTION}" \
@@ -52,13 +42,6 @@ runs:
            --options:host='&:xrpld=True' \
            --settings:all build_type="${BUILD_TYPE}" \
            --conf:all tools.build:jobs=${BUILD_NPROC} \
-            "${VERBOSITY_ARGS[@]}" \
-            --format=json \
-            . >"${RUNNER_TEMP}/conan-graph.json"
-
-    # Tools that run during the build may only load glibc from the Nix store,
-    # as their package ID survives a GCC runtime update.
-    - name: Check build-context packages for Nix store dependencies (Linux)
-      if: ${{ runner.os == 'Linux' }}
-      shell: bash
-      run: ./bin/nix/check-build-context-runtime.sh "${RUNNER_TEMP}/conan-graph.json"
+            --conf:all tools.build:verbosity="${LOG_VERBOSITY}" \
+            --conf:all tools.compilation:verbosity="${LOG_VERBOSITY}" \
+            .
@@ -174,10 +174,8 @@ test.server > xrpl.basics
 test.server > xrpl.config
 test.server > xrpld.app
 test.server > xrpld.core
-test.server > xrpld.rpc
 test.server > xrpl.json
 test.server > xrpl.protocol
-test.server > xrpl.resource
 test.server > xrpl.server
 test.unit_test > xrpl.basics
 test.unit_test > xrpl.protocol
@@ -1,5 +1,5 @@
 {
-  "image_tag": "sha-3d526d4",
+  "image_tag": "sha-060957e",
  "configs": {
    "ubuntu": [
      {
@@ -74,7 +74,7 @@
        "extra_cmake_args": "-Dvalidator_keys=ON",
        "package": {
          "type": "deb",
-          "image": "ghcr.io/xrplf/xrpld/packaging-debian:sha-e6055dd"
+          "image": "ghcr.io/xrplf/xrpld/packaging-debian:sha-3a2d19f"
        }
      },
      {
@@ -86,7 +86,7 @@
        "extra_cmake_args": "-Dvalidator_keys=ON -Dassert=ON",
        "package": {
          "type": "deb",
-          "image": "ghcr.io/xrplf/xrpld/packaging-debian:sha-e6055dd",
+          "image": "ghcr.io/xrplf/xrpld/packaging-debian:sha-3a2d19f",
          "variant": "assert"
        }
      }
@@ -101,7 +101,7 @@
        "extra_cmake_args": "-Dvalidator_keys=ON",
        "package": {
          "type": "rpm",
-          "image": "ghcr.io/xrplf/xrpld/packaging-rhel:sha-e6055dd"
+          "image": "ghcr.io/xrplf/xrpld/packaging-rhel:sha-3a2d19f"
        }
      }
    ]
@@ -12,7 +12,8 @@ on:
      - "!nix/docker/README.md"
      - "!nix/devshell.nix"
      - "!nix/check-tools/**"
-      - "bin/nix/default-loader-path.sh"
+      - "bin/default-loader-path.sh"
+      - "bin/install-sanitizer-libs.sh"
  pull_request:
    paths:
      - ".github/workflows/build-nix-images.yml"
@@ -24,8 +25,8 @@ on:
      - "!nix/devshell.nix"
      - "!nix/check-tools/**"
      - "bin/check-tools.sh"
-      - "bin/nix/default-loader-path.sh"
-      - "bin/install/sanitizer-libs.sh"
+      - "bin/default-loader-path.sh"
+      - "bin/install-sanitizer-libs.sh"
  workflow_dispatch:

 concurrency:
@@ -5,12 +5,13 @@ on:
    branches:
      - develop
    paths:
-      - "bin/install/packaging-tools.sh"
+      - ".github/workflows/build-packaging-images.yml"
+      - "bin/install-packaging-tools.sh"
      - "package/images/packaging/**"
  pull_request:
    paths:
      - ".github/workflows/build-packaging-images.yml"
-      - "bin/install/packaging-tools.sh"
+      - "bin/install-packaging-tools.sh"
      - "package/images/packaging/**"
  workflow_dispatch:

@@ -5,6 +5,7 @@ on:
    branches:
      - develop
    paths:
+      - ".github/workflows/build-pre-commit-image.yml"
      - "bin/pre-commit/Dockerfile"
      - "rust-toolchain.toml"
  pull_request:
@@ -34,7 +34,7 @@ permissions:
 jobs:
  audit:
    runs-on: ubuntu-latest
-    container: ghcr.io/xrplf/xrpld/nix-ubuntu:sha-3d526d4
+    container: ghcr.io/xrplf/xrpld/nix-ubuntu:sha-060957e
    permissions:
      contents: read
      # Needed to open an issue on scheduled failures.
@@ -66,7 +66,7 @@ jobs:

      - name: Create issue
        if: ${{ steps.audit.outcome != 'success' && github.event_name == 'schedule' }}
-        uses: XRPLF/actions/create-issue@b489c21d53928f7dab4874912fec40e4f6c9e85e
+        uses: XRPLF/actions/create-issue@2b8bc36af85b88bca0dd7bfac2e2dc05f94ad712
        with:
          title: "cargo audit found vulnerabilities"
          body_file: /tmp/cargo-audit-issue.md
@@ -53,7 +53,7 @@ jobs:

      - name: Create issue
        if: ${{ failure() && github.event_name == 'schedule' && github.repository == 'XRPLF/rippled' }}
-        uses: XRPLF/actions/create-issue@b489c21d53928f7dab4874912fec40e4f6c9e85e
+        uses: XRPLF/actions/create-issue@2b8bc36af85b88bca0dd7bfac2e2dc05f94ad712
        with:
          title: "Releases not merged back into develop"
          body_file: ${{ env.ISSUE_FILE }}
@@ -79,7 +79,7 @@ jobs:
        uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1

      - name: Prepare runner
-        uses: XRPLF/actions/prepare-runner@8e98d61f011fd2a67f3fde5f0f075662c099a527
+        uses: XRPLF/actions/prepare-runner@b3e255d74d785d053e4903da8ac90983cd7d9e82
        with:
          enable_ccache: false

@@ -108,7 +108,7 @@ jobs:

      - name: Upload regenerated snapshot
        if: failure()
-        uses: actions/upload-artifact@cf430e030ddbb5b0abf93d22962f4752f3646cd9 # v7.0.2
+        uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a # v7.0.1
        with:
          name: check-tools-${{ runner.os }}-${{ runner.arch }}
          path: ${{ matrix.snapshot }}
@@ -93,8 +93,9 @@ jobs:
            .github/workflows/reusable-upload-recipe.yml
            .clang-tidy
            .codecov.yml
+            bin/check-nix-store-refs.sh
            bin/check-tools.sh
-            bin/nix/**
+            bin/default-loader-path.sh
            cfg/**
            cmake/**
            conan/**
@@ -31,8 +31,9 @@ on:
      - ".github/workflows/reusable-upload-recipe.yml"
      - ".clang-tidy"
      - ".codecov.yml"
+      - "bin/check-nix-store-refs.sh"
      - "bin/check-tools.sh"
-      - "bin/nix/**"
+      - "bin/default-loader-path.sh"
      - "cfg/**"
      - "cmake/**"
      - "conan/**"
@@ -14,7 +14,7 @@ on:
 jobs:
  # Call the workflow in the XRPLF/actions repo that runs the pre-commit hooks.
  run-hooks:
-    uses: XRPLF/actions/.github/workflows/pre-commit.yml@d7f2acb47d9662e1e2f8b1f5feffa6d21cc3290a
+    uses: XRPLF/actions/.github/workflows/pre-commit.yml@279ec358f4a1be4088be3e024b07916fa97c75b6
    with:
      runs_on: ubuntu-latest
-      container: '{ "image": "ghcr.io/xrplf/xrpld/pre-commit:sha-70b8fd3" }'
+      container: '{ "image": "ghcr.io/xrplf/xrpld/pre-commit:sha-473fe44" }'
@@ -41,13 +41,13 @@ env:
 jobs:
  build:
    runs-on: ubuntu-latest
-    container: ghcr.io/xrplf/xrpld/nix-ubuntu:sha-3d526d4
+    container: ghcr.io/xrplf/xrpld/nix-ubuntu:sha-060957e
    steps:
      - name: Checkout repository
        uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1

      - name: Prepare runner
-        uses: XRPLF/actions/prepare-runner@8e98d61f011fd2a67f3fde5f0f075662c099a527
+        uses: XRPLF/actions/prepare-runner@b3e255d74d785d053e4903da8ac90983cd7d9e82
        with:
          enable_ccache: false

@@ -129,7 +129,7 @@ jobs:
        uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1

      - name: Prepare runner
-        uses: XRPLF/actions/prepare-runner@8e98d61f011fd2a67f3fde5f0f075662c099a527
+        uses: XRPLF/actions/prepare-runner@b3e255d74d785d053e4903da8ac90983cd7d9e82
        with:
          enable_ccache: ${{ inputs.ccache_enabled }}

@@ -186,6 +186,9 @@ jobs:
        with:
          build_nproc: ${{ steps.nproc.outputs.nproc }}
          build_type: ${{ inputs.build_type }}
+          # Set the verbosity to "quiet" for Windows to avoid an excessive
+          # amount of logs. For other OSes, the "verbose" logs are more useful.
+          log_verbosity: ${{ runner.os == 'Windows' && 'quiet' || 'verbose' }}
          sanitizers: ${{ inputs.sanitizers }}

      - name: Configure CMake
@@ -254,20 +257,20 @@ jobs:
      # cache included, since what it holds is what gets uploaded and reused.
      - name: Check the build output for Nix store references (Nix toolchain)
        if: ${{ inputs.toolchain == 'nix' }}
-        run: ./bin/nix/check-nix-store-refs.sh "${BUILD_DIR}"
+        run: ./bin/check-nix-store-refs.sh "${BUILD_DIR}"

      - name: Check the Conan cache for Nix store references (Nix toolchain)
        if: ${{ inputs.toolchain == 'nix' }}
-        run: ./bin/nix/check-nix-store-refs.sh "${CONAN_HOME}"
+        run: ./bin/check-nix-store-refs.sh "${CONAN_HOME}"

      # Only what PatchNixBinary.cmake retargets: the toolchain in the Linux
      # images always references the store. Same condition it uses.
      - name: Check for Nix store references (Linux)
        if: ${{ runner.os == 'Linux' && env.SANITIZERS_ENABLED == 'false' }}
        run: |
-          ./bin/nix/check-nix-store-refs.sh "${BUILD_DIR}/xrpld"
-          ./bin/nix/check-nix-store-refs.sh "${BUILD_DIR}/xrpl_tests"
-          ./bin/nix/check-nix-store-refs.sh "${BUILD_DIR}/xrpld_tests"
+          ./bin/check-nix-store-refs.sh "${BUILD_DIR}/xrpld"
+          ./bin/check-nix-store-refs.sh "${BUILD_DIR}/xrpl_tests"
+          ./bin/check-nix-store-refs.sh "${BUILD_DIR}/xrpld_tests"

      - name: Show ccache statistics
        if: ${{ inputs.ccache_enabled }}
@@ -280,7 +283,7 @@ jobs:

      - name: Upload the binary (Linux)
        if: ${{ env.PACKAGING_ARTIFACTS_ENABLED == 'true' && runner.os == 'Linux' }}
-        uses: actions/upload-artifact@cf430e030ddbb5b0abf93d22962f4752f3646cd9 # v7.0.2
+        uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a # v7.0.1
        with:
          name: xrpld-${{ inputs.config_name }}
          path: ${{ env.BUILD_DIR }}/xrpld
@@ -294,7 +297,7 @@ jobs:

      - name: Upload the validator-keys binary
        if: ${{ env.PACKAGING_ARTIFACTS_ENABLED == 'true' && env.VALIDATOR_KEYS_ENABLED == 'true' }}
-        uses: actions/upload-artifact@cf430e030ddbb5b0abf93d22962f4752f3646cd9 # v7.0.2
+        uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a # v7.0.1
        with:
          name: validator-keys-${{ inputs.config_name }}
          path: |
@@ -305,7 +308,7 @@ jobs:

      - name: Upload the test binary (Linux)
        if: ${{ github.event.repository.visibility == 'public' && runner.os == 'Linux' }}
-        uses: actions/upload-artifact@cf430e030ddbb5b0abf93d22962f4752f3646cd9 # v7.0.2
+        uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a # v7.0.1
        with:
          name: xrpl_tests-${{ inputs.config_name }}
          path: ${{ env.BUILD_DIR }}/xrpl_tests
@@ -321,7 +324,7 @@ jobs:

      - name: Upload server definitions
        if: ${{ github.event.repository.visibility == 'public' && inputs.config_name == 'ubuntu-gcc-debug-amd64-coverage' }}
-        uses: actions/upload-artifact@cf430e030ddbb5b0abf93d22962f4752f3646cd9 # v7.0.2
+        uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a # v7.0.1
        with:
          name: server-definitions
          path: ${{ env.BUILD_DIR }}/server_definitions.json
@@ -34,7 +34,7 @@ jobs:
    needs: [determine-files]
    if: ${{ needs.determine-files.outputs.cpp_changed_files != '' || needs.determine-files.outputs.need_full_run == 'true' }}
    runs-on: ["self-hosted", "Linux", "X64", "heavy"]
-    container: "ghcr.io/xrplf/xrpld/nix-debian:sha-3d526d4"
+    container: "ghcr.io/xrplf/xrpld/nix-debian:sha-060957e"
    permissions:
      contents: read
      issues: write
@@ -43,7 +43,7 @@ jobs:
        uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1

      - name: Prepare runner
-        uses: XRPLF/actions/prepare-runner@8e98d61f011fd2a67f3fde5f0f075662c099a527
+        uses: XRPLF/actions/prepare-runner@b3e255d74d785d053e4903da8ac90983cd7d9e82
        with:
          enable_ccache: false

@@ -73,6 +73,7 @@ jobs:
        with:
          build_nproc: ${{ steps.nproc.outputs.nproc }}
          build_type: ${{ env.BUILD_TYPE }}
+          log_verbosity: verbose

      - name: Configure CMake
        working-directory: ${{ env.BUILD_DIR }}
@@ -109,7 +110,7 @@ jobs:

      - name: Upload clang-tidy output
        if: ${{ github.event.repository.visibility == 'public' && steps.run_clang_tidy.outcome != 'success' }}
-        uses: actions/upload-artifact@cf430e030ddbb5b0abf93d22962f4752f3646cd9 # v7.0.2
+        uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a # v7.0.1
        with:
          path: ${{ env.OUTPUT_FILE }}
          archive: false
@@ -133,7 +134,7 @@ jobs:

      - name: Upload clang-tidy diff output
        if: ${{ github.event.repository.visibility == 'public' && steps.files_changed.outcome != 'success' }}
-        uses: actions/upload-artifact@cf430e030ddbb5b0abf93d22962f4752f3646cd9 # v7.0.2
+        uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a # v7.0.1
        with:
          path: ${{ env.DIFF_FILE }}
          archive: false
@@ -185,7 +186,7 @@ jobs:

      - name: Create issue
        if: ${{ steps.run_clang_tidy.outcome != 'success' && inputs.create_issue_on_failure }}
-        uses: XRPLF/actions/create-issue@b489c21d53928f7dab4874912fec40e4f6c9e85e
+        uses: XRPLF/actions/create-issue@2b8bc36af85b88bca0dd7bfac2e2dc05f94ad712
        with:
          title: "Clang-tidy check failed"
          body_file: ${{ env.ISSUE_FILE }}
@@ -44,7 +44,7 @@ jobs:
      # Every package lands in one directory; the step below picks its own,
      # which keeps this independent of the artifact names.
      - name: Download package artifacts
-        uses: actions/download-artifact@9000827ccba6bdab643e8b6fd33ac0654aef8333 # v8.0.2
+        uses: actions/download-artifact@3e5f45b2cfb9172054b4087a40e8e0b5a5461e7c # v8.0.1
        with:
          pattern: "*-pkg"
          merge-multiple: true
@@ -100,18 +100,18 @@ jobs:
        uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1

      - name: Prepare runner
-        uses: XRPLF/actions/prepare-runner@8e98d61f011fd2a67f3fde5f0f075662c099a527
+        uses: XRPLF/actions/prepare-runner@b3e255d74d785d053e4903da8ac90983cd7d9e82
        with:
          enable_ccache: false

      - name: Download pre-built xrpld binary
-        uses: actions/download-artifact@9000827ccba6bdab643e8b6fd33ac0654aef8333 # v8.0.2
+        uses: actions/download-artifact@3e5f45b2cfb9172054b4087a40e8e0b5a5461e7c # v8.0.1
        with:
          name: ${{ matrix.xrpld_artifact_name }}
          path: ${{ env.BUILD_DIR }}

      - name: Download pre-built validator-keys binary
-        uses: actions/download-artifact@9000827ccba6bdab643e8b6fd33ac0654aef8333 # v8.0.2
+        uses: actions/download-artifact@3e5f45b2cfb9172054b4087a40e8e0b5a5461e7c # v8.0.1
        with:
          name: ${{ matrix.validator_keys_artifact_name }}
          path: ${{ env.BUILD_DIR }}
@@ -151,7 +151,7 @@ jobs:
      # version starts with a digit and a longer name does not, so that one digit
      # is what tells 'xrpld-3.4.1-...' from 'xrpld-assert-3.4.1-...'.
      - name: Upload package artifact
-        uses: actions/upload-artifact@cf430e030ddbb5b0abf93d22962f4752f3646cd9 # v7.0.2
+        uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a # v7.0.1
        with:
          name: ${{ matrix.xrpld_artifact_name }}-pkg
          path: |
@@ -160,7 +160,7 @@ jobs:
          if-no-files-found: error

      - name: Upload debug symbol artifact
-        uses: actions/upload-artifact@cf430e030ddbb5b0abf93d22962f4752f3646cd9 # v7.0.2
+        uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a # v7.0.1
        with:
          name: ${{ matrix.xrpld_artifact_name }}-pkg-debug
          path: |
@@ -225,13 +225,13 @@ jobs:
        uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1

      - name: Prepare runner
-        uses: XRPLF/actions/prepare-runner@8e98d61f011fd2a67f3fde5f0f075662c099a527
+        uses: XRPLF/actions/prepare-runner@b3e255d74d785d053e4903da8ac90983cd7d9e82
        with:
          enable_ccache: false

      # Both artifacts, so the debug symbols are published alongside the package.
      - name: Download package artifacts
-        uses: actions/download-artifact@9000827ccba6bdab643e8b6fd33ac0654aef8333 # v8.0.2
+        uses: actions/download-artifact@3e5f45b2cfb9172054b4087a40e8e0b5a5461e7c # v8.0.1
        with:
          pattern: ${{ matrix.xrpld_artifact_name }}-pkg*
          merge-multiple: true
@@ -287,7 +287,7 @@ jobs:
        run: echo "PUSH=${PUSH}" | tee -a "${GITHUB_ENV}"

      - name: Download package artifacts
-        uses: actions/download-artifact@9000827ccba6bdab643e8b6fd33ac0654aef8333 # v8.0.2
+        uses: actions/download-artifact@3e5f45b2cfb9172054b4087a40e8e0b5a5461e7c # v8.0.1
        with:
          pattern: "*-pkg"
          merge-multiple: true
@@ -295,7 +295,7 @@ jobs:

      - name: Download voidstar binary
        if: ${{ matrix.target == 'voidstar' }}
-        uses: actions/download-artifact@9000827ccba6bdab643e8b6fd33ac0654aef8333 # v8.0.2
+        uses: actions/download-artifact@3e5f45b2cfb9172054b4087a40e8e0b5a5461e7c # v8.0.1
        with:
          name: xrpld-ubuntu-clang-debug-amd64-voidstar
          path: ${{ env.CONTEXT }}
@@ -28,7 +28,7 @@ permissions:
 jobs:
  clippy:
    runs-on: ubuntu-latest
-    container: ghcr.io/xrplf/xrpld/nix-ubuntu:sha-3d526d4
+    container: ghcr.io/xrplf/xrpld/nix-ubuntu:sha-060957e
    steps:
      - name: Checkout repository
        uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1
@@ -41,7 +41,7 @@ jobs:

  coverage:
    runs-on: ubuntu-latest
-    container: ghcr.io/xrplf/xrpld/nix-ubuntu:sha-3d526d4
+    container: ghcr.io/xrplf/xrpld/nix-ubuntu:sha-060957e
    steps:
      - name: Checkout repository
        uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1
@@ -70,7 +70,7 @@ jobs:

  doc:
    runs-on: ubuntu-latest
-    container: ghcr.io/xrplf/xrpld/nix-ubuntu:sha-3d526d4
+    container: ghcr.io/xrplf/xrpld/nix-ubuntu:sha-060957e
    steps:
      - name: Checkout repository
        uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1
@@ -40,7 +40,7 @@ defaults:
 jobs:
  upload:
    runs-on: ubuntu-latest
-    container: ghcr.io/xrplf/xrpld/nix-ubuntu:sha-3d526d4
+    container: ghcr.io/xrplf/xrpld/nix-ubuntu:sha-060957e
    env:
      REMOTE_NAME: ${{ inputs.remote_name }}
      CONAN_LOGIN_USERNAME_XRPLF: ${{ secrets.remote_username }}
@@ -50,7 +50,7 @@ jobs:
        uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1

      - name: Prepare runner
-        uses: XRPLF/actions/prepare-runner@8e98d61f011fd2a67f3fde5f0f075662c099a527
+        uses: XRPLF/actions/prepare-runner@b3e255d74d785d053e4903da8ac90983cd7d9e82
        with:
          enable_ccache: false

@@ -68,7 +68,7 @@ jobs:
        uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1

      - name: Prepare runner
-        uses: XRPLF/actions/prepare-runner@8e98d61f011fd2a67f3fde5f0f075662c099a527
+        uses: XRPLF/actions/prepare-runner@b3e255d74d785d053e4903da8ac90983cd7d9e82
        with:
          enable_ccache: false

@@ -108,11 +108,14 @@ jobs:
          build_nproc: ${{ steps.nproc.outputs.nproc }}
          build_type: ${{ matrix.build_type }}
          force_build: ${{ github.event_name == 'schedule' || github.event.inputs.force_source_build == 'true' }}
+          # Set the verbosity to "quiet" for Windows to avoid an excessive
+          # amount of logs. For other OSes, the "verbose" logs are more useful.
+          log_verbosity: ${{ runner.os == 'Windows' && 'quiet' || 'verbose' }}
          sanitizers: ${{ matrix.sanitizers }}

      - name: Check the Conan cache for Nix store references (Nix toolchain)
        if: ${{ matrix.toolchain == 'nix' }}
-        run: ./bin/nix/check-nix-store-refs.sh "${CONAN_HOME}"
+        run: ./bin/check-nix-store-refs.sh "${CONAN_HOME}"

      - name: Log into Conan remote
        if: ${{ github.repository == 'XRPLF/rippled' && (github.event_name == 'push' || github.event_name == 'workflow_dispatch') }}
@@ -60,7 +60,7 @@ repos:
        types_or: [c++, c]

  - repo: https://github.com/pre-commit/mirrors-clang-format
-    rev: a9a8a861f30ed207ead7d5a3b7e8032283ba5da7 # frozen: v23.1.2
+    rev: e2b496dc2bd8340c2524cb9a2d2a943cde1bb6df # frozen: v23.1.1
    hooks:
      - id: clang-format
        args: [--style=file]
@@ -82,10 +82,9 @@ repos:
        files: ^crates/.*\.rs$

  - repo: https://github.com/BlankSpruce/gersemi-pre-commit
-    rev: 28010ddd6016e1a0f7bd232acb6536ef996ae897 # frozen: 0.29.2
+    rev: f1c4833f8cf23c6d952673abc73525411a5719e8 # frozen: 0.29.1
    hooks:
      - id: gersemi
-        args: [-i, --warnings-as-errors]

  - repo: https://github.com/rbubley/mirrors-prettier
    rev: ef4a397f916211b4a39ccf9d3d9cbb6562157251 # frozen: v3.9.9
@@ -96,19 +95,19 @@ repos:
  # Scoped to package/: the rest of the repo's Python has pre-existing findings,
  # so widening these is its own change.
  - repo: https://github.com/astral-sh/ruff-pre-commit
-    rev: f12be1ebaa5351c1fc76472de98db2c3446c8253 # frozen: v0.16.10
+    rev: a56c0b927e6465d37cae3e97d35d4d18ab2b96cd # frozen: v0.16.9
    hooks:
      - id: ruff-check
        args: [--fix]
        files: ^package/.*\.py$

  - repo: https://github.com/psf/black-pre-commit-mirror
-    rev: 96ae3e5802f3fe2d551e703e18f0a367d1a81ac2 # frozen: 26.10.0
+    rev: 4160603246a6b365d4a2af661c6d71b0a0f50478 # frozen: 26.5.1
    hooks:
      - id: black

  - repo: https://github.com/pre-commit/mirrors-mypy
-    rev: 2834ec6639549dd6796205c8f011dedcd587288b # frozen: v2.4.0
+    rev: 7ff8d35ae36a7d2b968f2f90b4c723e292e594ee # frozen: v2.3.1
    hooks:
      - id: mypy
        args: [--strict]
@@ -14,8 +14,6 @@ For a log of breaking changes, see the **API Version [number]** headings. In gen

 API version 3 is currently a beta API. It requires enabling `[beta_rpc_api]` in the xrpld configuration to use. See [API-VERSION-3.md](API-VERSION-3.md) for the full list of changes in API version 3.

-From this version a message a subscription pushes over WebSocket is sent as a JSON-RPC 2.0 notification: `method` names the event where `type` did, the message's content sits under `params`, and there is no `id`. A `path_find` update is sent the same way. A subscriber named by a `url` receives the legacy `event` call at every version. See [API-VERSION-3.md](API-VERSION-3.md).
-
 ## API Version 2

 API version 2 is available in `xrpld` version 2.0.0 and later. See [API-VERSION-2.md](API-VERSION-2.md) for the full list of changes in API version 2.
@@ -24,55 +22,19 @@ API version 2 is available in `xrpld` version 2.0.0 and later. See [API-VERSION-

 This version is supported by all `xrpld` versions. For WebSocket and HTTP JSON-RPC requests, it is currently the default API version used when no `api_version` is specified.

-## Unreleased
-
-### Breaking changes
-
- The `ripplerpc` request field, which selects the shape of the JSON-RPC reply envelope, is now validated, and a value that is not exactly `"1.0"`, `"2.0"` or `"3.0"` is rejected. Two things it does not reach: a WebSocket session, which reads the field only to echo it back, and API version 3, where the field selects nothing and so is neither read nor checked, `ripplerpc: "banana"` and no `ripplerpc` at all being served alike. A request sending `"2"`, `" 2.0"`, `"2.00"`, `"02.0"`, `"2.0.0"`, `"10.0"` or any other text, such as `"abc"` or `"x2"`, gets a working reply today. After this ships, such a request sent alone gets HTTP 400 with `ripplerpc is not a supported version`, charged as a malformed request, and such an entry of a `"method": "batch"` body gets that error in its own reply while the batch answers 200. A client that spells the version loosely must therefore be corrected to one of the three exact values, or omit the field. Previously the value was compared as a string, which both accepted values that name no version and ordered multi-digit versions incorrectly: `"abc"` and `"x2"` sorted above `"3.0"` and so selected version 3, and `"10.0"` sorted below `"2.0"` and so selected version 1. Requests that send one of the three supported values, or omit the field, are unaffected.
- `subscribe`: every subscription a connection holds must name the same `api_version`. The first `subscribe` establishes it, and a later one naming a different version is refused with `apiVersionConflict`, registering nothing; the message names the version the connection's subscriptions are served at. The HTTP status is 400 where the envelope derives it from the error, with `ripplerpc: "3.0"` or at API version 3, and 200 with `ripplerpc` `"1.0"` or `"2.0"`, as for every error; a WebSocket frame carries no status. A `subscribe` refused for another reason, a stream name the server does not have for one, still fixes that version, since the registrations it makes are spread through the handler. A connection that subscribed one stream at `api_version` 2 and another at 1, in two calls, was previously served both at version 1, the later call's, so the first stream's shape changed under it. The version is fixed for the connection's life: a client that wants another version opens another connection. On the webhook path the subscriber is keyed on its `url` alone, so the refusal reaches a second admin because of a first admin's version, and the message names the version the url's subscriptions are served at, and no remedy: releasing a url means unsubscribing its streams, which would remove that first admin's subscription, and the server cannot tell the two callers apart. Naming the url alone does not release it, a subscriber a stream map still holds not being evicted. This reaches every API version, since version 1 and version 2 content is not interchangeable either: version 2 renames `transaction` to `tx_json` and hoists `hash`, so a version 1 client handed version 2 content finds no `transaction` member.
- `batch`: An entry of a `"method": "batch"` request that names `api_version` both inside `params` and at its own top level is now served at the version inside `params`. It was previously served at the top-level version whenever the one inside `params` resolved to version 1, so an entry asking for version 1 explicitly and something else at the top level changes which version answers it. A lone request can name both values too, and for it the version inside `params` has always won; reversing that precedence was confined to a batch entry.
- `batch`: every entry of one batch request is now served at the same API version. An entry may name the version the body names, or name none and inherit it, and a body whose entries name two different versions is refused whole with `Batch entries name different versions` at HTTP 400, charged as a malformed request. **This is visible to a version 1 or 2 client:** a `"method": "batch"` body could mix versions across its entries, each entry being dispatched at its own, and the difference is observable because the reply shape changes with the version. Three bodies change. One whose entries name two different versions is now refused instead of answered. One where some entries name a version and others name none now serves all of them at the named version, where an entry naming none was served at version 1; the same applies to a version named on the request's own top level, which the entries now inherit. One whose top level names a version and whose entries name a different one is now refused, where the top-level value was ignored and each entry was served at the version it named. A body whose top level names no version or the same version as its entries, and whose entries agree with one another or name none at all, is unaffected. A `"method": "batch"` body naming at its own top level a version the server does not serve is refused with `invalid_API_version` at HTTP 400, where it was previously served at version 1 as if it had named none. No test pinned the mixing behavior and the line that produced it was a fallback about where the field sits rather than a decision that entries may differ, so this is a correction rather than a removal of a feature: a per-entry version made the entry cap above unanswerable, since a cap can only ask about a body that has one version.
-
-### Additions
-
- A JSON-RPC (HTTP) request may send its parameters as an object, `"params": {"account": "r..."}`, as well as the array of one object that was already accepted. This reaches every API version: such a request was previously rejected with HTTP 400 and `params unparsable`, and is now served. A request that already sends the array form is unaffected. An entry of a `"method": "batch"` request that carries a by-name `params` object is read for `api_version` and credentials from that object. Previously such an entry had its `api_version` read from its top level and its credentials ignored, since credentials were read only from an array-form `params`.
- A JSON-RPC (HTTP) request may name its `api_version` at its own top level, beside `method` and `id`, as well as inside `params`. The member is an XRPL extension, which the JSON-RPC 2.0 specification does not define; placing it beside the specification's own members is what lets a version 3 client send a specification-shaped request. This is honored from API version 3 up, so it needs `[beta_rpc_api]`; a version below that named at the top level is ignored, exactly as it always has been. Previously only an entry of a `"method": "batch"` request had its own top level read; a WebSocket request, which is one flat object, has always been read there.
-
-### Bugfixes
-
- A request echoed back in an error reply now has every credential-bearing field masked: `admin_password`, `admin_user`, `passphrase`, `password`, `secret`, `seed`, `seed_hex`, `url_password`, `url_username` and `username`. Nesting no longer matters, so a credential inside `params` is masked too. The same masking is applied to every request and reply written to the log, and it covers six further names that only a reply carries: `master_key`, `master_seed`, `master_seed_hex`, `validation_key`, `validation_private_key` and `validation_seed`, which is how `wallet_propose` and `validation_create` used to write a live private key to the log. A request or reply written to the log is truncated at 10,000 characters. An `[rpc_startup]` line that does not parse is reported by its position in the section and the parser's message, never by its text, and no command runs for it.
- The command line client no longer prints a credential the operator did not type. A failing command echoes the request it built under `request_sent`, which carries the `admin_password` the client copies out of `[port_rpc]` in the config, so `./xrpld account_info rBogus` printed that password to stdout and into any captured output. `request_sent` is now masked. The `rpc` member beside it, which echoes the arguments as they were typed, is unchanged. The command line client also no longer writes an unparsed `json` or `ripple_path_find` argument to its trace log before parsing it, where a `secret` inside that argument could not be masked; it logs the parsed request instead, masked. The reply it receives is logged the same way, parsed and masked, where the raw body was written before, a `validation_create` answer included.
- A WebSocket frame that does not parse, or exceeds the request size limit, is answered `{"type": "error", "error": "jsonInvalid", "size": <bytes>}`. The frame's body is reported by size rather than echoed back in a `value` member, since a body that does not parse has no fields to mask. A client that read `value` gets `size` instead.
- Four error codes that named no HTTP status of their own, and so answered 200 on a reply reporting an error, now name one: `actMalformed`, `alreadyMultisig` and `alreadySingleSig` answer 400, and `actNotFound` answers 404. **Only API version 3 reports these four.** A request sending `ripplerpc: "3.0"` at API version 1 or 2 still receives 200 for all four, as it always has, so `account_info` on a malformed account or one the ledger does not hold answers 200 for those clients exactly as before.
- `submit`, `simulate`, `transaction_entry`, `ledger_entry` and `ledger_accept`: Errors from these methods now include `error_code` and `error_message` alongside the `error` token, as every other method already did. Each error now answers the status its code names: 400 for a malformed request, 404 for `transactionNotFound`, 500 for an internal failure, and 501 for `notYetImplemented` and `notStandAlone`. That status change reaches only a request sending `ripplerpc: "3.0"`, or API version 3, which are the envelopes that derive the status from the error. With `ripplerpc` `"1.0"` the status stays 200 and the two new members appear beside `error`; with `"2.0"` the status stays 200, `error_code` appears, and the `code` and `message` members carry the code and the message rather than null, since that envelope copies them from `error_code` and `error_message` and drops `error_message`. From API version 3 the twenty-one `malformed*` tokens `ledger_entry` names each report the code they own instead of 31.
- A reply reporting HTTP 402 or 502 now carries a status line. Those two statuses named no case in the switch that writes one, so such a reply began with a header instead and did not parse as an HTTP response at all. Both are reachable at any API version with `ripplerpc: "3.0"`, which derives the status from the error code: 402 through `highFee` from `sign`, `sign_for` or `submit` with a low `fee_mult_max`, and 502 through `dbDeserialization` from `tx`. The eleven statuses that already named a case report the same phrase they always have.
- An error reply to a request sending `ripplerpc: "2.0"` or `"3.0"` no longer carries a stray `"error_message": null` beside the error it reports. The member appeared only when the `Server` log partition was set to debug or lower, because the log statement read `error_message` after the reply had renamed it to `message`, and reading it put it back as null. So the reply a client received depended on the server's log level, and the log line itself printed an empty message. Both are fixed.
- A body the server rejects before it reads a request out of it now says what was wrong. A body over the size limit answers `Request is too large`, and a body that parses to `{}` or `null` answers `Request is empty`; both previously answered `Unable to parse request: ` with nothing after the colon, the parser having recorded no error for them. Any other body that is not an array does not parse, which includes one that is only whitespace and one whose top-level value is a string, number or boolean; it answers `Unable to parse request: ` followed by the parser's own reason, as it did before. The status is 400 for all three, as before, and this reaches every API version. A top-level array, the one other body nothing can read a request from, is answered with a JSON error object instead, which the entry on top-level arrays below describes.
- `batch`: An entry that is not identified through a secure gateway no longer clears the connection's `X-User` and forwarded-for values for the entries after it, so every entry of one body reports the role and username it would have reported on its own.
- `batch`: Every entry of a `"method": "batch"` request is now charged against the sender's resource allowance, including one rejected before it reaches a handler, and the request stops at the first entry the connection is too loaded to serve. A reply array can therefore be shorter than the request array, and its last element depends on where the batch stopped. An entry that is not a JSON object, or names an API version the server does not serve, is charged before its role is known and answered with its own rejection; if that charge took the connection over the drop threshold, the batch stops there and that rejection is the last answer. Every other entry met over the threshold is answered `Server is overloaded` and nothing follows. A client should read any reply array shorter than its request as a connection over the drop threshold, whatever the last element says. A well-behaved client is unaffected; one that sends thousands of entries in a single body no longer gets every one of them answered.
- A body the server rejects before it reads a request out of it is now charged against the sender's resource allowance, as a malformed request already was. Ten conditions were free: a body over the size limit, one that does not parse, one carrying no document, one that is neither a JSON object nor an array, a `"method": "batch"` naming no entry array, and six ways a top-level array is not a batch this server can serve. The answer to the first five is unchanged; an array's answer changes as the entry on top-level arrays below describes, the one generic rejection it received having become six. A well-behaved client is unaffected; one that repeats such a body exhausts its allowance and is refused the next request a handler would have served. A WebSocket frame that does not parse, or exceeds the request size limit, is charged the same way, and a connection that sends only such frames is closed once it crosses the drop threshold.
- `subscribe`, `path_find`: A subscription is now served at the API version its own `subscribe` call named, rather than at the version of the most recent `subscribe` or `path_find` on the same connection. A `path_find` no longer changes the version anything that connection has subscribed is served at. An ordinary request is unaffected: it names its own version and is answered at it, so a connection subscribed at `api_version` 3 still calls `account_info` at version 1 and reads the version 1 shape.
- `batch`: A batch request served at API version 3 or above is now refused when it holds more than 100 entries, before any entry is dispatched. The cap reads the one version the body is served at, so no ordering of the entries evades it. A body served below version 3 is unaffected and stays bounded only by the request size limit.
- A top-level JSON array, which is the JSON-RPC 2.0 batch form accepted from API version 3, is now rejected with a JSON-RPC error object rather than a plain text body when it cannot be served: when it is empty, holds more than 100 entries, holds nothing that could be a request, names no API version of 3 or above that the server serves, or names two different versions the server serves. An array naming no served version of 3 or above is refused for requiring version 3, or, where every version it names is one the server does not serve, with `invalid_API_version`. One entry naming a version the server does not serve does not reject an array in which another entry names a served version: it is answered on its own, inside the reply array, with the same `-32606` code. This reaches every API version, since no earlier version accepts an array at all: a client that sent one received HTTP 400 with `Unable to parse request: ` and now receives the same status with a JSON body. The message also names the version rather than the shape where the version is what is wrong, which is what a client with `[beta_rpc_api]` disabled receives.
-
 ## XRP Ledger server version 3.5.0

 Version 3.5.0 is not yet released.

 ### Additions in 3.5.0

- `subscribe`, `unsubscribe`: Added an optional `mpt_issuances` request field, an array of MPT issuance IDs (hex strings). Subscribers receive the same `transaction` message as the `transactions` stream for each validated transaction whose metadata affects a subscribed issuance. MPT issuance subscriptions count toward the per-connection subscription limit. An empty array, a non-array value, or an invalid ID returns `invalidParams`. ([#5671](https://github.com/XRPLF/rippled/pull/5671))
- `ledger_entry`: Add full support for checks, NFT offers, payment channels, and signer lists. ([#6319](https://github.com/XRPLF/rippled/pull/6319))
+- `subscribe`, `unsubscribe`: Added an optional `mpt_issuances` request field, an array of MPT issuance IDs (hex strings). Subscribers receive a message with `type` `mptTransaction` for each validated transaction whose metadata affects a subscribed issuance; the message has the same fields as the `transactions` stream. MPT issuance subscriptions count toward the per-connection subscription limit. An empty array, a non-array value, or an invalid ID returns `invalidParams`. ([#5671](https://github.com/XRPLF/rippled/pull/5671))

 ### Bugfixes in 3.5.0

 - `channel_authorize`: The `channel_id` field now returns an `invalidParams` error if the value is not a string. [#7582](https://github.com/XRPLF/rippled/pull/7582)
 - `channel_verify`: The `channel_id` and `signature` fields now return an `invalidParams` error if the value is not a string. [#7582](https://github.com/XRPLF/rippled/pull/7582)

-### Bugfixes in 3.5.0
-
- `feature`: The admin-only `vetoed` field now returns `invalidParams` unless its value is a boolean. [#7583](https://github.com/XRPLF/rippled/pull/7583)
-
 ## XRP Ledger server version 3.4.0

 Version 3.4.0 is not yet released. These changes are available in the 3.4.0 beta releases.
@@ -93,7 +55,6 @@ Version 3.4.0 is not yet released. These changes are available in the 3.4.0 beta
 - `account_lines`: The `peer` field now returns an error if the value is not a string. [#7728](https://github.com/XRPLF/rippled/pull/7728)
 - `ledger`: `delivered_amount` is now included in the metadata of successful `AccountDelete` transactions when transactions are expanded (`expand`, or admin-only `full`). Previously it was only added for `Payment` and `CheckCash`, which made `ledger` inconsistent with `tx` and `account_tx`. [#5706](https://github.com/XRPLF/rippled/pull/5706)
 - `noripple_check`: The `transactions` field is no longer included in error responses; it is still returned (possibly as an empty array) whenever `transactions` is `true` and the request succeeds. A malformed `account` is now rejected before the ledger is looked up, so that error response no longer carries the `ledger_hash`, `ledger_index`, and `validated` fields ([#6303](https://github.com/XRPLF/rippled/pull/6303)).
- `transaction_entry`: An object or an array in `tx_hash` now returns `malformedRequest`, like any other value that is not a hex hash, instead of an `internal` error.

 ## XRP Ledger server version 3.3.0

@@ -10,7 +10,7 @@
 # alone; the scripts in a Conan cache are all git hook samples and autotools
 # scratch, 36 false positives to 0 real.
 #
-# Usage: bin/nix/check-nix-store-refs.sh <path>
+# Usage: bin/check-nix-store-refs.sh <path>

 set -euo pipefail

@@ -20,11 +20,9 @@
 # development setups, but not in the macOS CI environment. They are checked
 # everywhere except when running in CI on macOS.
 #
-# Tools that Nix also exposes under a version-suffixed name
-# (`clang-tidy-<v>`, `g++-<v>`, ...) are probed under both names:
-# a suffixed name can break while the plain one still works
-# (see mkVersionedToolLinks in nix/packages.nix).
-# The suffix is the major version of the plain `clang` / `gcc` on PATH.
+# Tools that Nix also exposes under a version-suffixed name (`clang-tidy-22`,
+# `g++-15`, ...) are probed under both names: a suffixed name can break while
+# the plain one still works (see mkVersionedToolLinks in nix/packages.nix).
 #
 # Tools scoped to a single dev shell rather than to commonPackages are checked
 # only in that shell, keyed off XRPL_DEVSHELL.
@@ -36,16 +34,9 @@

 set -uo pipefail

-# major_version <compiler>
-# Major version of a compiler on PATH, or "unknown" when it isn't there.
-major_version() {
-    local version
-    version="$("$1" -dumpversion 2>/dev/null)" || version=""
-    version="${version%%.*}"
-    printf '%s' "${version:-unknown}"
-}
-
-llvm_version="$(major_version clang)"
+# Version suffixes of the Nix tool links, tracking nix/packages.nix.
+gcc_version=15
+llvm_version=22

 missing=()
 checked=0
@@ -121,7 +112,6 @@ if [ "${os}" = "linux" ] || [ "${os}" = "macos" ]; then
    check ClangBuildAnalyzer
    check curl
    check file
-    check jq
    check less
    check make
    # net-tools netstat reports "net-tools X.Y"; macOS ships BSD netstat with no
@@ -190,7 +180,6 @@ fi
 if [ "${os}" = "linux" ]; then
    echo
    echo "GCC toolchain:"
-    gcc_version="$(major_version gcc)"
    check gcc
    check "gcc-${gcc_version}"
    check g++
@@ -1,65 +0,0 @@
-#!/usr/bin/env bash
-# Fail if a binary of a build-context Conan package loads anything from the Nix
-# store other than glibc, or cannot resolve a library at all.
-#
-# Only binaries linked by the Nix toolchain are checked, i.e. those recording a
-# store path as their interpreter or RUNPATH. Prebuilt upstream binaries (such
-# as the ones the cmake package ships) use the system loader instead.
-#
-# Build-context packages provide the tools that run during the build (protoc,
-# grpc_cpp_plugin, ...). Their package ID does not change when a Nix toolchain
-# update moves the GCC runtime to a new store path, so a cached binary has to
-# get by with the pinned glibc alone. See docs/build/nix.md.
-#
-# Usage: bin/nix/check-build-context-runtime.sh <graph.json>
-#   <graph.json> is the output of `conan install --format=json`.
-
-set -euo pipefail
-
-if [ "$#" -ne 1 ]; then
-    echo "usage: $0 <graph.json>" >&2
-    exit 2
-fi
-
-if [ "$(uname -s)" != "Linux" ]; then
-    echo "$0: Linux only" >&2
-    exit 2
-fi
-
-folders="$(jq -r '.graph.nodes[] | select(.context == "build" and .package_folder) | .package_folder' "$1" | sort -u)"
-
-checked=0
-failed=0
-
-while IFS= read -r file; do
-    case "$(file -b "${file}")" in
-        ELF*) ;;
-        *) continue ;;
-    esac
-    [[ "$(readelf -ldW "${file}")" == */nix/store/* ]] || continue
-    checked=$((checked + 1))
-
-    # `ldd` lists the interpreter and every library as the loader resolves them.
-    if deps="$(ldd "${file}" 2>&1)"; then
-        bad="$(printf '%s\n' "${deps}" |
-            grep -E 'not found|/nix/store/' |
-            grep -vE '/nix/store/[^/]+-glibc-[^/]+/' || true)"
-    else
-        case "${deps}" in
-            *"not a dynamic executable"*) continue ;;
-        esac
-        bad="${deps}"
-    fi
-    if [ -n "${bad}" ]; then
-        failed=$((failed + 1))
-        echo "::error file=${file}::loads a library from the Nix store other than glibc"
-        echo "${file}"
-        echo "${bad}" | sed 's/^/    /'
-    fi
-done < <(
-    # shellcheck disable=SC2086 # one folder per line, no spaces in Conan paths
-    [ -z "${folders}" ] || find ${folders} -type f \( -perm -u+x -o -name '*.so*' \)
-)
-
-echo "Build-context packages: checked ${checked} binaries, ${failed} failed."
-[ "${failed}" -eq 0 ]
@@ -23,7 +23,7 @@ apt-get clean
 rm -rf /var/lib/apt/lists/*
 EOF

-ARG PRE_COMMIT_VERSION=4.6.2
+ARG PRE_COMMIT_VERSION=4.6.0
 RUN pip install --no-cache --break-system-packages \
    pre-commit==${PRE_COMMIT_VERSION}

@@ -26,6 +26,8 @@ import sys
 import tempfile
 from pathlib import Path

+CLANG_TIDY_VERSION = 22
+
 # Extensions run-clang-tidy can analyse: `.cpp` translation units and, thanks to
 # the `verify_headers` build option, `.h`/`.hpp` headers (each has its own
 # compile_commands.json entry). `.ipp` fragments have no entry and are skipped.
@@ -37,21 +39,8 @@ TIDY_EXTENSIONS = {".cpp", ".h", ".hpp"}
 FILEPATH_RE = re.compile(r"^(\s*(?:-\s+)?FilePath:\s*)'((?:[^']|'')*)'\s*$")


-def clang_tidy_major() -> str | None:
-    """Major version of the `clang-tidy` on PATH, which run-clang-tidy invokes."""
-    if not (clang_tidy := shutil.which("clang-tidy")):
-        return None
-    output = subprocess.run(
-        [clang_tidy, "--version"], capture_output=True, text=True
-    ).stdout
-    m = re.search(r"LLVM version (\d+)", output)
-    return m.group(1) if m else None
-
-
-def find_tool(name: str, version: str | None) -> str | None:
-    """Prefer `<name>-<version>`, so a host tool of another version can't win."""
-    candidates = ([f"{name}-{version}"] if version else []) + [name]
-    for candidate in candidates:
+def find_tool(name: str) -> str | None:
+    for candidate in (f"{name}-{CLANG_TIDY_VERSION}", name):
        if path := shutil.which(candidate):
            return path
    return None
@@ -114,9 +103,8 @@ def main():
    if not files:
        return 0

-    version = clang_tidy_major()
-    run_clang_tidy = find_tool("run-clang-tidy", version)
-    clang_apply_replacements = find_tool("clang-apply-replacements", version)
+    run_clang_tidy = find_tool("run-clang-tidy")
+    clang_apply_replacements = find_tool("clang-apply-replacements")
    missing = [
        name
        for name, path in (
@@ -126,10 +114,9 @@ def main():
        if not path
    ]
    if missing:
-        tried = f" (tried the '-{version}' suffix too)" if version else ""
        print(
            f"clang-tidy check failed: TIDY is enabled but {' and '.join(missing)} "
-            f"was not found in PATH{tried}.",
+            f"was not found in PATH (tried the '-{CLANG_TIDY_VERSION}' suffix too).",
            file=sys.stderr,
        )
        return 1
@@ -17,7 +17,7 @@
   (runtime libraries resolved through the rpath) are skipped too.
   Everywhere else `patch_nix_binary` is a no-op.

-   The default loader is resolved by bin/nix/default-loader-path.sh.
+   The default loader is resolved by bin/default-loader-path.sh.
 #]===================================================================]

 include_guard(GLOBAL)
@@ -25,7 +25,7 @@ include_guard(GLOBAL)
 include(CompilationEnv)

 # Resolves the system default ELF loader path for the current architecture.
-set(_loader_path_script "${CMAKE_SOURCE_DIR}/bin/nix/default-loader-path.sh")
+set(_loader_path_script "${CMAKE_SOURCE_DIR}/bin/default-loader-path.sh")

 if(
    is_linux
@@ -292,14 +292,6 @@ if(xrpld)
    )
    target_sources(xrpld PRIVATE ${sources})

-    rpcspec_generate_instantiations(
-        OUT_VAR rpcspec_instantiations
-        VALUE_TYPE "::json::Value"
-        VIEW_HEADER "xrpld/rpc/detail/JsonObjectView.hpp"
-        HANDLERS book_changes ledger transaction_entry
-    )
-    target_sources(xrpld PRIVATE ${rpcspec_instantiations})
-
    if(tests)
        file(
            GLOB_RECURSE sources
@@ -330,7 +322,6 @@ if(xrpld)
        # antithesis_instrumentation.h, which is not exported as INTERFACE
        target_include_directories(
            xrpld
-            SYSTEM
            PRIVATE ${CMAKE_SOURCE_DIR}/external/antithesis-sdk
        )
    endif()
@@ -77,24 +77,19 @@ if(is_clang)
    message(STATUS "  Ignorelist: ${ignorelist_path}")
 endif()

-# Define the SANITIZERS macro for BuildInfo.cpp, plus one of XRPL_ASAN,
-# XRPL_TSAN and XRPL_UBSAN per active sanitizer, so that code can test for a
-# specific one with #ifdef instead of parsing the dot-joined SANITIZERS string.
+# Define SANITIZERS macro for BuildInfo.cpp
 set(sanitizers_list)
 if(SANITIZERS MATCHES "address")
    set(enable_asan ON)
    list(APPEND sanitizers_list "ASAN")
-    target_compile_definitions(common INTERFACE XRPL_ASAN)
 endif()
 if(SANITIZERS MATCHES "thread")
    set(enable_tsan ON)
    list(APPEND sanitizers_list "TSAN")
-    target_compile_definitions(common INTERFACE XRPL_TSAN)
 endif()
 if(SANITIZERS MATCHES "undefinedbehavior")
    set(enable_ubsan ON)
    list(APPEND sanitizers_list "UBSAN")
-    target_compile_definitions(common INTERFACE XRPL_UBSAN)
 endif()

 if(sanitizers_list)
@@ -31,7 +31,7 @@ namespace xrpl::ledger_entries {
 // builder's STObject and the wrapper's SLE.
 TEST(${name}Tests, BuilderSettersRoundTrip)
 {
-    UInt256 const index{1u};
+    constexpr UInt256 index{1};

 % for field in fields:
    auto const ${field["paramName"]}Value = ${canonical_expr(field)};
@@ -85,7 +85,7 @@ TEST(${name}Tests, BuilderSettersRoundTrip)
 // from that SLE, build a new wrapper, and verify all fields (and validate()).
 TEST(${name}Tests, BuilderFromSleRoundTrip)
 {
-    UInt256 const index{2u};
+    constexpr UInt256 index{2};

 % for field in fields:
    auto const ${field["paramName"]}Value = ${canonical_expr(field)};
@@ -146,7 +146,7 @@ TEST(${name}Tests, BuilderFromSleRoundTrip)
 // 3) Verify wrapper throws when constructed from wrong ledger entry type.
 TEST(${name}Tests, WrapperThrowsOnWrongEntryType)
 {
-    UInt256 const index{3u};
+    constexpr UInt256 index{3};

    // Build a valid ledger entry of a different type
    // Ticket requires: Account, OwnerNode, TicketSequence, PreviousTxnID, PreviousTxnLgrSeq
@@ -177,7 +177,7 @@ TEST(${name}Tests, WrapperThrowsOnWrongEntryType)
 // 4) Verify builder throws when constructed from wrong ledger entry type.
 TEST(${name}Tests, BuilderThrowsOnWrongEntryType)
 {
-    UInt256 const index{4u};
+    constexpr UInt256 index{4};

    // Build a valid ledger entry of a different type
 % if wrong_le_include == "Ticket":
@@ -207,7 +207,7 @@ TEST(${name}Tests, BuilderThrowsOnWrongEntryType)
 // 5) Build with only required fields and verify optional fields return nullopt.
 TEST(${name}Tests, OptionalFieldsReturnNullopt)
 {
-    UInt256 const index{3u};
+    constexpr UInt256 index{3};

 % for field in required_fields:
    auto const ${field["paramName"]}Value = ${canonical_expr(field)};
@@ -3,7 +3,7 @@
    "requires": [
        "zlib/1.3.2#1cb806da49011867778ffb6ac7190fcb%1782392402.122708",
        "xxhash/0.8.3#681d36a0a6111fc56e5e45ea182c19cc%1782392402.420688",
-        "xrpl-rpc-spec/0.1.21#d536f87a2ae7d313452746cfa3ac4404%1790869005.122975",
+        "xrpl-rpc-spec/0.1.19#870b2d3abcfbbf13b61c2d3c69495060%1790348286.187549",
        "sqlite3/3.53.0#324ada52333108388a9a6108bfa96734%1782392403.185447",
        "soci/4.0.3#e726491a03468795453f7c83fc924a96%1782392402.679521",
        "snappy/1.1.10#968fef506ff261592ec30c574d4a7809%1782307151.633168",
@@ -20,7 +20,7 @@
        "libarchive/3.8.7#c446109bd1f1d8ba7936c94189bc50e6%1782392403.066892",
        "jemalloc/5.3.1#1fc58d55316041f10fbc1e8a2eae632a%1776700028.228",
        "gtest/1.17.0#5224b3b3ff3b4ce1133cbdd27d53ee7d%1782392402.791979",
-        "grpc/1.81.1#aaa93ab6cda2f2baa6a84490582c8adf%1791284951.826256",
+        "grpc/1.81.1#b87796a4269034856cbc1a2522db16eb%1788275071.530512",
        "fast_float/8.2.10#f6f28d6bb22112078e7dbda611caf681%1785888854.601666",
        "ed25519/2015.03#ae761bdc52730a843f0809bdf6c1b1f6%1782307148.15562",
        "date/3.0.4#862e11e80030356b53c2c38599ceb32b%1782392402.538492",
@@ -34,15 +34,11 @@
    "build_requires": [
        "zlib/1.3.2#1cb806da49011867778ffb6ac7190fcb%1782392402.122708",
        "strawberryperl/5.32.1.1#8d114504d172cfea8ea1662d09b6333e%1782395692.540639",
-        "re2/20251105#8579cfd0bda4daf0683f9e3898f964b4%1782392402.431897",
        "protobuf/6.33.5#ff253ead763bd8d9904a52979cd21e81%1782392410.233933",
-        "openssl/3.6.3#f806de8933e3bf6f01016c6a888cee2e%1783945160.863288",
        "nasm/2.16.01#31e26f2ee3c4346ecd347911bd126904%1782395690.33162",
        "msys2/cci.latest#d22fe7b2808f5fd34d0a7923ace9c54f%1770657326.649",
        "m4/1.4.19#1727f439cf74e83826ec96d0b4904eee%1784541921.659",
-        "grpc/1.81.1#aaa93ab6cda2f2baa6a84490582c8adf%1791284951.826256",
        "cmake/4.3.3#840cf00ea09777e05c2050a50a82c722%1782392418.696091",
-        "c-ares/1.34.6#545240bb1c40e2cacd4362d6b8967650%1782392402.681654",
        "b2/5.4.2#ffd6084a119587e70f11cd45d1a386e2%1782392402.624226",
        "automake/1.16.5#b91b7c384c3deaa9d535be02da14d04f%1755524470.56",
        "autoconf/2.71#51077f068e61700d65bb05541ea1e4b0%1731054366.86",
@@ -3,13 +3,5 @@
 core:non_interactive=True
 core.download:parallel={{ os.cpu_count() }}
 core.upload:parallel={{ os.cpu_count() }}
-# Fall back to Conan Center's source backups when a recipe's upstream URL is down
-# (e.g. the GNU FTP mirrors), see
-# https://github.com/conan-io/conan-center-index/issues/28147#issuecomment-3183544772
-# The backups are only tried once every upstream URL has used up its retries,
-# so keep the retries low.
-core.sources:download_urls=["origin", "https://c3i.jfrog.io/artifactory/conan-center-backup-sources"]
-tools.files.download:retry=1
+tools.files.download:retry=5
 tools.files.download:retry_wait=10
-# Fail fast on unreachable hosts (default connect timeout is 30s), keep the 60s read timeout.
-core.net.http:timeout=(5, 60)
@@ -29,9 +29,9 @@ os.version={{ min_macos_version }}
 [conf]
 {# The Boost recipe builds with b2, which doesn't use Conan's toolchain files. #}
 {# Instead it hand-rolls the compiler for user-config.jam, #}
-{# and its fallback probes a version-suffixed binary (e.g. `g++-<major>`) before plain `g++`. #}
-{# Inside the Nix shell the wrapper only provides `g++`/`gcc` (no `-<major>` suffix), #}
-{# so on a host that also has a system `g++-<major>` the probe escapes Nix #}
+{# and its fallback probes a version-suffixed binary (e.g. `g++-15`) before plain `g++`. #}
+{# Inside the Nix shell the wrapper only provides `g++`/`gcc` (no `-15` suffix), #}
+{# so on a host that also has a system `g++-15` the probe escapes Nix #}
 {# and picks the system compiler, which is mismatched with the Nix libraries #}
 {# and breaks the build (e.g. Boost.Stacktrace link checks fail). #}
 {# Pinning the executables here short-circuits that probe so Boost (and the rest of the toolchain) #}
@@ -49,31 +49,8 @@ tools.build:compiler_executables={'c':'{{ cc_exe }}','cpp':'{{ cxx_exe }}'}
 user.package:cppstd_version=23
 tools.info.package_id:confs+=["user.package:cppstd_version"]

-{% if os == "Linux" and context == "build" %}
-{# Build-context executables (protoc, grpc_cpp_plugin, build tools) run during the build #}
-{# and would otherwise load libstdc++/libgcc from a Nix store path #}
-{# that might change with a Nix toolchain update. #}
-{# --as-needed drops the ones they link but don't use, #}
-{# such as libatomic for grpc_cpp_plugin on arm64. #}
-{% set static_runtime_flags = ["-static-libstdc++", "-static-libgcc", "-Wl,--as-needed"] %}
-tools.build:exelinkflags+={{ static_runtime_flags }}
-tools.info.package_id:confs+=["tools.build:exelinkflags"]
-{% endif %}
-
-{% if os == "Linux" and context == "build" %}
-{# b2 links itself with its own script, which ignores exelinkflags #}
-{# and only takes CXXFLAGS when use_cxx_env is set (see [buildenv] below). #}
-[options]
-b2/*:use_cxx_env=True
-{% endif %}
-
-[buildenv]
-{# gRPC emits thousands of compiler warnings that we cannot act on. #}
-{# CMake picks up CXXFLAGS, and unlike tools.build:cxxflags, #}
-{# this is not part of the package ID, so binaries stay shareable. #}
-grpc/*:CXXFLAGS=-w
-
 {% if os == "Macos" %}
+[buildenv]
 {# os.version adds -mmacosx-version-min to compiler command lines, #}
 {# but Boost.Context's b2 assembly (.S) rule ignores it, #}
 {# so those objects keep the host SDK version and still warn at link time. #}
@@ -81,7 +58,3 @@ grpc/*:CXXFLAGS=-w
 {# Scoped to boost/* since it is the only gap. #}
 boost/*:MACOSX_DEPLOYMENT_TARGET={{ min_macos_version }}
 {% endif %}
-
-{% if os == "Linux" and context == "build" %}
-b2/*:CXXFLAGS={{ static_runtime_flags | join(" ") }}
-{% endif %}
@@ -5,9 +5,6 @@ include(default)

 {% if not sanitizers %}
 {# Sanitizers not configured; no additional settings needed #}
-{% elif context == "build" %}
-{# Build-context packages are tools we run, not code we test, #}
-{# so don't instrument them #}
 {% else %}

 {% if compiler == "msvc" %}
@@ -40,7 +40,7 @@ class Xrpl(ConanFile):
        "nudb/2.0.9",
        "openssl/3.6.3",
        "soci/4.0.3",
-        "xrpl-rpc-spec/0.1.21",
+        "xrpl-rpc-spec/0.1.19",
        "zlib/1.3.2",
    ]

@@ -49,8 +49,7 @@ class Xrpl(ConanFile):
    ]

    tool_requires = [
-        "grpc/<host_version>",
-        "protobuf/<host_version>",
+        "protobuf/6.33.5",
    ]

    default_options = {
@@ -124,10 +123,6 @@ class Xrpl(ConanFile):
        "xxhash/*:shared": False,
    }

-    # default_options only reach the host context;
-    # give tool_requires (and their dependencies) the same dependency options.
-    default_build_options = {k: v for k, v in default_options.items() if "/" in k}
-
    def set_version(self):
        self.version = self.version or DEV_VERSION

@@ -154,7 +149,7 @@ class Xrpl(ConanFile):
        self.requires("xxhash/0.8.3", transitive_headers=True)

    exports_sources = (
-        "bin/nix/default-loader-path.sh",
+        "bin/default-loader-path.sh",
        "CMakeLists.txt",
        "cfg/*",
        "cmake/*",
@@ -80,12 +80,13 @@ function(add_xrpl_crate name)
    # `cc` picks its runtime flag from `crt-static` alone, so it compiles a
    # crate's C++ with `-MT`; Debug needs `-MTd` (to match cmake/XrplCompiler.cmake).
    if(is_msvc)
-        corrosion_set_env_vars(${ARG_CRATE} "$<$<CONFIG:Debug>:CXXFLAGS=-MTd>")
+        corrosion_set_env_vars(
+            ${ARG_CRATE}
+            "$<$<CONFIG:Debug>:CXXFLAGS=-MTd>"
+        )
    endif()
-    corrosion_add_cxxbridge(
-        ${name}_cxxbridge
-        CRATE ${ARG_CRATE}
-        FILES ${ARG_FILES}
+    corrosion_add_cxxbridge(${name}_cxxbridge CRATE ${ARG_CRATE} FILES
+                            ${ARG_FILES}
    )
    # Generated cxxbridge headers don't exist at configure time; CMake 3.28+
    # validates INTERFACE_SOURCES on consuming targets. Clear it to skip the
@@ -57,9 +57,9 @@ dependencies = [

 [[package]]
 name = "cxx"
-version = "1.0.205"
+version = "1.0.202"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "c8a02bfeba9074b2bbb39492c9b0908ec854f455a3cccdcde7a1de55e2315e40"
+checksum = "13f6de320895f42e6e081abb5c7983bedcf0b6d0ff9323de0d33f620c8ac1199"
 dependencies = [
 "cc",
 "cxx-build",
@@ -72,9 +72,9 @@ dependencies = [

 [[package]]
 name = "cxx-build"
-version = "1.0.205"
+version = "1.0.202"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "bbc2ac76997619d2d888b58ad79488f1c835de41ca4dd3dd7082acf4ecb0a50f"
+checksum = "4fde53ca86b9704a943fef0f1e1d836239a6aedca5de3c65fa9f97ec0bd46d39"
 dependencies = [
 "cc",
 "codespan-reporting",
@@ -87,9 +87,9 @@ dependencies = [

 [[package]]
 name = "cxxbridge-cmd"
-version = "1.0.205"
+version = "1.0.202"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "54a486ed16a3f2a053fe7f5b98db4c1a59846a23e8cb9e49f9d1bd48830d51a2"
+checksum = "07bae89236c811fd4d08ed3441759ac4ac98d752cbfd3de341315ba16ad20ec3"
 dependencies = [
 "clap",
 "codespan-reporting",
@@ -101,15 +101,15 @@ dependencies = [

 [[package]]
 name = "cxxbridge-flags"
-version = "1.0.205"
+version = "1.0.202"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "c72569be94e2283764d334cd285980b747dbe221e27e315484eb6bcd93fa808a"
+checksum = "49045042e5fced01b80742aba5508de82aa4f13677ed3fa2b4cda709c40c5918"

 [[package]]
 name = "cxxbridge-macro"
-version = "1.0.205"
+version = "1.0.202"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "423e795284a7f7b3f843c956a5322c4bc497b60c2dd1a6339da10313cb18d575"
+checksum = "b181252e2e3d3b5d183afbdc033a3958b445eb3e1a0ae65fc3d4f5259f5da6fd"
 dependencies = [
 "indexmap",
 "proc-macro2",
@@ -10,15 +10,14 @@ This document explains how to set one up.
 support it — see [compiler support for C++23][cpp23-support].
 The versions currently tested in CI are:

-| Compiler    | Version                         |
-| ----------- | ------------------------------- |
-| GCC         | `gccVersion` in [packages.nix]  |
-| Clang       | `llvmVersion` in [packages.nix] |
-| Apple Clang | 21                              |
-| MSVC        | Visual Studio 2026              |
+| Compiler    | Version            |
+| ----------- | ------------------ |
+| GCC         | 15.2               |
+| Clang       | 22                 |
+| Apple Clang | 21                 |
+| MSVC        | Visual Studio 2026 |

-LLVM tools (`clang-tidy` and `clang-format`)
-come from the same LLVM release as Clang.
+LLVM tools (`clang-tidy` and `clang-format`) are also pinned to version 22.

 ### Older compilers

@@ -157,4 +156,3 @@ version out of the box — run it via `run-clang-tidy`. No separate installation
 is needed.

 [cpp23-support]: https://en.cppreference.com/w/cpp/compiler_support/23
-[packages.nix]: ../../nix/packages.nix
@@ -183,36 +183,24 @@ at link or run time.
 > configuration CI covers, and no dependency binaries are published for it.

 This is checked rather than assumed.
-[`bin/nix/check-nix-store-refs.sh`](../../bin/nix/check-nix-store-refs.sh) takes one
-file or directory and fails if a binary under it resolves a store path at run time.
+[`bin/check-nix-store-refs.sh`](../../bin/check-nix-store-refs.sh) takes one file
+or directory and fails if a binary under it resolves a store path at run time.
 CI runs it over the build output and the Conan cache, and again in the upload job
 before anything is published. You can run it yourself:

 ```bash
-bin/nix/check-nix-store-refs.sh build
-bin/nix/check-nix-store-refs.sh ~/.conan2-nix
+bin/check-nix-store-refs.sh build
+bin/check-nix-store-refs.sh ~/.conan2-nix
 ```

 It works on Linux too, but asserts something narrower there: the toolchain always
-writes the store into `PT_INTERP` and `RUNPATH`. That is fine for the pinned
-glibc, whose path does not move, but not for the GCC runtime, which moves with
-every GCC update. So [`conan/profiles/default`](../../conan/profiles/default)
-links build-context packages, whose executables run during the build, with
-`-static-libstdc++ -static-libgcc -Wl,--as-needed`, and
-[`conan/profiles/sanitizers`](../../conan/profiles/sanitizers) does not
-instrument them. CI checks that they load nothing from the store but glibc, from
-the graph `conan install --format=json` writes:
+writes the store into `PT_INTERP` and `RUNPATH`, and CI builds inside an image
+whose store is fixed for its lifetime, so that is fine. Only the binaries
+[`PatchNixBinary.cmake`](../../cmake/PatchNixBinary.cmake) retargets to the
+system loader have to be clean, and those are what CI checks:

 ```bash
-bin/nix/check-build-context-runtime.sh graph.json
-```
-
-Only the binaries [`PatchNixBinary.cmake`](../../cmake/PatchNixBinary.cmake)
-retargets to the system loader have to be fully clean, and those are what CI
-checks:
-
-```bash
-bin/nix/check-nix-store-refs.sh build/xrpld
+bin/check-nix-store-refs.sh build/xrpld
 ```

 ### The libresolv stub
@@ -178,11 +178,11 @@ A binary stops starting after a `nix flake update`, or after
 dyld[57271]: Library not loaded: /nix/store/…-libresolv-93/lib/libresolv.9.dylib
 ```

-[`bin/nix/check-nix-store-refs.sh`](../../bin/nix/check-nix-store-refs.sh) finds the
-same thing without having to run anything, and names the file:
+[`bin/check-nix-store-refs.sh`](../../bin/check-nix-store-refs.sh) finds the same
+thing without having to run anything, and names the file:

 ```
-$ bin/nix/check-nix-store-refs.sh ~/.conan2-nix
+$ bin/check-nix-store-refs.sh ~/.conan2-nix
 ::error file=/Users/you/.conan2-nix/p/b/c-area24ded30c388c/p/bin/adig::references the Nix store at run time
 /Users/you/.conan2-nix/p/b/c-area24ded30c388c/p/bin/adig
    /nix/store/p4lp3xq4imd1qzqh08x8vcq2zfhi7rca-libresolv-93/lib/libresolv.9.dylib
@@ -6,7 +6,7 @@ project(antithesis-sdk-cpp VERSION 0.4.4 LANGUAGES CXX)
 add_library(antithesis-sdk-cpp INTERFACE antithesis_sdk.h)

 # Note, both sections below created by xrpld project
-target_include_directories(antithesis-sdk-cpp SYSTEM INTERFACE
+target_include_directories(antithesis-sdk-cpp INTERFACE
  $<INSTALL_INTERFACE:${CMAKE_INSTALL_INCLUDEDIR}>
  $<BUILD_INTERFACE:${CMAKE_CURRENT_SOURCE_DIR}>
 )
@@ -34,7 +34,7 @@ lz4Compress(void const* in, std::size_t inSize, BufferFactory&& bf)
    // data
    auto compressed = bf(outCapacity);

-    auto const compressedSize = LZ4_compress_default(
+    auto compressedSize = LZ4_compress_default(
        reinterpret_cast<char const*>(in),
        reinterpret_cast<char*>(compressed),
        inSize,
@@ -1,29 +1,31 @@
 #pragma once

-#include <xrpl/beast/type_name.h>
+#include <xrpl/beast/utility/instrumentation.h>

+#include <boost/core/type_name.hpp>
+
+#include <algorithm>
 #include <atomic>
+#include <cstddef>
+#include <cstdint>
+#include <iterator>
 #include <string>
+#include <type_traits>
 #include <utility>
-#include <vector>

 namespace xrpl {

 /**
 * Manages all counted object types.
+ *
+ * Counters register themselves on a lock-free intrusive list maintained
+ * by this object when constructed. Because counters are never destroyed
+ * or removed, the ABA problem does not apply.
+ *
+ * The registry is iterable as a forward range.
 */
 class CountedObjects
 {
-public:
-    static CountedObjects&
-    getInstance() noexcept;
-
-    using Entry = std::pair<std::string, int>;
-    using List = std::vector<Entry>;
-
-    [[nodiscard]] List
-    getCounts(int minimumThreshold) const;
-
 public:
    /**
     * Implementation for @ref CountedObject.
@@ -33,68 +35,147 @@ public:
    class Counter
    {
    public:
-        Counter(std::string name) noexcept : name_(std::move(name)), count_(0)
-        {
-            // Insert ourselves at the front of the lock-free linked list
-            CountedObjects& instance = CountedObjects::getInstance();
-            Counter* head = nullptr;
+        Counter(std::string name) noexcept;

-            do
-            {
-                head = instance.head_.load();
-                next_ = head;
-            } while (instance.head_.exchange(this) != head);
+        // Counters are intrusive list nodes whose addresses are published
+        // in the registry; they must never be copied or moved. The atomic
+        // members already force this, but we make it explicit.
+        Counter(Counter const&) = delete;
+        Counter&
+        operator=(Counter const&) = delete;
+        Counter(Counter&&) = delete;
+        Counter&
+        operator=(Counter&&) = delete;

-            ++instance.count_;
-        }
-
-        ~Counter() noexcept = default;
-
-        int
+        std::uint32_t
        increment() noexcept
        {
-            return ++count_;
+            auto const newCount = count_.fetch_add(1, std::memory_order::relaxed) + 1;
+            XRPL_ASSERT(newCount != 0, "xrpl::CountedObjects::Counter::increment : no overflow");
+
+            auto maxCount = maxCount_.load(std::memory_order::relaxed);
+
+            while (newCount > maxCount &&
+                   !maxCount_.compare_exchange_weak(maxCount, newCount, std::memory_order::relaxed))
+            {
+            }
+
+            return newCount;
        }

-        int
+        std::uint32_t
        decrement() noexcept
        {
-            return --count_;
+            auto const prev = count_.fetch_sub(1, std::memory_order::relaxed);
+            XRPL_ASSERT(prev != 0, "xrpl::CountedObjects::Counter::decrement : no underflow");
+            return prev - 1;
        }

-        [[nodiscard]] int
-        getCount() const noexcept
+        [[nodiscard]] std::uint32_t
+        count() const noexcept
        {
-            return count_.load();
+            return count_.load(std::memory_order::relaxed);
        }

-        [[nodiscard]] Counter*
-        getNext() const noexcept
+        [[nodiscard]] std::uint32_t
+        max() const noexcept
        {
-            return next_;
+            return std::max(
+                count_.load(std::memory_order::relaxed),
+                maxCount_.load(std::memory_order::relaxed));
        }

        [[nodiscard]] std::string const&
-        getName() const noexcept
+        name() const noexcept
        {
            return name_;
        }

    private:
+        friend class CountedObjects;
+
+        Counter* next_ = nullptr;
+        std::atomic<std::uint32_t> count_ = 0;
+        std::atomic<std::uint32_t> maxCount_ = 0;
        std::string const name_;
-        std::atomic<int> count_;
-        Counter* next_;
    };

-private:
-    CountedObjects() noexcept;
-    ~CountedObjects() noexcept = default;
+    class Iterator
+    {
+    public:
+        using value_type = Counter const;
+        using reference = value_type&;
+        using pointer = value_type*;
+        using difference_type = std::ptrdiff_t;
+        using iterator_category = std::forward_iterator_tag;
+
+        explicit Iterator(Counter* c = nullptr) noexcept : current_(c)
+        {
+        }
+
+        reference
+        operator*() const noexcept
+        {
+            return *current_;
+        }
+
+        pointer
+        operator->() const noexcept
+        {
+            return current_;
+        }
+
+        Iterator&
+        operator++() noexcept
+        {
+            current_ = current_->next_;
+            return *this;
+        }
+
+        Iterator
+        operator++(int) noexcept
+        {
+            auto tmp = *this;
+            ++*this;
+            return tmp;
+        }
+
+        bool
+        operator==(Iterator const&) const noexcept = default;
+
+    private:
+        Counter* current_;
+    };
+
+    constexpr CountedObjects() noexcept = default;
+
+    [[nodiscard]] auto
+    begin() const noexcept
+    {
+        return Iterator{head_.load(std::memory_order::acquire)};
+    }
+
+    [[nodiscard]] auto
+    end() const noexcept
+    {
+        return Iterator{};
+    }

 private:
-    std::atomic<int> count_;
-    std::atomic<Counter*> head_;
+    std::atomic<Counter*> head_ = nullptr;
 };

+/** The global counted object registry. */
+inline constinit CountedObjects gCountedObjects;
+
+inline CountedObjects::Counter::Counter(std::string name) noexcept
+    : next_(gCountedObjects.head_.load(std::memory_order::relaxed)), name_(std::move(name))
+{
+    while (!gCountedObjects.head_.compare_exchange_weak(
+        next_, this, std::memory_order::release, std::memory_order::relaxed))
+        ;
+}
+
 //------------------------------------------------------------------------------

 /**
@@ -103,27 +184,37 @@ private:
 * Derived classes have their instances counted automatically. This is used
 * for reporting purposes.
 *
+ * The constructors are private and `Object` is befriended so that the
+ * CRTP parameter must be the deriving class itself: a copy-paste error
+ * like `class B : public CountedObject<A>` fails to compile instead of
+ * silently polluting A's count.
+ *
+ * @note This class has no move operations by design: a derived class's
+ *       move constructor falls back to the copy constructor for this
+ *       base, so the newly created instance is counted. This keeps the
+ *       invariant that count is the number of outstanding subobjects.
+ *
+ * @warning Counted objects constructed during dynamic initialization of
+ *          other translation units may have their increments discarded when
+ *          counter itself is dynamically initialized. Do not create counted
+ *          objects before main() begins.
+ *
 * @ingroup basics
 */
 template <class Object>
+    requires std::is_class_v<Object>
 class CountedObject
 {
-private:
-    static auto&
-    getCounter() noexcept
-    {
-        static CountedObjects::Counter kC{beast::typeName<Object>()};
-        return kC;
-    }
+    static inline CountedObjects::Counter counter{boost::core::type_name<Object>()};

    CountedObject() noexcept
    {
-        getCounter().increment();
+        counter.increment();
    }

    CountedObject(CountedObject const&) noexcept
    {
-        getCounter().increment();
+        counter.increment();
    }

    CountedObject&
@@ -132,7 +223,7 @@ private:
 public:
    ~CountedObject() noexcept
    {
-        getCounter().decrement();
+        counter.decrement();
    }

    friend Object;
@@ -20,14 +20,12 @@ SharedIntrusive<T>::SharedIntrusive(T* p, TAdoptTag) noexcept : ptr_{p}

 template <class T>
 SharedIntrusive<T>::SharedIntrusive(SharedIntrusive const& rhs)
-    : ptr_{
-          [&] {
-              auto p = rhs.unsafeGetRawPtr();
-              if (p)
-                  p->addStrongRef();
-              return p;
-          }(),
-      }
+    : ptr_{[&] {
+        auto p = rhs.unsafeGetRawPtr();
+        if (p)
+            p->addStrongRef();
+        return p;
+    }()}
 {
 }

@@ -35,14 +33,12 @@ template <class T>
 template <class TT>
    requires std::convertible_to<TT*, T*>
 SharedIntrusive<T>::SharedIntrusive(SharedIntrusive<TT> const& rhs)
-    : ptr_{
-          [&] {
-              auto p = rhs.unsafeGetRawPtr();
-              if (p)
-                  p->addStrongRef();
-              return p;
-          }(),
-      }
+    : ptr_{[&] {
+        auto p = rhs.unsafeGetRawPtr();
+        if (p)
+            p->addStrongRef();
+        return p;
+    }()}
 {
 }

@@ -144,14 +140,12 @@ SharedIntrusive<T>::~SharedIntrusive()
 template <class T>
 template <class TT>
 SharedIntrusive<T>::SharedIntrusive(StaticCastTagSharedIntrusive, SharedIntrusive<TT> const& rhs)
-    : ptr_{
-          [&] {
-              auto p = static_cast<T*>(rhs.unsafeGetRawPtr());
-              if (p)
-                  p->addStrongRef();
-              return p;
-          }(),
-      }
+    : ptr_{[&] {
+        auto p = static_cast<T*>(rhs.unsafeGetRawPtr());
+        if (p)
+            p->addStrongRef();
+        return p;
+    }()}
 {
 }

@@ -165,14 +159,12 @@ SharedIntrusive<T>::SharedIntrusive(StaticCastTagSharedIntrusive, SharedIntrusiv
 template <class T>
 template <class TT>
 SharedIntrusive<T>::SharedIntrusive(DynamicCastTagSharedIntrusive, SharedIntrusive<TT> const& rhs)
-    : ptr_{
-          [&] {
-              auto p = dynamic_cast<T*>(rhs.unsafeGetRawPtr());
-              if (p)
-                  p->addStrongRef();
-              return p;
-          }(),
-      }
+    : ptr_{[&] {
+        auto p = dynamic_cast<T*>(rhs.unsafeGetRawPtr());
+        if (p)
+            p->addStrongRef();
+        return p;
+    }()}
 {
 }

@@ -268,7 +260,7 @@ SharedIntrusive<T>::unsafeReleaseAndStore(T* next)
        return;

    using enum ReleaseStrongRefAction;
-    auto const action = prev->releaseStrongRef();
+    auto action = prev->releaseStrongRef();
    switch (action)
    {
        case NoOp:
@@ -369,7 +361,7 @@ WeakIntrusive<T>::unsafeReleaseNoStore()
        return;

    using enum ReleaseWeakRefAction;
-    auto const action = ptr_->releaseWeakRef();
+    auto action = ptr_->releaseWeakRef();
    switch (action)
    {
        case NoOp:
@@ -588,7 +580,7 @@ SharedWeakUnion<T>::convertToStrong()
    auto p = unsafeGetRawPtr();
    if (p && p->checkoutStrongRefFromWeak())
    {
-        [[maybe_unused]] auto const action = p->releaseWeakRef();
+        [[maybe_unused]] auto action = p->releaseWeakRef();
        XRPL_ASSERT(
            (action == ReleaseWeakRefAction::NoOp),
            "xrpl::SharedWeakUnion::convertToStrong : "
@@ -611,7 +603,7 @@ SharedWeakUnion<T>::convertToWeak()
        return false;

    using enum ReleaseStrongRefAction;
-    auto const action = p->addWeakReleaseStrongRef();
+    auto action = p->addWeakReleaseStrongRef();
    switch (action)
    {
        case NoOp:
@@ -678,7 +670,7 @@ SharedWeakUnion<T>::unsafeReleaseNoStore()
    if (isStrong())
    {
        using enum ReleaseStrongRefAction;
-        auto const strongAction = p->releaseStrongRef();
+        auto strongAction = p->releaseStrongRef();
        switch (strongAction)
        {
            case NoOp:
@@ -696,7 +688,7 @@ SharedWeakUnion<T>::unsafeReleaseNoStore()
    else
    {
        using enum ReleaseWeakRefAction;
-        auto const weakAction = p->releaseWeakRef();
+        auto weakAction = p->releaseWeakRef();
        switch (weakAction)
        {
            case NoOp:
@@ -406,10 +406,10 @@ IntrusiveRefCounts::useCount() const noexcept
 inline IntrusiveRefCounts::~IntrusiveRefCounts() noexcept
 {
 #ifndef NDEBUG
-    auto const v = refCounts_.load(std::memory_order_acquire);
+    auto v = refCounts_.load(std::memory_order_acquire);
    XRPL_ASSERT(
        (!(v & kValueMask)), "xrpl::IntrusiveRefCounts::~IntrusiveRefCounts : count must be zero");
-    auto const t = v & kTagMask;
+    auto t = v & kTagMask;
    XRPL_ASSERT((!t || t == kTagMask), "xrpl::IntrusiveRefCounts::~IntrusiveRefCounts : valid tag");
 #endif
 }
@@ -3,84 +3,9 @@
 #include <algorithm>
 #include <cassert>
 #include <cstddef>
-#include <cstdint>
-#include <limits>
-#include <optional>

 namespace xrpl {

-/**
- * Add two signed 64-bit integers, returning std::nullopt when the exact
- * mathematical sum is not representable in std::int64_t.
- */
-[[nodiscard]] constexpr std::optional<std::int64_t>
-checkedAdd(std::int64_t a, std::int64_t b) noexcept
-{
-    using L = std::numeric_limits<std::int64_t>;
-
-    if ((b > 0 && a > L::max() - b) || (b < 0 && a < L::min() - b))
-        return std::nullopt;
-
-    return a + b;
-}
-
-/**
- * Subtract two signed 64-bit integers, returning std::nullopt when the exact
- * mathematical difference is not representable in std::int64_t.
- */
-[[nodiscard]] constexpr std::optional<std::int64_t>
-checkedSub(std::int64_t a, std::int64_t b) noexcept
-{
-    using L = std::numeric_limits<std::int64_t>;
-
-    if ((b > 0 && a < L::min() + b) || (b < 0 && a > L::max() + b))
-        return std::nullopt;
-
-    return a - b;
-}
-
-static_assert(checkedAdd(0, 0) == 0);
-static_assert(checkedAdd(1, -1) == 0);
-static_assert(checkedAdd(-5, 2) == -3);
-static_assert(!checkedAdd(std::numeric_limits<std::int64_t>::max(), 1).has_value());
-static_assert(!checkedAdd(std::numeric_limits<std::int64_t>::min(), -1).has_value());
-static_assert(
-    checkedAdd(std::numeric_limits<std::int64_t>::max() - 1, 1) ==
-    std::numeric_limits<std::int64_t>::max());
-static_assert(
-    checkedAdd(
-        std::numeric_limits<std::int64_t>::min(),
-        std::numeric_limits<std::int64_t>::max()) == -1);
-static_assert(
-    checkedAdd(
-        std::numeric_limits<std::int64_t>::max(),
-        std::numeric_limits<std::int64_t>::min()) == -1);
-static_assert(
-    !checkedAdd(std::numeric_limits<std::int64_t>::max(), std::numeric_limits<std::int64_t>::max())
-         .has_value());
-static_assert(
-    !checkedAdd(std::numeric_limits<std::int64_t>::min(), std::numeric_limits<std::int64_t>::min())
-         .has_value());
-
-static_assert(checkedSub(0, 0) == 0);
-static_assert(checkedSub(1, 1) == 0);
-static_assert(checkedSub(-5, 2) == -7);
-static_assert(checkedSub(-5, -2) == -3);
-static_assert(!checkedSub(std::numeric_limits<std::int64_t>::min(), 1).has_value());
-static_assert(!checkedSub(std::numeric_limits<std::int64_t>::max(), -1).has_value());
-static_assert(
-    checkedSub(std::numeric_limits<std::int64_t>::min() + 1, 1) ==
-    std::numeric_limits<std::int64_t>::min());
-static_assert(
-    checkedSub(-1, std::numeric_limits<std::int64_t>::max()) ==
-    std::numeric_limits<std::int64_t>::min());
-static_assert(
-    !checkedSub(std::numeric_limits<std::int64_t>::max(), std::numeric_limits<std::int64_t>::min())
-         .has_value());
-static_assert(
-    !checkedSub(std::numeric_limits<std::int64_t>::min(), std::numeric_limits<std::int64_t>::max())
-         .has_value());
-
 /**
 * Calculate one number divided by another number in percentage.
 * The result is rounded up to the next integer, and capped in the range [0,100]
@@ -253,4 +253,27 @@ makeSlice(std::basic_string_view<char, Traits> s)
    return Slice(s.data(), s.size());
 }

+/**
+ * Checks whether two byte ranges share at least one byte of storage.
+ *
+ *  Each slice is treated as the half-open range [data, data + size), so
+ *  ranges that merely touch do not overlap. An empty slice overlaps
+ *  nothing, wherever it points.
+ *
+ *  @param a The first range.
+ *  @param b The second range.
+ *  @return true if the ranges overlap, false otherwise.
+ */
+[[nodiscard]] inline bool
+overlap(Slice a, Slice b) noexcept
+{
+    if (a.empty() || b.empty())
+        return false;
+
+    auto const a0 = reinterpret_cast<std::uintptr_t>(a.data());
+    auto const b0 = reinterpret_cast<std::uintptr_t>(b.data());
+
+    return a0 < b0 + b.size() && b0 < a0 + a.size();
+}
+
 }  // namespace xrpl
@@ -1,6 +1,7 @@
 #pragma once

 #include <xrpl/basics/Blob.h>
+#include <xrpl/basics/Slice.h>

 #include <boost/utility/string_view.hpp>

@@ -28,7 +29,7 @@ namespace xrpl {
 * @return The input, encoded as a blob literal.
 */
 std::string
-sqlBlobLiteral(Blob const& blob);
+sqlBlobLiteral(Slice blob);

 namespace detail {

@@ -36,7 +37,7 @@ template <typename T>
 concept SomeChar = std::same_as<std::remove_cvref_t<T>, int8_t> ||
    std::same_as<std::remove_cvref_t<T>, char> || std::same_as<std::remove_cvref_t<T>, uint8_t>;

-inline constexpr std::array<std::optional<int>, 256> const kDigitLookupTable = [] {
+inline constexpr std::array<std::optional<int>, 256> const kDigitLookupTable = []() {
    std::array<std::optional<int>, 256> t{};

    for (int i = 0; i < 10; ++i)
@@ -300,7 +300,7 @@ TaggedCache<Key, T, IsKeyCache, SharedWeakUnionPointer, SharedPointerType, Hash,
    // Remove from cache, if !valid, remove from map too. Returns true if removed from cache
    std::scoped_lock const lock(mutex_);

-    auto const cit = cache_.find(key);
+    auto cit = cache_.find(key);

    if (cit == cache_.end())
        return false;
@@ -356,7 +356,7 @@ TaggedCache<Key, T, IsKeyCache, SharedWeakUnionPointer, SharedPointerType, Hash,

    std::scoped_lock const lock(mutex_);

-    auto const cit = cache_.find(key);
+    auto cit = cache_.find(key);

    if (cit == cache_.end())
    {
@@ -371,7 +371,7 @@ TaggedCache<Key, T, IsKeyCache, SharedWeakUnionPointer, SharedPointerType, Hash,
    Entry& entry = cit->second;
    entry.touch(clock_.now());

-    auto const shouldReplaceCached = [&] {
+    auto shouldReplaceCached = [&] {
        if constexpr (replaceCached)
        {
            return true;
@@ -400,7 +400,7 @@ TaggedCache<Key, T, IsKeyCache, SharedWeakUnionPointer, SharedPointerType, Hash,
        return true;
    }

-    auto const cachedData = entry.lock();
+    auto cachedData = entry.lock();

    if (cachedData)
    {
@@ -563,7 +563,7 @@ TaggedCache<Key, T, IsKeyCache, SharedWeakUnionPointer, SharedPointerType, Hash,
    retrieve(key_type const& key, T& data)
 {
    // retrieve the value of the stored data
-    auto const entry = fetch(key);
+    auto entry = fetch(key);

    if (!entry)
        return false;
@@ -687,7 +687,7 @@ TaggedCache<Key, T, IsKeyCache, SharedWeakUnionPointer, SharedPointerType, Hash,
            return ret;
    }

-    auto const sle = h();
+    auto sle = h();
    if (!sle)
        return {};

@@ -713,7 +713,7 @@ inline SharedPointerType
 TaggedCache<Key, T, IsKeyCache, SharedWeakUnionPointer, SharedPointerType, Hash, KeyEqual, Mutex>::
    initialFetch(key_type const& key, std::scoped_lock<MutexType> const& l)
 {
-    auto const cit = cache_.find(key);
+    auto cit = cache_.find(key);
    if (cit == cache_.end())
        return {};

@@ -783,7 +783,7 @@ TaggedCache<Key, T, IsKeyCache, SharedWeakUnionPointer, SharedPointerType, Hash,
        std::atomic<int>& allRemovals,
        std::scoped_lock<std::recursive_mutex> const&)
 {
-    return std::thread([&, this] {
+    return std::thread([&, this]() {
        int cacheRemovals = 0;
        int mapRemovals = 0;

@@ -863,7 +863,7 @@ TaggedCache<Key, T, IsKeyCache, SharedWeakUnionPointer, SharedPointerType, Hash,
        std::atomic<int>& allRemovals,
        std::scoped_lock<std::recursive_mutex> const&)
 {
-    return std::thread([&, this] {
+    return std::thread([&, this]() {
        // NOLINTBEGIN https://github.com/XRPLF/rippled/issues/7056
        int cacheRemovals = 0;
        int mapRemovals = 0;
@@ -1,7 +1,8 @@
 #pragma once

 #include <xrpl/basics/sanitizers.h>
-#include <xrpl/beast/type_name.h>
+
+#include <boost/core/type_name.hpp>

 #include <exception>
 #include <string>
@@ -55,7 +56,7 @@ Throw(Args&&... args)
        std::is_convertible_v<E*, std::exception*>, "Exception must derive from std::exception.");

    E e(std::forward<Args>(args)...);
-    logThrow(std::string("Throwing exception of type " + beast::typeName<E>() + ": ") + e.what());
+    logThrow("Throwing exception of type " + boost::core::type_name<E>() + ": " + e.what());
    throw std::move(e);
 }

@@ -0,0 +1,200 @@
+#pragma once
+
+#include <concepts>
+#include <type_traits>
+#include <utility>
+
+namespace xrpl {
+
+/**
+ * Opt-in bitwise operators for scoped enumerations.
+ *
+ * Provides `&`, `|`, `^`, `~` and the corresponding compound assignment forms
+ * for any scoped enumeration with an unsigned underlying type that opts in by
+ * specializing @ref OptIn:
+ *
+ * @code
+ * namespace xrpl {
+ *
+ * enum class MyFlags : std::uint32_t { a = 1, b = 2, c = 4 };
+ *
+ * template <>
+ * struct enum_bitops::OptIn<MyFlags> : std::true_type {};
+ *
+ * }  // namespace xrpl
+ * @endcode
+ *
+ * The operators are constexpr and noexcept, and return the enumeration type.
+ *
+ * @par Where the specialization may be declared
+ *   In namespace xrpl, or at global scope with full qualification, and after
+ *   the enumeration is defined but before the operators are first used. This
+ *   cannot be done in a nested namespace or at class scope; for enumerations
+ *   nested in a class this means after the class definition.
+ *
+ * @par Where the operators are found
+ *   The operators are declared in namespace xrpl, so argument-dependent lookup
+ *   will find them only for enumerations declared in xrpl or nested in a class
+ *   declared directly in xrpl. Enumerations in a nested namespace will only be
+ *   found from code inside xrpl, and only if no enclosing scope declares an
+ *   operator of the same name; elsewhere they require using-declarations.
+ */
+namespace enum_bitops {
+
+/**
+ * Types that may be opted in to the bitwise operators.
+ *
+ * Satisfied by a scoped enumeration whose underlying type is an unsigned
+ * integer type.
+ *
+ * @tparam T The type to test.
+ */
+template <typename T>
+concept Eligible = std::is_scoped_enum_v<T> && std::unsigned_integral<std::underlying_type_t<T>>;
+
+/**
+ * Opt-in switch for the bitwise operators.
+ *
+ * The primary template derives from std::false_type. Specialize it to derive
+ * from std::true_type to enable the operators for an @ref Eligible enumeration.
+ *
+ * @tparam T The scoped enumeration to opt in.
+ */
+template <Eligible T>
+struct OptIn : std::false_type
+{
+};
+
+/**
+ * Enumerations for which the bitwise operators are enabled.
+ *
+ * Satisfied when @p T satisfies @ref Eligible and @ref OptIn has been
+ * specialized for it to derive from std::true_type.
+ *
+ * @tparam T The type to test.
+ */
+template <typename T>
+concept Candidate = Eligible<T> && OptIn<T>::value;
+
+}  // namespace enum_bitops
+
+// These are declared in xrpl, not in enum_bitops, so that argument-dependent
+// lookup finds them for enumerations whose associated namespace is xrpl.
+
+/**
+ * @name Bitwise operators for opted-in scoped enumerations
+ *
+ * Each operator applies the corresponding built-in operator to the
+ * underlying values and converts the result back to the enumeration type.
+ * Both operands must have the same enumeration type; there is no implicit
+ * conversion to or from the underlying type.
+ */
+/** @{ */
+
+/**
+ * Bitwise AND.
+ *
+ * @param lhs The left operand.
+ * @param rhs The right operand.
+ * @return The bits set in both @p lhs and @p rhs.
+ */
+template <enum_bitops::Candidate T>
+constexpr T
+operator&(T lhs, T rhs) noexcept
+{
+    return static_cast<T>(std::to_underlying(lhs) & std::to_underlying(rhs));
+}
+
+/**
+ * Bitwise OR.
+ *
+ * @param lhs The left operand.
+ * @param rhs The right operand.
+ * @return The bits set in @p lhs, in @p rhs, or in both.
+ */
+template <enum_bitops::Candidate T>
+constexpr T
+operator|(T lhs, T rhs) noexcept
+{
+    return static_cast<T>(std::to_underlying(lhs) | std::to_underlying(rhs));
+}
+
+/**
+ * Bitwise exclusive OR.
+ *
+ * @param lhs The left operand.
+ * @param rhs The right operand.
+ * @return The bits set in exactly one of @p lhs and @p rhs.
+ */
+template <enum_bitops::Candidate T>
+constexpr T
+operator^(T lhs, T rhs) noexcept
+{
+    return static_cast<T>(std::to_underlying(lhs) ^ std::to_underlying(rhs));
+}
+
+/**
+ * Bitwise complement.
+ *
+ * The result has every bit of the underlying type that is clear in
+ * @p val, including bits that no enumerator names. It is intended for
+ * clearing flags, as in `flags & ~flag`.
+ *
+ * @param val The operand.
+ * @return The complement of @p val.
+ */
+template <enum_bitops::Candidate T>
+constexpr T
+operator~(T val) noexcept
+{
+    return static_cast<T>(~std::to_underlying(val));
+}
+
+/**
+ * Bitwise AND assignment.
+ *
+ * @param lhs The value to modify.
+ * @param rhs The right operand.
+ * @return A reference to @p lhs.
+ */
+template <enum_bitops::Candidate T>
+constexpr T&
+operator&=(T& lhs, T rhs) noexcept
+{
+    lhs = lhs & rhs;
+    return lhs;
+}
+
+/**
+ * Bitwise OR assignment.
+ *
+ * @param lhs The value to modify.
+ * @param rhs The right operand.
+ * @return A reference to @p lhs.
+ */
+template <enum_bitops::Candidate T>
+constexpr T&
+operator|=(T& lhs, T rhs) noexcept
+{
+    lhs = lhs | rhs;
+    return lhs;
+}
+
+/**
+ * Bitwise exclusive OR assignment.
+ *
+ * @param lhs The value to modify.
+ * @param rhs The right operand.
+ * @return A reference to @p lhs.
+ */
+template <enum_bitops::Candidate T>
+constexpr T&
+operator^=(T& lhs, T rhs) noexcept
+{
+    lhs = lhs ^ rhs;
+    return lhs;
+}
+
+/** @} */
+
+}  // namespace xrpl
@@ -17,7 +17,7 @@
 namespace xrpl {

 template <typename Key>
-std::size_t
+static std::size_t
 extract(Key const& key)
 {
    return key;
@@ -365,7 +365,7 @@ public:
    size() const
    {
        std::size_t ret = 0;
-        for (auto const& p : map_)
+        for (auto& p : map_)
            ret += p.size();
        return ret;
    }
@@ -2,107 +2,212 @@

 #include <xrpl/beast/utility/instrumentation.h>  // IWYU pragma: keep

+#include <concepts>
+#include <cstdint>
+#include <limits>
+#include <memory>
 #include <type_traits>
+#include <utility>

 namespace xrpl {

-// safe_cast adds compile-time checks to a static_cast to ensure that
-// the destination can hold all values of the source.  This is particularly
-// handy when the source or destination is an enumeration type.
+/** Every value of @p Src can be represented by @p Dest.

-template <class Src, class Dest>
-concept SafeToCast = (std::is_integral_v<Src> && std::is_integral_v<Dest>) &&
-    (std::is_signed_v<Src> || std::is_unsigned_v<Dest>) &&
-    (std::is_signed_v<Src> != std::is_signed_v<Dest> ? sizeof(Dest) > sizeof(Src)
-                                                     : sizeof(Dest) >= sizeof(Src));
+    Given two integral types, the cast is safe when the destination can
+    hold every possible value of the source.

-template <class Dest, class Src>
+    Comparing the bounds requires care: we use @c std::cmp_less_equal and
+    @c std::cmp_greater_equal; the plain relational operators would apply
+    arithmetic conversions, resulting in incorrect results when comparing
+    across signedness.
+
+    Because @c std::cmp_* requires standard signed or unsigned integer
+    arguments, which excludes character types and bool, we first widen
+    all type bounds to the maximum-width integer type while preserving
+    signedness.
+
+    @note Extended integer types, like __int128 on gcc, cannot be safely
+          widened to a standard integer type, so the concept will reject
+          them.
+*/
+template <typename Src, typename Dest>
+concept SafeToCast = std::is_integral_v<Src> && std::is_integral_v<Dest> && []() consteval {
+    using WideSrc = std::conditional_t<std::is_signed_v<Src>, std::intmax_t, std::uintmax_t>;
+    using WideDest = std::conditional_t<std::is_signed_v<Dest>, std::intmax_t, std::uintmax_t>;
+
+    // Note: this guard must be an evaluated branch, and not a
+    // static_assert. The lambda body is outside the immediate
+    // context, so a substitution-time failure here would be a
+    // hard error rather than leaving the concept unsatisfied.
+    if constexpr (sizeof(Src) > sizeof(WideSrc) || sizeof(Dest) > sizeof(WideDest))
+    {
+        return false;
+    }
+    else
+    {
+        return std::cmp_less_equal(
+                   static_cast<WideDest>(std::numeric_limits<Dest>::min()),
+                   static_cast<WideSrc>(std::numeric_limits<Src>::min())) &&
+            std::cmp_greater_equal(
+                   static_cast<WideDest>(std::numeric_limits<Dest>::max()),
+                   static_cast<WideSrc>(std::numeric_limits<Src>::max()));
+    }
+}();
+
+/** Compile-time-checked static_cast that rejects non-value preserving casts.
+
+    @note There is deliberately no enum-to-enum overload, and no overload
+          returning the underlying type of an enum. For the latter, use
+          @c std::to_underlying.
+*/
+/** @{ */
+template <typename Dest, typename Src>
+    requires(std::is_integral_v<Dest> && std::is_integral_v<Src>)
 constexpr Dest
 safeCast(Src s) noexcept
-    requires(std::is_integral_v<Dest> && std::is_integral_v<Src>)
 {
    static_assert(
-        std::is_signed_v<Dest> || std::is_unsigned_v<Src>, "Cannot cast signed to unsigned");
-    constexpr unsigned kNotSame = std::is_signed_v<Dest> != std::is_signed_v<Src>;
-    static_assert(
-        sizeof(Dest) >= sizeof(Src) + kNotSame,
-        "Destination is too small to hold all values of source");
+        SafeToCast<Src, Dest>, "This cast is not value-preserving. Please use unsafeCast instead.");
    return static_cast<Dest>(s);
 }

-template <class Dest, class Src>
+template <typename Dest, typename Src>
+    requires(std::is_enum_v<Dest> && std::is_integral_v<Src>)
 constexpr Dest
 safeCast(Src s) noexcept
-    requires(std::is_enum_v<Dest> && std::is_integral_v<Src>)
 {
    return static_cast<Dest>(safeCast<std::underlying_type_t<Dest>>(s));
 }

-template <class Dest, class Src>
+template <typename Dest, typename Src>
+    requires(std::is_integral_v<Dest> && std::is_enum_v<Src>)
 constexpr Dest
 safeCast(Src s) noexcept
-    requires(std::is_integral_v<Dest> && std::is_enum_v<Src>)
 {
-    return safeCast<Dest>(static_cast<std::underlying_type_t<Src>>(s));
+    return safeCast<Dest>(std::to_underlying(s));
 }
+/** @} */

-// unsafe_cast explicitly flags a static_cast as not necessarily able to hold
-// all values of the source. It includes a compile-time check so that if
-// underlying types become safe, it can be converted to a safe_cast.
+/** Integral-to-integral cast that is known to be narrowing or sign-erasing.

-template <class Dest, class Src>
+    This explicitly flags a conversion that can lose information for some
+    values of the source type, where the call site accepts that loss (or
+    truncation is the intended behavior).
+
+    The compile-time check ensures the cast remains "unsafe": if the types
+    involved later change such that the conversion becomes inherently
+    value-preserving, the static assertion fires with instructions to
+    migrate the call site to @ref safeCast.
+
+    If the conversion's safety depends on a runtime precondition rather
+    than on the types, or varies across instantiations of generic code,
+    use @ref checkedCast instead.
+*/
+/** @{ */
+template <typename Dest, typename Src>
+    requires(std::is_integral_v<Dest> && std::is_integral_v<Src>)
 constexpr Dest
 unsafeCast(Src s) noexcept
-    requires(std::is_integral_v<Dest> && std::is_integral_v<Src>)
 {
    static_assert(
-        !SafeToCast<Src, Dest>,
-        "Only unsafe if casting signed to unsigned or "
-        "destination is too small");
+        !SafeToCast<Src, Dest>, "This cast is value-preserving. Please use safeCast instead.");
    return static_cast<Dest>(s);
 }

-template <class Dest, class Src>
+template <typename Dest, typename Src>
+    requires(std::is_enum_v<Dest> && std::is_integral_v<Src>)
 constexpr Dest
 unsafeCast(Src s) noexcept
-    requires(std::is_enum_v<Dest> && std::is_integral_v<Src>)
 {
    return static_cast<Dest>(unsafeCast<std::underlying_type_t<Dest>>(s));
 }

-template <class Dest, class Src>
+template <typename Dest, typename Src>
+    requires(std::is_integral_v<Dest> && std::is_enum_v<Src>)
 constexpr Dest
 unsafeCast(Src s) noexcept
-    requires(std::is_integral_v<Dest> && std::is_enum_v<Src>)
 {
-    return unsafeCast<Dest>(static_cast<std::underlying_type_t<Src>>(s));
+    return unsafeCast<Dest>(std::to_underlying(s));
+}
+/** @} */
+
+/** Integral-to-integral cast when the caller has performed a bounds check.
+
+    This documents that a runtime precondition or external invariant, which
+    is not necessarily visible to the compiler, guarantees that the requested
+    conversion is value-preserving for the values that can actually occur.
+
+    This is primarily meant for generic code, where the same expression may
+    be value-preserving for one instantiation but not for another, making
+    both @ref safeCast and @ref unsafeCast unusable.
+
+    Unlike @ref safeCast and @ref unsafeCast, this imposes no static check
+    on the type relationship: the caller's claim is about runtime values,
+    not about types.
+*/
+/** @{ */
+template <typename Dest, typename Src>
+    requires(std::is_integral_v<Dest> && std::is_integral_v<Src>)
+constexpr Dest
+checkedCast(Src s) noexcept
+{
+    return static_cast<Dest>(s);
 }

+template <typename Dest, typename Src>
+    requires(std::is_enum_v<Dest> && std::is_integral_v<Src>)
+constexpr Dest
+checkedCast(Src s) noexcept
+{
+    return static_cast<Dest>(checkedCast<std::underlying_type_t<Dest>>(s));
+}
+
+template <typename Dest, typename Src>
+    requires(std::is_integral_v<Dest> && std::is_enum_v<Src>)
+constexpr Dest
+checkedCast(Src s) noexcept
+{
+    return checkedCast<Dest>(std::to_underlying(s));
+}
+/** @} */
+
+/** Downcast within a class hierarchy, verified in debug builds.
+
+    Performs a static_cast down a hierarchy, but in debug builds verifies
+    via dynamic_cast that the object's dynamic type actually permits the
+    downcast. Both build modes execute the same conversion; debug builds
+    merely add the check.
+
+    The pointer form passes null through unchanged, as dynamic_cast does.
+
+    @note The check requires @p Src to be polymorphic; in release builds
+          an invalid downcast is undefined behavior on use, exactly as
+          with a bare static_cast.
+*/
+/** @{ */
 template <class Dest, class Src>
-    requires std::is_pointer_v<Dest>
+    requires(
+        std::is_pointer_v<Dest> && std::is_polymorphic_v<Src> &&
+        std::derived_from<std::remove_pointer_t<Dest>, Src> &&
+        (std::is_const_v<std::remove_pointer_t<Dest>> || !std::is_const_v<Src>) &&
+        (std::is_volatile_v<std::remove_pointer_t<Dest>> || !std::is_volatile_v<Src>))
 inline Dest
 safeDowncast(Src* s) noexcept
 {
-#ifdef NDEBUG
+    XRPL_ASSERT(
+        s == nullptr || dynamic_cast<Dest>(s) != nullptr, "xrpl::safeDowncast : valid downcast");
    return static_cast<Dest>(s);  // NOLINT(cppcoreguidelines-pro-type-static-cast-downcast)
-#else
-    auto* result = dynamic_cast<Dest>(s);
-    XRPL_ASSERT(result != nullptr, "xrpl::safeDowncast : pointer downcast is valid");
-    return result;
-#endif
 }

 template <class Dest, class Src>
-    requires std::is_lvalue_reference_v<Dest>
+    requires(
+        std::is_lvalue_reference_v<Dest> && std::is_polymorphic_v<Src> &&
+        std::derived_from<std::remove_reference_t<Dest>, Src>)
 inline Dest
 safeDowncast(Src& s) noexcept
 {
-#ifndef NDEBUG
-    XRPL_ASSERT(
-        dynamic_cast<std::add_pointer_t<std::remove_reference_t<Dest>>>(&s) != nullptr,
-        "xrpl::safeDowncast : reference downcast is valid");
-#endif
-    return static_cast<Dest>(s);  // NOLINT(cppcoreguidelines-pro-type-static-cast-downcast)
+    return *safeDowncast<std::add_pointer_t<std::remove_reference_t<Dest>>>(std::addressof(s));
 }
+/** @} */

 }  // namespace xrpl
@@ -4,88 +4,215 @@

 #include <xrpl/beast/utility/instrumentation.h>

+#include <boost/predef/architecture.h>
+
 #include <atomic>
+#include <concepts>
 #include <limits>
 #include <type_traits>

-#ifndef __aarch64__
+#if BOOST_ARCH_X86
 #include <immintrin.h>
 #endif

 namespace xrpl {

+/** An unsigned integral type suitable for use as a spinlock.
+
+    The type must be always lock-free when wrapped in std::atomic, so
+    that lock operations cannot themselves take a (library-level) lock.
+ */
+template <typename T>
+concept SpinlockValueType = std::is_unsigned_v<T> && std::atomic<T>::is_always_lock_free;
+
+/** A spinlock value type that additionally supports the atomic bitwise
+    operations required to pack multiple locks into a single integer.
+ */
+template <typename T>
+concept PackedSpinlockValueType = SpinlockValueType<T> && requires(std::atomic<T>& a, T v) {
+    { a.fetch_or(v) } -> std::same_as<T>;
+    { a.fetch_and(v) } -> std::same_as<T>;
+};
+
 namespace detail {
-/**
- * Inform the processor that we are in a tight spin-wait loop.
- *
- * Spinlocks caught in tight loops can result in the processor's pipeline
- * filling up with comparison operations, resulting in a misprediction at
- * the time the lock is finally acquired, necessitating pipeline flushing
- * which is ridiculously expensive and results in very high latency.
- *
- * This function instructs the processor to "pause" for some architecture
- * specific amount of time, to prevent this.
+
+/** Inform the processor that we are in a tight spin-wait loop.
+
+    Spinlocks caught in tight loops can result in the processor's pipeline
+    filling up with comparison operations, resulting in a misprediction at
+    the time the lock is finally acquired, necessitating pipeline flushing
+    which is ridiculously expensive and results in very high latency.
+
+    This function instructs the processor to "pause" for some architecture
+    specific amount of time, to prevent this.
 */
 inline void
 spinPause() noexcept
 {
-#ifdef __aarch64__
-    asm volatile("yield");
-#else
+#if BOOST_ARCH_X86
    _mm_pause();
+#elif BOOST_ARCH_ARM
+    asm volatile("yield" ::: "memory");
+#else
+#error No implementation available for spinPause to use
 #endif
 }

 }  // namespace detail

-/** @{ */
-/**
- * Classes to handle arrays of spinlocks packed into a single atomic integer:
- *
- * Packed spinlocks allow for tremendously space-efficient lock-sharding
- * but they come at a cost.
- *
- * First, the implementation is necessarily low-level and uses advanced
- * features like memory ordering and highly platform-specific tricks to
- * maximize performance. This imposes a significant and ongoing cost to
- * developers.
- *
- * Second, and perhaps most important, is that the packing of multiple
- * locks into a single integer which, albeit space-efficient, also has
- * performance implications stemming from data dependencies, increased
- * cache-coherency traffic between processors and heavier loads on the
- * processor's load/store units.
- *
- * To be sure, these locks can have advantages but they are definitely
- * not general purpose locks and should not be thought of or used that
- * way. The use cases for them are likely few and far between; without
- * a compelling reason to use them, backed by profiling data, it might
- * be best to use one of the standard locking primitives instead. Note
- * that in most common platforms, `std::mutex` is so heavily optimized
- * that it can, usually, outperform spinlocks.
- *
- * @tparam T An unsigned integral type (e.g. std::uint16_t)
- */
+//------------------------------------------------------------------------------

-/**
- * A class that grabs a single packed spinlock from an atomic integer.
- *
- * This class meets the requirements of Lockable:
- *     https://en.cppreference.com/w/cpp/named_req/Lockable
+/** Attempt to acquire a spinlock without blocking.
+
+    @note This interface is primarily intended for one-shot attempts to
+          acquire the lock. Avoid calling this function directly from a
+          loop and use @ref spinLock instead.
+
+    @tparam T An unsigned integral type.
+    @param lock The atomic variable used as the lock.
+    @return true if the lock was acquired, false if it was already held.
 */
-template <class T>
+template <SpinlockValueType T>
+[[nodiscard]] bool
+spinTryLock(std::atomic<T>& lock) noexcept
+{
+    // A compare-exchange is required here, not an unconditional exchange:
+    // a failed attempt must not modify the lock word, in case the atomic
+    // is shared with PackedSpinlock).
+    T expected = 0;
+
+    return lock.compare_exchange_strong(
+        expected,
+        std::numeric_limits<T>::max(),
+        std::memory_order::acquire,
+        std::memory_order::relaxed);
+    ;
+}
+
+/** Acquire a spinlock, blocking until available.
+
+    Uses a TTAS (test-and-test-and-set) pattern, so waiters share the cache
+    line read-only, helping to avoid unnecessary coherency traffic.
+
+    @tparam T An unsigned integral type.
+    @param lock The atomic variable used as the lock.
+ */
+template <SpinlockValueType T>
+void
+spinLock(std::atomic<T>& lock) noexcept
+{
+    do
+    {
+        // Relaxed ordering is sufficient for the spin: this load is only
+        // a filter. The acquire on the successful exchange in spinTryLock
+        // is what synchronizes the critical section.
+        while (lock.load(std::memory_order::relaxed) != 0)
+            detail::spinPause();
+    } while (!spinTryLock(lock));
+}
+
+/** Release a spinlock.
+
+    @tparam T An unsigned integral type.
+    @param lock The atomic variable used as the lock.
+ */
+template <SpinlockValueType T>
+void
+spinUnlock(std::atomic<T>& lock) noexcept
+{
+    lock.store(0, std::memory_order::release);
+}
+
+//------------------------------------------------------------------------------
+
+/** A Lockable interface to a spinlock implemented on top of an atomic.
+
+    @tparam T An unsigned integral type.
+
+    @note Using `PackedSpinlock` and `Spinlock` against the same underlying
+          atomic integer is possible but can result in `Spinlock` not being
+          able to acquire the lock during periods of high contention due to
+          the way the two locks operate: `Spinlock` spins and tries to grab
+          all the bits at once, whereas any given `PackedSpinlock` instance
+          only tries to grab one bit at a time. Caveat emptor.
+
+    This class meets the requirements of Lockable:
+        https://en.cppreference.com/w/cpp/named_req/Lockable
+ */
+template <SpinlockValueType T>
+class Spinlock
+{
+    std::atomic<T>& lock_;
+
+public:
+    Spinlock(Spinlock const&) = delete;
+    Spinlock&
+    operator=(Spinlock const&) = delete;
+
+    /** Construct a spinlock handle.
+
+        @param lock The atomic integer to spin against.
+
+        @note For performance reasons, you should strive to have `lock` be
+              on a cacheline by itself.
+     */
+    explicit Spinlock(std::atomic<T>& lock) noexcept : lock_(lock)
+    {
+    }
+
+    [[nodiscard]] bool
+    try_lock() noexcept  // NOLINT(readability-identifier-naming)
+    {
+        return spinTryLock(lock_);
+    }
+
+    void
+    lock() noexcept
+    {
+        spinLock(lock_);
+    }
+
+    void
+    unlock() noexcept
+    {
+        spinUnlock(lock_);
+    }
+};
+
+//------------------------------------------------------------------------------
+
+/** A Lockable interface to a packed spinlock implemented on top of an atomic.
+
+    Packed spinlocks offer tremendous space-efficient lock-sharding but
+    they come at a cost.
+
+    First, the implementation is necessarily low-level and uses advanced
+    features like memory ordering and highly platform-specific tricks to
+    maximize performance. This imposes a significant and ongoing cost to
+    developers.
+
+    Second, and perhaps most important, is that the packing of multiple
+    locks into a single integer which, albeit space-efficient, also has
+    performance implications stemming from data dependencies, increased
+    cache-coherency traffic between processors and heavier loads on the
+    processor's load/store units.
+
+    To be sure, these locks can have advantages but they are definitely
+    not general purpose locks and should not be thought of or used that
+    way. The use cases for them are likely few and far between; without
+    a compelling reason to use them, backed by profiling data, it might
+    be best to use one of the standard locking primitives instead. Note
+    that in most common platforms, `std::mutex` is so heavily optimized
+    that it can, usually, outperform spinlocks.
+
+    @tparam T An unsigned integral type (e.g. std::uint16_t)
+
+    This class meets the requirements of Lockable:
+        https://en.cppreference.com/w/cpp/named_req/Lockable
+ */
+template <PackedSpinlockValueType T>
 class PackedSpinlock
 {
-    // clang-format off
-    static_assert(std::is_unsigned_v<T>);
-    static_assert(std::atomic<T>::is_always_lock_free);
-    static_assert(
-        std::is_same_v<decltype(std::declval<std::atomic<T>&>().fetch_or(0)), T> &&
-        std::is_same_v<decltype(std::declval<std::atomic<T>&>().fetch_and(0)), T>,
-        "std::atomic<T>::fetch_and(T) and std::atomic<T>::fetch_and(T) are required by packed_spinlock");
-    // clang-format on
-
-private:
    std::atomic<T>& bits_;
    T const mask_;

@@ -94,120 +221,49 @@ public:
    PackedSpinlock&
    operator=(PackedSpinlock const&) = delete;

-    /**
-     * A single spinlock packed inside the specified atomic
-     *
-     * @param lock The atomic integer inside which the spinlock is packed.
-     * @param index The index of the spinlock this object acquires.
-     *
-     * @note For performance reasons, you should strive to have `lock` be
-     *       on a cacheline by itself.
+    /** Construct a packed spinlock handle for a single bit.
+
+        @param lock The atomic integer inside which the spinlock is packed.
+        @param index The index of the spinlock this object acquires.
+
+        @note For performance reasons, you should strive to have `lock` be
+              on a cacheline by itself.
     */
-    PackedSpinlock(std::atomic<T>& lock, int index) : bits_(lock), mask_(static_cast<T>(1) << index)
-    {
-        XRPL_ASSERT(
-            index >= 0 && (mask_ != 0),
-            "xrpl::PackedSpinlock::PackedSpinlock : valid index and mask");
-    }
-
-    [[nodiscard]] bool
-    try_lock()  // NOLINT(readability-identifier-naming)
-    {
-        return (bits_.fetch_or(mask_, std::memory_order_acquire) & mask_) == 0;
-    }
-
-    void
-    lock()
-    {
-        while (!try_lock())
-        {
-            // The use of relaxed memory ordering here is intentional and
-            // serves to help reduce cache coherency traffic during times
-            // of contention by avoiding writes that would definitely not
-            // result in the lock being acquired.
-            while ((bits_.load(std::memory_order_relaxed) & mask_) != 0)
-                detail::spinPause();
-        }
-    }
-
-    void
-    unlock()
-    {
-        bits_.fetch_and(~mask_, std::memory_order_release);
-    }
-};
-
-/**
- * A spinlock implemented on top of an atomic integer.
- *
- * @note Using `packed_spinlock` and `spinlock` against the same underlying
- *       atomic integer can result in `spinlock` not being able to actually
- *       acquire the lock during periods of high contention, because of how
- *       the two locks operate: `spinlock` will spin trying to grab all the
- *       bits at once, whereas any given `packed_spinlock` will only try to
- *       grab one bit at a time. Caveat emptor.
- *
- * This class meets the requirements of Lockable:
- *     https://en.cppreference.com/w/cpp/named_req/Lockable
- */
-template <class T>
-class Spinlock
-{
-    static_assert(std::is_unsigned_v<T>);
-    static_assert(std::atomic<T>::is_always_lock_free);
-
-private:
-    std::atomic<T>& lock_;
-
-public:
-    Spinlock(Spinlock const&) = delete;
-    Spinlock&
-    operator=(Spinlock const&) = delete;
-
-    /**
-     * Grabs the
-     *
-     * @param lock The atomic integer to spin against.
-     *
-     * @note For performance reasons, you should strive to have `lock` be
-     *       on a cacheline by itself.
-     */
-    Spinlock(std::atomic<T>& lock) : lock_(lock)
+    PackedSpinlock(std::atomic<T>& lock, int index) noexcept
+        : bits_(lock), mask_([index]() {
+            XRPL_ASSERT(
+                index >= 0 && index < std::numeric_limits<T>::digits,
+                "xrpl::PackedSpinlock::PackedSpinlock : valid index");
+            return static_cast<T>(T{1} << index);
+        }())
    {
    }

    [[nodiscard]] bool
-    try_lock()  // NOLINT(readability-identifier-naming)
+    try_lock() noexcept  // NOLINT(readability-identifier-naming)
    {
-        T expected = 0;
-
-        return lock_.compare_exchange_weak(
-            expected,
-            std::numeric_limits<T>::max(),
-            std::memory_order_acquire,
-            std::memory_order_relaxed);
+        return (bits_.fetch_or(mask_, std::memory_order::acquire) & mask_) == 0;
    }

    void
-    lock()
+    lock() noexcept
    {
-        while (!try_lock())
+        do
        {
            // The use of relaxed memory ordering here is intentional and
            // serves to help reduce cache coherency traffic during times
-            // of contention by avoiding writes that would definitely not
-            // result in the lock being acquired.
-            while (lock_.load(std::memory_order_relaxed) != 0)
+            // of contention by avoiding writes that are unlikely to grab
+            // the requested lock.
+            while ((bits_.load(std::memory_order::relaxed) & mask_) != 0)
                detail::spinPause();
-        }
+        } while (!try_lock());
    }

    void
-    unlock()
+    unlock() noexcept
    {
-        lock_.store(0, std::memory_order_release);
+        bits_.fetch_and(~mask_, std::memory_order::release);
    }
 };
-/** @} */

 }  // namespace xrpl
@@ -1604,7 +1604,7 @@ AgedOrderedContainer<IsMulti, IsMap, Key, T, Clock, Compare, Allocator>::erase(
    beast::detail::AgedContainerIterator<IsConst, Iterator> pos)
    requires(!IsBoostReverseIterator<Iterator>::value)
 {
-    unlinkAndDeleteElement(&*(pos++).iterator());
+    unlinkAndDeleteElement(&*((pos++).iterator()));
    return beast::detail::AgedContainerIterator<false, Iterator>(pos.iterator());
 }

@@ -1617,7 +1617,7 @@ AgedOrderedContainer<IsMulti, IsMap, Key, T, Clock, Compare, Allocator>::erase(
    requires(!IsBoostReverseIterator<Iterator>::value)
 {
    for (; first != last;)
-        unlinkAndDeleteElement(&*(first++).iterator());
+        unlinkAndDeleteElement(&*((first++).iterator()));

    return beast::detail::AgedContainerIterator<false, Iterator>(first.iterator());
 }
@@ -2404,7 +2404,7 @@ beast::detail::AgedContainerIterator<false, Iterator>
 AgedUnorderedContainer<IsMulti, IsMap, Key, T, Clock, Hash, KeyEqual, Allocator>::erase(
    beast::detail::AgedContainerIterator<IsConst, Iterator> pos)
 {
-    unlinkAndDeleteElement(&*(pos++).iterator());
+    unlinkAndDeleteElement(&*((pos++).iterator()));
    return beast::detail::AgedContainerIterator<false, Iterator>(pos.iterator());
 }

@@ -2424,7 +2424,7 @@ AgedUnorderedContainer<IsMulti, IsMap, Key, T, Clock, Hash, KeyEqual, Allocator>
    beast::detail::AgedContainerIterator<IsConst, Iterator> last)
 {
    for (; first != last;)
-        unlinkAndDeleteElement(&*(first++).iterator());
+        unlinkAndDeleteElement(&*((first++).iterator()));

    return beast::detail::AgedContainerIterator<false, Iterator>(first.iterator());
 }
@@ -2536,7 +2536,7 @@ AgedUnorderedContainer<IsMulti, IsMap, Key, T, Clock, Hash, KeyEqual, Allocator>
        return false;
    for (auto iter(cbegin()), last(cend()), otherLast(other.cend()); iter != last; ++iter)
    {
-        auto const otherIter(other.find(extract(*iter)));
+        auto otherIter(other.find(extract(*iter)));
        if (otherIter == otherLast)
            return false;
    }
@@ -68,7 +68,7 @@ struct LexicalCast<Out, std::string_view>
        if (first != last && *first == '+')
            ++first;

-        auto const ret = std::from_chars(first, last, out);
+        auto ret = std::from_chars(first, last, out);

        return ret.ec == std::errc() && ret.ptr == last;
    }
@@ -390,7 +390,7 @@ void
 hash_append(Hasher& h, boost::container::flat_set<Key, Compare, Alloc> const& v) noexcept
    requires(IsContiguouslyHashable<Key, Hasher>::value)
 {
-    h(&v.begin(), v.size() * sizeof(Key));
+    h(&(v.begin()), v.size() * sizeof(Key));
 }
 // tuple

@@ -8,7 +8,7 @@ namespace beast::insight {
 class HookImpl : public std::enable_shared_from_this<HookImpl>
 {
 public:
-    using HandlerType = std::function<void()>;
+    using HandlerType = std::function<void(void)>;

    virtual ~HookImpl() = 0;
 };
@@ -61,7 +61,7 @@ isMulticast(Address const& addr)
 inline bool
 isPrivate(Address const& addr)
 {
-    return addr.is_v4() ? isPrivate(addr.to_v4()) : isPrivate(addr.to_v6());
+    return (addr.is_v4()) ? isPrivate(addr.to_v4()) : isPrivate(addr.to_v6());
 }

 /**
@@ -70,7 +70,7 @@ isPrivate(Address const& addr)
 inline bool
 isPublic(Address const& addr)
 {
-    return addr.is_v4() ? isPublic(addr.to_v4()) : isPublic(addr.to_v6());
+    return (addr.is_v4()) ? isPublic(addr.to_v4()) : isPublic(addr.to_v6());
 }

 }  // namespace ip
@@ -83,7 +83,7 @@ trimRight(String const& s)
 {
    using std::begin;
    using std::end;
-    auto const first(begin(s));
+    auto first(begin(s));
    auto last(end(s));
    last = trimRight(first, last);
    return {first, last};
@@ -1,47 +0,0 @@
-#pragma once
-
-#include <cstdlib>
-#include <string>
-#include <type_traits>
-#include <typeinfo>
-
-#ifndef _MSC_VER
-#include <cxxabi.h>
-#endif
-
-namespace beast {
-
-template <typename T>
-std::string
-typeName()
-{
-    using TR = std::remove_reference_t<T>;
-
-    std::string name = typeid(TR).name();
-
-#ifndef _MSC_VER
-    if (auto s = abi::__cxa_demangle(name.c_str(), nullptr, nullptr, nullptr))
-    {
-        name = s;
-        // NOLINTNEXTLINE(cppcoreguidelines-no-malloc)
-        std::free(s);
-    }
-#endif
-
-    if (std::is_const_v<TR>)
-        name += " const";
-    if (std::is_volatile_v<TR>)
-        name += " volatile";
-    if (std::is_lvalue_reference_v<T>)
-    {
-        name += '&';
-    }
-    else if (std::is_rvalue_reference_v<T>)
-    {
-        name += "&&";
-    }
-
-    return name;
-}
-
-}  // namespace beast
@@ -32,7 +32,7 @@ public:
        Module,

        // Match nothing (used internally)
-        None,
+        None
    };

 private:
@@ -20,7 +20,7 @@ namespace beast::unit_test {
 namespace detail {

 template <class String>
-std::string
+static std::string
 makeReason(String const& reason, char const* file, int line)
 {
    std::string s(reason);
@@ -47,8 +47,8 @@ public:
    template <class F, class... Args>
    explicit Thread(Suite& s, F&& f, Args&&... args) : s_(&s)
    {
-        std::function<void()> b = [f = std::forward<F>(f),
-                                   ... args = std::forward<Args>(args)] mutable {
+        std::function<void(void)> b = [f = std::forward<F>(f),
+                                       ... args = std::forward<Args>(args)]() mutable {
            std::invoke(f, args...);
        };
        t_ = std::thread(&Thread::run, this, std::move(b));
@@ -94,7 +94,7 @@ public:

 private:
    void
-    run(std::function<void()> f)
+    run(std::function<void(void)> f)
    {
        try
        {
@@ -24,7 +24,7 @@ enum class Severity : std::uint8_t {
    Fatal = 5,

    Disabled = 6,
-    None = Disabled,
+    None = Disabled
 };

 /**
@@ -2,26 +2,24 @@

 #pragma once

+#include <compare>
+#include <concepts>
+
 namespace beast {

 /**
 * Zero allows classes to offer efficient comparisons to zero.
 *
- * Zero is a struct to allow classes to efficiently compare with zero without
- * requiring an rvalue construction.
- *
 * It's often the case that we have classes which combine a number and a unit.
 * In such cases, comparisons like t > 0 or t != 0 make sense, but comparisons
- * like t > 1 or t != 1 do not.
+ * like t > 1 or t != 1 do not. Comparing against kZero expresses exactly that,
+ * without constructing a T.
 *
- * The class Zero allows such comparisons to be easily made.
- *
- * The comparing class T either needs to have a method called signum() which
- * returns a positive number, 0, or a negative; or there needs to be a signum
- * function which resolves in the namespace which takes an instance of T and
- * returns a positive, zero or negative number.
+ * A type T participates if either `t.signum()` or an unqualified `signum(t)`
+ * found by argument-dependent lookup returns an integer that is negative,
+ * zero, or positive according to the sign of t. Both `t == kZero` and
+ * `kZero == t` work, as do all six relational operators in either order.
 */
-
 struct Zero
 {
    explicit Zero() = default;
@@ -30,115 +28,44 @@ struct Zero
 inline constexpr Zero kZero{};

 /**
- * Default implementation of signum calls the method on the class.
+ * Default implementation of signum: call the member function.
 */
-template <typename T>
-auto
-signum(T const& t)
+template <class T>
+    requires requires(T const& t) {
+        { t.signum() } -> std::integral;
+    }
+[[nodiscard]] constexpr auto
+signum(T const& t) noexcept(noexcept(t.signum()))
 {
    return t.signum();
 }

-namespace detail::zero_helper {
+namespace detail {

-// For argument dependent lookup to function properly, calls to signum must
-// be made from a namespace that does not include overloads of the function..
+/**
+ * A type with a usable signum: either the member-based default above, or a
+ * `signum(t)` overload in T's own namespace, found by ADL. A user overload
+ * that is a better match than the template wins, as usual.
+ */
 template <class T>
-auto
-callSignum(T const& t)
+concept HasSignum = requires(T const& t) {
+    { signum(t) } -> std::integral;
+};
+
+}  // namespace detail
+
+template <detail::HasSignum T>
+[[nodiscard]] constexpr bool
+operator==(T const& t, Zero) noexcept(noexcept(signum(t)))
 {
-    return signum(t);
+    return signum(t) == 0;
 }

-}  // namespace detail::zero_helper
-
-// Handle operators where T is on the left side using signum.
-
-template <typename T>
-bool
-operator==(T const& t, Zero)
+template <detail::HasSignum T>
+[[nodiscard]] constexpr std::strong_ordering
+operator<=>(T const& t, Zero) noexcept(noexcept(signum(t)))
 {
-    return detail::zero_helper::callSignum(t) == 0;
-}
-
-template <typename T>
-bool
-operator!=(T const& t, Zero)
-{
-    return detail::zero_helper::callSignum(t) != 0;
-}
-
-template <typename T>
-bool
-operator<(T const& t, Zero)
-{
-    return detail::zero_helper::callSignum(t) < 0;
-}
-
-template <typename T>
-bool
-operator>(T const& t, Zero)
-{
-    return detail::zero_helper::callSignum(t) > 0;
-}
-
-template <typename T>
-bool
-operator>=(T const& t, Zero)
-{
-    return detail::zero_helper::callSignum(t) >= 0;
-}
-
-template <typename T>
-bool
-operator<=(T const& t, Zero)
-{
-    return detail::zero_helper::callSignum(t) <= 0;
-}
-
-// Handle operators where T is on the right side by
-// reversing the operation, so that T is on the left side.
-
-template <typename T>
-bool
-operator==(Zero, T const& t)
-{
-    return t == kZero;
-}
-
-template <typename T>
-bool
-operator!=(Zero, T const& t)
-{
-    return t != kZero;
-}
-
-template <typename T>
-bool
-operator<(Zero, T const& t)
-{
-    return t > kZero;
-}
-
-template <typename T>
-bool
-operator>(Zero, T const& t)
-{
-    return t < kZero;
-}
-
-template <typename T>
-bool
-operator>=(Zero, T const& t)
-{
-    return t <= kZero;
-}
-
-template <typename T>
-bool
-operator<=(Zero, T const& t)
-{
-    return t >= kZero;
+    return signum(t) <=> 0;
 }

 }  // namespace beast
@@ -18,7 +18,7 @@ enum class Type : std::uint8_t {
    PrefixSha256 = 1,
    ThresholdSha256 = 2,
    RsaSha256 = 3,
-    Ed25519Sha256 = 4,
+    Ed25519Sha256 = 4
 };

 class Condition
@@ -49,7 +49,7 @@ public:

        using namespace der;

-        auto const p = parsePreamble(s, ec);
+        auto p = parsePreamble(s, ec);
        if (ec)
            return nullptr;

@@ -21,7 +21,7 @@ enum class Error {
    UnexpectedTag,
    LongTag,
    LargeSize,
-    PreimageTooLong,
+    PreimageTooLong
 };

 std::error_code
@@ -396,7 +396,7 @@ inline bool
 getIfExists<bool>(Section const& section, std::string const& name, bool& v)
 {
    int intVal = 0;
-    auto const stat = getIfExists(section, name, intVal);
+    auto stat = getIfExists(section, name, intVal);
    if (stat)
        v = bool(intVal);
    return stat;
@@ -93,14 +93,14 @@ public:
    void
    check(std::vector<TxID> accepted, Predicate&& pred)
    {
-        auto const acceptTxid = accepted.begin();
+        auto acceptTxid = accepted.begin();
        auto const ae = accepted.end();
        std::sort(acceptTxid, ae);

        // We want to remove all tracking entries for transactions that were
        // accepted as well as those which match the predicate.

-        auto const i = removeIfIntersectOrMatch(
+        auto i = removeIfIntersectOrMatch(
            tracker_.begin(),
            tracker_.end(),
            accepted.begin(),
@@ -779,7 +779,7 @@ Consensus<Adaptor>::peerProposalInternal(

    {
        // update current position
-        auto const peerPosIt = currPeerPositions_.find(peerID);
+        auto peerPosIt = currPeerPositions_.find(peerID);

        if (peerPosIt != currPeerPositions_.end())
        {
@@ -1110,7 +1110,7 @@ Consensus<Adaptor>::checkLedger(std::unique_ptr<std::stringstream> const& clog)
 {
    CLOG(clog) << "checkLedger. ";

-    auto const netLgr = adaptor_.getPrevLedger(prevLedgerID_, previousLedger_, mode_.get());
+    auto netLgr = adaptor_.getPrevLedger(prevLedgerID_, previousLedger_, mode_.get());
    CLOG(clog) << "network ledgerid " << netLgr << ",  "
               << "previous ledger " << prevLedgerID_ << ". ";

@@ -1159,8 +1159,8 @@ Consensus<Adaptor>::phaseOpen(std::unique_ptr<std::stringstream> const& clog)

    // it is shortly before ledger close time
    bool const anyTransactions = adaptor_.hasOpenTransactions();
-    auto const proposersClosed = currPeerPositions_.size();
-    auto const proposersValidated = adaptor_.proposersValidated(prevLedgerID_);
+    auto proposersClosed = currPeerPositions_.size();
+    auto proposersValidated = adaptor_.proposersValidated(prevLedgerID_);

    openTime_.tick(clock_.now());

@@ -1669,7 +1669,7 @@ Consensus<Adaptor>::haveConsensus(std::unique_ptr<std::stringstream> const& clog
    // CHECKME: should possibly count unacquired TX sets as disagreeing
    int agree = 0, disagree = 0;

-    auto const ourPosition = result_->position.position();
+    auto ourPosition = result_->position.position();

    // Count number of agreements/disagreements with our position
    for (auto const& [nodeId, peerPos] : currPeerPositions_)
@@ -1686,7 +1686,7 @@ Consensus<Adaptor>::haveConsensus(std::unique_ptr<std::stringstream> const& clog
            ++disagree;
        }
    }
-    auto const currentFinished = adaptor_.proposersFinished(previousLedger_, prevLedgerID_);
+    auto currentFinished = adaptor_.proposersFinished(previousLedger_, prevLedgerID_);

    JLOG(j_.debug()) << "Checking for TX consensus: agree=" << agree << ", disagree=" << disagree;

@@ -1812,7 +1812,7 @@ Consensus<Adaptor>::createDisputes(TxSetT const& o, std::unique_ptr<std::strings
               << ". ";
    JLOG(j_.debug()) << "createDisputes " << result_->txns.id() << " to " << o.id();

-    auto const differences = result_->txns.compare(o);
+    auto differences = result_->txns.compare(o);

    int dc = 0;

@@ -1826,7 +1826,7 @@ Consensus<Adaptor>::createDisputes(TxSetT const& o, std::unique_ptr<std::strings
            "xrpl::Consensus::createDisputes : has disputed transactions");

        TxT const tx = inThisSet ? result_->txns.find(txId) : o.find(txId);
-        auto const txID = tx.id();
+        auto txID = tx.id();

        if (result_->disputes.find(txID) != result_->disputes.end())
            continue;
@@ -1837,8 +1837,7 @@ Consensus<Adaptor>::createDisputes(TxSetT const& o, std::unique_ptr<std::strings
            tx,
            result_->txns.exists(txID),
            std::max(prevProposers_, currPeerPositions_.size()),
-            j_,
-        };
+            j_};

        // Update all of the available peer's votes on the disputed transaction
        for (auto const& [nodeId, peerPos] : currPeerPositions_)
@@ -145,26 +145,18 @@ struct ConsensusParms
    std::map<AvalancheState, AvalancheCutoff> const avalancheCutoffs{
        // {state, {time, percent, nextState}},
        // Initial state: 50% of nodes must vote yes
-        {
-            AvalancheState::Init,
-            {.consensusTime = 0, .consensusPct = 50, .next = AvalancheState::Mid},
-        },
+        {AvalancheState::Init,
+         {.consensusTime = 0, .consensusPct = 50, .next = AvalancheState::Mid}},
        // mid-consensus starts after 50% of the previous round time, and
        // requires 65% yes
-        {
-            AvalancheState::Mid,
-            {.consensusTime = 50, .consensusPct = 65, .next = AvalancheState::Late},
-        },
+        {AvalancheState::Mid,
+         {.consensusTime = 50, .consensusPct = 65, .next = AvalancheState::Late}},
        // late consensus starts after 85% time, and requires 70% yes
-        {
-            AvalancheState::Late,
-            {.consensusTime = 85, .consensusPct = 70, .next = AvalancheState::Stuck},
-        },
+        {AvalancheState::Late,
+         {.consensusTime = 85, .consensusPct = 70, .next = AvalancheState::Stuck}},
        // we're stuck after 2x time, requires 95% yes votes
-        {
-            AvalancheState::Stuck,
-            {.consensusTime = 200, .consensusPct = 95, .next = AvalancheState::Stuck},
-        },
+        {AvalancheState::Stuck,
+         {.consensusTime = 200, .consensusPct = 95, .next = AvalancheState::Stuck}},
    };

    /**
@@ -59,7 +59,7 @@ enum class ConsensusMode {
     * if we entered the round observing, but is used to indicate we did
     * have the wrongLedger at some point.
     */
-    SwitchedLedger,
+    SwitchedLedger
 };

 inline std::string
@@ -222,7 +222,7 @@ medianCloseOffset(ConsensusCloseTimes const& times)

    // Accumulate weight in time order; the first bin to reach halfWeight is
    // the (lower) weighted median. Returns true once that bin is found.
-    auto const step = [&](time_point t, std::int64_t w) {
+    auto step = [&](time_point t, std::int64_t w) {
        XRPL_ASSERT(tally < halfWeight, "xrpl::medianCloseOffset::step : median not yet found");
        tally += w;
        if (tally >= halfWeight)
@@ -267,7 +267,7 @@ enum class ConsensusState {
    No,       ///< We do not have consensus
    MovedOn,  ///< The network has consensus without us
    Expired,  ///< Consensus time limit has hard-expired
-    Yes,      ///< We have consensus along with the network
+    Yes       ///< We have consensus along with the network
 };

 /**
@@ -266,7 +266,7 @@ template <class Tx, class NodeId>
 void
 DisputedTx<Tx, NodeId>::unVote(NodeId const& peer)
 {
-    auto const it = votes_.find(peer);
+    auto it = votes_.find(peer);

    if (it != votes_.end())
    {
@@ -211,7 +211,7 @@ struct Node
    void
    erase(Node const* child)
    {
-        auto const it = std::ranges::find_if(
+        auto it = std::ranges::find_if(
            children, [child](std::unique_ptr<Node> const& curr) { return curr.get() == child; });
        XRPL_ASSERT(it != children.end(), "xrpl::Node::erase : valid input");
        std::swap(*it, children.back());
@@ -184,7 +184,7 @@ enum class ValStatus {
    /**
     * Multiple validations by a validator for different ledgers
     */
-    Conflicting,
+    Conflicting
 };

 inline std::string
@@ -366,7 +366,7 @@ private:
    removeTrie(std::scoped_lock<Mutex> const&, NodeID const& nodeID, Validation const& val)
    {
        {
-            auto const it = acquiring_.find(std::make_pair(val.seq(), val.ledgerID()));
+            auto it = acquiring_.find(std::make_pair(val.seq(), val.ledgerID()));
            if (it != acquiring_.end())
            {
                it->second.erase(nodeID);
@@ -375,7 +375,7 @@ private:
            }
        }
        {
-            auto const it = lastLedger_.find(nodeID);
+            auto it = lastLedger_.find(nodeID);
            if (it != lastLedger_.end() && it->second.id() == val.ledgerID())
            {
                trie_.remove(it->second);
@@ -443,7 +443,7 @@ private:
        // Clear any prior acquiring ledger for this node
        if (prior)
        {
-            auto const it = acquiring_.find(*prior);
+            auto it = acquiring_.find(*prior);
            if (it != acquiring_.end())
            {
                it->second.erase(nodeID);
@@ -455,7 +455,7 @@ private:
        checkAcquired(lock);

        std::pair<Seq, ID> const valPair{val.seq(), val.ledgerID()};
-        auto const it = acquiring_.find(valPair);
+        auto it = acquiring_.find(valPair);
        if (it != acquiring_.end())
        {
            it->second.insert(nodeID);
@@ -529,7 +529,7 @@ private:
            }
            else
            {
-                auto const cit = typename decltype(current_)::const_iterator{it};
+                auto cit = typename decltype(current_)::const_iterator{it};
                // contains a live record
                f(cit->first, cit->second);
                ++it;
@@ -554,7 +554,7 @@ private:
    void
    byLedger(std::scoped_lock<Mutex> const&, ID const& ledgerID, Pre&& pre, F&& f)
    {
-        auto const it = byLedger_.find(ledgerID);
+        auto it = byLedger_.find(ledgerID);
        if (it != byLedger_.end())
        {
            // Update set time since it is being used
@@ -857,7 +857,7 @@ public:
        if (!preferred)
        {
            // fall back to majority over acquiring ledgers
-            auto const it = std::ranges::max_element(acquiring_, [](auto const& a, auto const& b) {
+            auto it = std::ranges::max_element(acquiring_, [](auto const& a, auto const& b) {
                std::pair<Seq, ID> const& aKey = a.first;
                typename HashSet<NodeID>::size_type const& aSize = a.second.size();
                std::pair<Seq, ID> const& bKey = b.first;
@@ -937,7 +937,7 @@ public:
            return (preferred->first >= minSeq) ? preferred->second : lcl.id();

        // Otherwise, rely on peer ledgers
-        auto const it = std::ranges::max_element(peerCounts, [](auto const& a, auto const& b) {
+        auto it = std::ranges::max_element(peerCounts, [](auto const& a, auto const& b) {
            // Prefer larger counts, then larger ids on ties
            // (max_element expects this to return true if a < b)
            return std::tie(a.second, a.first) < std::tie(b.second, b.first);
@@ -156,7 +156,7 @@ public:
        {
            if (!allClosuresDoneCond_.wait_for(lock, wait, [this] { return closureCount_ == 0; }))
            {
-                if (auto const stream = j.error())
+                if (auto stream = j.error())
                    stream << name << " waiting for ClosureCounter::join().";
                allClosuresDoneCond_.wait(lock, [this] { return closureCount_ == 0; });
            }
@@ -54,7 +54,7 @@ JobQueue::Coro::post()
    }

    // sp keeps 'this' alive
-    if (jq_.addJob(type_, name_, [this, sp = shared_from_this()] { resume(); }))
+    if (jq_.addJob(type_, name_, [this, sp = shared_from_this()]() { resume(); }))
    {
        return true;
    }
@@ -130,7 +130,7 @@ inline void
 JobQueue::Coro::join()
 {
    std::unique_lock<std::mutex> lk(mutexRun_);
-    cv_.wait(lk, [this] { return !running_; });
+    cv_.wait(lk, [this]() { return !running_; });
 }

 }  // namespace xrpl
@@ -37,7 +37,7 @@ enum class HashRouterFlags : std::uint16_t {
    PRIVATE6 = 0x2000,
    // Used in apply.cpp
    PRIVATE7 = 0x4000,
-    PRIVATE8 = 0x8000,
+    PRIVATE8 = 0x8000
 };

 constexpr HashRouterFlags
@@ -30,12 +30,12 @@ private:
        using namespace std::chrono_literals;
        int const maxLimit = std::numeric_limits<int>::max();

-        auto const add = [this](
-                             JobType jt,
-                             std::string name,
-                             int limit,
-                             std::chrono::milliseconds avgLatency,
-                             std::chrono::milliseconds peakLatency) {
+        auto add = [this](
+                       JobType jt,
+                       std::string name,
+                       int limit,
+                       std::chrono::milliseconds avgLatency,
+                       std::chrono::milliseconds peakLatency) {
            XRPL_ASSERT(!map.contains(jt), "xrpl::JobTypes::JobTypes::add : unique job type input");

            [[maybe_unused]] auto const inserted =
@@ -177,13 +177,12 @@ measureDurationAndLog(
    std::chrono::duration<Rep, Period> maxDelay,
    beast::Journal const& journal)
 {
-    auto const startTime = std::chrono::high_resolution_clock::now();
+    auto startTime = std::chrono::high_resolution_clock::now();

    auto result = func();

-    auto const endTime = std::chrono::high_resolution_clock::now();
-    auto const duration =
-        std::chrono::duration_cast<std::chrono::milliseconds>(endTime - startTime);
+    auto endTime = std::chrono::high_resolution_clock::now();
+    auto duration = std::chrono::duration_cast<std::chrono::milliseconds>(endTime - startTime);
    if (duration > maxDelay)
    {
        JLOG(journal.warn()) << actionDescription << " took " << duration.count() << " ms";
@@ -95,7 +95,7 @@ private:
        ArraySeparator,
        MemberSeparator,
        Comment,
-        Error,
+        Error
    };

    class Token
@@ -3,13 +3,11 @@
 #include <xrpl/basics/Number.h>
 #include <xrpl/json/json_forwards.h>

-#include <concepts>
 #include <cstring>
 #include <iterator>
 #include <limits>
 #include <map>
 #include <string>
-#include <string_view>
 #include <vector>

 /**
@@ -28,7 +26,7 @@ enum class ValueType {
    String,    ///< UTF-8 string value
    Boolean,   ///< bool value
    Array,     ///< array value (ordered list)
-    Object,    ///< object value (collection of name/value pairs).
+    Object     ///< object value (collection of name/value pairs).
 };

 /**
@@ -200,15 +198,6 @@ public:
     */
    Value(StaticString const& value);
    Value(std::string const& value);
-    /**
-     * @brief Constructs a value from a string view.
-     *
-     * The characters are copied, so the view need not outlive the call and need
-     * not be NUL-terminated.
-     *
-     * @param value The characters to copy.
-     */
-    Value(std::string_view value);
    Value(bool value);
    Value(Value const& other);
    ~Value();
@@ -483,32 +472,6 @@ toJson(xrpl::Number const& number)
 bool
 operator==(Value const&, Value const&);

-/**
- * Compares a value with a string view, reading the value's characters in place
- * rather than building a Value from the view.
- *
- * Constrained to the exact type: a string literal converts equally well to a
- * view and to a Value, so a plain overload makes `value == "literal"`
- * ambiguous.
- *
- * @param x The value to compare.
- * @param y The characters to compare it against.
- * @return Whether `x` is a string whose characters up to its first NUL are
- *         exactly the characters of `y`.
- */
-template <class T>
-    requires std::same_as<T, std::string_view>
-bool
-operator==(Value const& x, T y)
-{
-    if (!x.isString())
-        return false;
-
-    // A string `Value` can hold a null pointer, which names no characters, so it equals no view.
-    char const* const s = x.asCString();
-    return s != nullptr && std::string_view{s} == y;
-}
-
 bool
 operator<(Value const&, Value const&);

@@ -165,7 +165,7 @@ public:
        beast::Journal j)
    {
        // Ask implementation what to do
-        auto const actions = doVoting(
+        auto actions = doVoting(
            lastClosedLedger->rules(),
            lastClosedLedger->parentCloseTime(),
            getEnabledAmendments(*lastClosedLedger),
@@ -1,7 +1,7 @@
 #pragma once

 #include <xrpl/basics/base_uint.h>
-#include <xrpl/basics/safe_cast.h>
+#include <xrpl/basics/enum_bitops.h>
 #include <xrpl/beast/utility/instrumentation.h>
 #include <xrpl/ledger/OwnerCounts.h>
 #include <xrpl/ledger/ReadView.h>
@@ -22,80 +22,39 @@

 namespace xrpl {

-// Bitwise flag enum with existing operator overloads
-// NOLINTNEXTLINE(cppcoreguidelines-use-enum-class)
-enum ApplyFlags : std::uint32_t {
-    TapNone = 0x00,
+enum class ApplyFlags : std::uint32_t {
+    None = 0x00,

    // This is a local transaction with the
    // fail_hard flag set.
-    TapFailHard = 0x10,
+    FailHard = 0x10,

    // This is not the transaction's last pass
    // Transaction can be retried, soft failures allowed
-    TapRetry = 0x20,
+    Retry = 0x20,

    // Transaction came from a privileged source
-    TapUnlimited = 0x400,
+    Unlimited = 0x400,

    // Transaction is executing as part of a batch
-    TapBatch = 0x800,
+    Batch = 0x800,

    // Transaction shouldn't be applied
    // Signatures shouldn't be checked
-    TapDryRun = 0x1000,
+    DryRun = 0x1000,

    // Transaction is being preflighted as the payload of a
    // TransactionProposalCreate. Its signatures are collected on-ledger
    // afterward, so signature-presence checks (e.g. Batch signer matching)
    // are skipped at proposal-creation time (On-Chain Cosigner spec
    // §5.3.1.2).
-    TapProposal = 0x2000,
+    Proposal = 0x2000
 };

-constexpr ApplyFlags
-operator|(ApplyFlags const& lhs, ApplyFlags const& rhs)
+template <>
+struct enum_bitops::OptIn<ApplyFlags> : std::true_type
 {
-    return safeCast<ApplyFlags>(
-        safeCast<std::underlying_type_t<ApplyFlags>>(lhs) |
-        safeCast<std::underlying_type_t<ApplyFlags>>(rhs));
-}
-
-static_assert((TapFailHard | TapRetry) == safeCast<ApplyFlags>(0x30u), "ApplyFlags operator |");
-static_assert((TapRetry | TapFailHard) == safeCast<ApplyFlags>(0x30u), "ApplyFlags operator |");
-
-constexpr ApplyFlags
-operator&(ApplyFlags const& lhs, ApplyFlags const& rhs)
-{
-    return safeCast<ApplyFlags>(
-        safeCast<std::underlying_type_t<ApplyFlags>>(lhs) &
-        safeCast<std::underlying_type_t<ApplyFlags>>(rhs));
-}
-
-static_assert((TapFailHard & TapRetry) == TapNone, "ApplyFlags operator &");
-static_assert((TapRetry & TapFailHard) == TapNone, "ApplyFlags operator &");
-
-constexpr ApplyFlags
-operator~(ApplyFlags const& flags)
-{
-    return safeCast<ApplyFlags>(~safeCast<std::underlying_type_t<ApplyFlags>>(flags));
-}
-
-static_assert(~TapRetry == safeCast<ApplyFlags>(0xFFFFFFDFu), "ApplyFlags operator ~");
-
-inline ApplyFlags
-operator|=(ApplyFlags& lhs, ApplyFlags const& rhs)
-{
-    lhs = lhs | rhs;
-    return lhs;
-}
-
-inline ApplyFlags
-operator&=(ApplyFlags& lhs, ApplyFlags const& rhs)
-{
-    lhs = lhs & rhs;
-    return lhs;
-}
+};

 //------------------------------------------------------------------------------

@@ -20,8 +20,7 @@ constexpr std::chrono::seconds kLedgerPossibleTimeResolutions[] = {
    std::chrono::seconds{30},
    std::chrono::seconds{60},
    std::chrono::seconds{90},
-    std::chrono::seconds{120},
-};
+    std::chrono::seconds{120}};

 /**
 * Initial resolution of ledger close time.
@@ -35,7 +35,7 @@ public:
    {
        std::scoped_lock const lock(mutex_);

-        auto const it = map_.find(seq);
+        auto it = map_.find(seq);

        if ((it == map_.end()) || it->second)
        {
@@ -88,7 +88,7 @@ public:
        std::unique_lock<std::mutex> lock(mutex_);
        do
        {
-            auto const it = map_.find(seq);
+            auto it = map_.find(seq);

            if (it == map_.end())
            {
@@ -164,7 +164,7 @@ getCandidateLedger(LedgerIndex requested)
 areCompatible(
    ReadView const& validLedger,
    ReadView const& testLedger,
-    beast::Journal::Stream const& s,
+    beast::Journal::Stream& s,
    char const* reason);

 [[nodiscard]] bool
@@ -172,7 +172,7 @@ areCompatible(
    UInt256 const& validHash,
    LedgerIndex validIndex,
    ReadView const& testLedger,
-    beast::Journal::Stream const& s,
+    beast::Journal::Stream& s,
    char const* reason);

 //------------------------------------------------------------------------------
@@ -185,7 +185,7 @@ areCompatible(
 dirLink(
    ApplyView& view,
    AccountID const& owner,
-    SLE::pointer const& object,
+    SLE::pointer& object,
    SF_UINT64 const& node = sfOwnerNode);

 /**
@@ -30,17 +30,13 @@ public:
    static constexpr size_t kInitialBufferSize = kilobytes(256);

    RawStateTable()
-        : monotonicResource_{
-              std::make_unique<boost::container::pmr::monotonic_buffer_resource>(
-                  kInitialBufferSize),
-          }
+        : monotonicResource_{std::make_unique<boost::container::pmr::monotonic_buffer_resource>(
+              kInitialBufferSize)}
        , items_{monotonicResource_.get()} {};

    RawStateTable(RawStateTable const& rhs)
-        : monotonicResource_{
-              std::make_unique<boost::container::pmr::monotonic_buffer_resource>(
-                  kInitialBufferSize),
-          }
+        : monotonicResource_{std::make_unique<boost::container::pmr::monotonic_buffer_resource>(
+              kInitialBufferSize)}
        , items_{rhs.items_, monotonicResource_.get()}
        , dropsDestroyed_{rhs.dropsDestroyed_} {};

@@ -1,17 +1,11 @@
 #pragma once

-#include <xrpl/basics/base_uint.h>
-#include <xrpl/basics/chrono.h>
 #include <xrpl/beast/utility/Journal.h>
 #include <xrpl/ledger/ApplyView.h>
 #include <xrpl/ledger/ReadView.h>
 #include <xrpl/ledger/entries/SLEBase.h>
 #include <xrpl/protocol/Indexes.h>
 #include <xrpl/protocol/LedgerFormats.h>
-#include <xrpl/protocol/SField.h>
-
-#include <map>
-#include <set>

 namespace xrpl {

@@ -31,52 +25,6 @@ public:
        : Base(keylet::amendments(), view, j)
    {
    }
-
-    /**
-     * Returns the set of amendments this entry reports as enabled.
-     *
-     * @return the set of enabled amendments.
-     */
-    [[nodiscard]] std::set<UInt256>
-    enabledAmendments() const
-    {
-        std::set<UInt256> amendments;
-
-        if (this->exists() && (*this)->isFieldPresent(sfAmendments))
-        {
-            auto const& v = (*this)->getFieldV256(sfAmendments);
-            amendments.insert_range(v);
-        }
-
-        return amendments;
-    }
-
-    /**
-     * Returns a map of amendments that have achieved majority, to the time
-     * majority was reached.
-     *
-     * @return a map of amendment to the time majority was reached.
-     */
-    [[nodiscard]] std::map<UInt256, NetClock::time_point>
-    majorityAmendments() const
-    {
-        std::map<UInt256, NetClock::time_point> ret;
-
-        if (this->exists() && (*this)->isFieldPresent(sfMajorities))
-        {
-            using TimePoint = NetClock::time_point;
-            using Duration = TimePoint::duration;
-
-            auto const majorities = (*this)->getFieldArray(sfMajorities);
-
-            for (auto const& m : majorities)
-            {
-                ret[m.getFieldH256(sfAmendment)] = TimePoint(Duration(m.getFieldU32(sfCloseTime)));
-            }
-        }
-
-        return ret;
-    }
 };

 using AmendmentsEntryR = AmendmentsEntry<ReadView>;
Author	SHA1	Message	Date
Nik Bougalis	ef0635d26c	Simplify and harden Serializer and SerialIter: Reduce both classes to a small, header-only interface built on Slice. Serializer becomes a strictly append-only buffer and SerialIter a strictly forward-only reader; all deprecated and duplicated paths are removed and data copying is entirely eliminated. But the wire format, you'll be happy to know, remains unchanged. Serializer: - Make the buffer append-only. Remove modData, resize, chop, mutable begin/end and getDataPtr, the (void const, size_t) constructor, and the read-side helpers get8, getInteger and getBitString. - Collapse the accessors (peekData, getData, getDataPtr, getDataLength, getLength, getString, begin/end, operator==) into slice(), data() and size(). - Add takeData(), which moves the buffer out. Use it wherever callers copied via getData/peekData or reached in with std::move(modData()). - addRaw and addVL take a Slice only; the Blob, (void const, int), Serializer and iterator-range overloads are removed. Both throw if the input aliases the serializer's own storage. - Funnel every append through one private write(), which reserves for the combined size of its parts before inserting. An add that throws leaves the serializer unchanged. - Constrain add8/add16/add32/add64 on the exact argument type: no implicit conversions, no bool or character types, and the width must match the function name. Call sites that relied on silent narrowing now cast explicitly. - The add functions return void rather than the offset written to. - Replace getSHA512Half() with a hash_append hidden friend; callers use sha512Half(s). - Move the VL length-prefix constants into xrpl::detail as std::size_t and fold header encoding into addVL. Remove addEncoded, encodeLengthLength, decodeLengthLength and decodeVLLength; decoding now lives in SerialIter::getVL. SerialIter: - Construct only, and explicitly, from a Slice. Remove the (void const*, size_t) and array-reference constructors. A null pointer with a non-zero size throws. - Remove reset() and skip(), and with them the used_ counter. A reader cannot be rewound; copy it to retain a position. - getVL() returns a Slice into the underlying buffer instead of a Blob copy. Remove getVLDataLength, getSlice, getRaw, getVLBuffer and getRawHelper. - Route every read through a single bounds-checked consume(). Integers are decoded with std::bit_cast and std::byteswap, which drops the boost::endian dependency. - Replace getBytesLeft() with size(), returning std::size_t, and add slice() for the unread remainder. - getBitString<T> takes the BaseUInt type rather than <Bits, Tag>, so tagged types (AccountID, Currency, MPTID) are read directly. - Add geti16() and mark the accessors [[nodiscard]]. Error reporting: - VL length errors, on both the encode and the decode side, now throw std::runtime_error; they previously threw std::overflow_error. - Short reads all throw std::runtime_error("short buffer") in place of the per-function messages. Other changes: - Slice: add overlap() global utility function, used for the aliasing checks above. - sqlBlobLiteral takes a Slice. - LedgerHeader: remove deserializePrefixedHeader. Its caller in InboundLedger skips the prefix locally; the prefix is still not validated, which is now marked with a FIXME. - STAccount: deserialize from the getVL slice via AccountID::fromRaw and remove the (SField, Buffer) constructor. - STVector256: serialize the whole vector with a single addVL, relying on contiguous storage and the absence of padding in UInt256. - STInteger: static_assert that the type is serializable and dispatch add() on sizeof(Integer). Read STInteger<std::int32_t> with geti32(). Log an unknown transaction result code as a number instead of a raw character. - STTx: fold the four-argument getMetaSQL overload into its only caller. - SHAMapTreeNode: the two leaf deserializers share a splitTag helper that slices off the trailing tag, replacing the copy into a Serializer followed by getBitString-at-offset and chop. The size check is done once, up front. - SHAMap::getNodeFat: serialize each node into its own Serializer and move the buffer into the reply instead of copying out of a shared one. - LedgerMaster::getCloseTimeByHash: parse the stored header with deserializeHeader instead of skipping hard-coded offsets. - ConsensusTransSetSF::gotNode and LedgerReplayMsgHandler parse directly from the incoming bytes without an intermediate Serializer copy. - TransactionSign: use sterilize() for the serialization round trip. Tests: - Port to the new interfaces; byte payloads become std::array so they can go through makeSlice. - STAccount: the bad-size case now covers both a short and a long field, and fails if no exception is thrown. Previously it passed silently in that case.	2026-10-05 19:50:58 -07:00
Nik Bougalis	3c6279b735	Restructure the slab allocator - Define size classes as template parameters, validated at compile time, allowing allocators to be constinit-constructed - Move slab metadata out of the data buffer into a cache-line-aligned global pool - Replace per-block mutexes with spinlocks, and add lock-free "parking" slots so most alloc/dealloc pairs avoid the lock entirely - Release a slab's buffer when it fully drains; cache one spare buffer to avoid thrashing around slab boundaries - Make the fallback path a compile-time policy (heap or none) - Add sized deallocate() overloads that route directly to the owning size class	2026-10-05 19:42:53 -07:00
Nik Bougalis	55b265a8c2	Improve BaseUInt: This commit reworks the BaseUInt class, which was originally taken from Bitcoin and has since been heavily modified. The internal limb type now widens from 32 to 64 bits when the requested width is a multiple of 64. This halves the iteration count when working on limbs; coupled with the use of add-with-carry intrinsics, the result is better code generation and optimized use of processor resources. The constructors have been reshuffled, constraining how an instance can be initialized. As a result, many common initialization errors will now fail at compile time instead of run time. A `BaseUInt` can now be initialized with: - A byte sequence of fixed size, i.e an `std::array`, a C array, or a fixed-extent `std::span` of `unsigned char` or `std::byte`. - A string literal, formatted as a hexadecimal string with no leading 0x. Malformed or improperly sized inputs are a compile-time error. - An unsigned integer. This constructor is mostly used for tests, and must execute at compile time. - A byte sequence whose length is not known at compile time, via the fromRaw() function which returns a seated `std::optional` only if the input is byte sequence is properly sized. The runtime raw-pointer and `std::uint64_t` constructors are gone, as are the container assignment operator and the the `fromVoid` and `fromVoidChecked` functions. Other changes: - The SFINAE-based container trait is replaced by the ByteCopySource and FixedByteRange concepts. - operator== and operator<=> are now hidden friends. The workaround in the old operator<=> is gone. - Efficient tag conversion is now possible. - `BaseUInt` is now `noexcept` and (almost) fully `constexpr`: every operation except byte access, hashing and text output is usable in constant evaluation. Cleanups: - Most invocations of `stringIsUInt256Sized` in PeerImp are now gone and leverage `fromRaw`.	2026-10-05 19:42:52 -07:00
Nik Bougalis	02e40e7a00	Restructure the spinlock code: - Add function-based spinlock API - Improve comments - Reduce bouncing with contested locks	2026-10-05 19:42:50 -07:00
Nik Bougalis	f14395427e	Modernize CountedObject infrastructure: - Replace the lazy singleton with a constinit registry of per-type static counters. The `getInstance` method is removed. - A concurrency bug that would corrupt the list of counters during insertion has been fixed. - Debug asserts can detect incorrect accounting that can result in counter under- or overflow. - `CountedObject` can only be used as a CRTP base for its own type parameter. - Track the current and maximum values of counter. This results in a change in the output of `get_counts`, with individual counters now reported as sub-objects of a `counters` object.	2026-10-05 19:42:50 -07:00
Nik Bougalis	b707bac74f	Improve AccountID base58 conversion cache: The previous cache guarded a configurably-sized vector with 64 packed spinlocks. Two aspects of that setup were not worth the cost: - Callers were forced to perform atomic RMW operations on a single word which was shared by all 64 locks; this caused the lock line to bounce between cores on every lookup. - The configurable size and runtime initialization resulted in complexity that did not yield a meaningful improvement in performance. This commit ditches the 64 packed spinlocks, replacing them with per-entry sequence locks, allowing readers on the fast path to avoid performing any stores at all. Cache entries are now carefully sized and aligned to fit into a typical cache line, which helps avoid false sharing and any unnecessary coherence traffic. The cache is now a fixed array of 65,536 entries. The size was chosen to balance capacity against memory overhead and to keep the indexing operation fast and simple: since AccountID values are uniformly distributed, two bytes can serve directly as the index without requiring additional hashing. Note: The reader path will perform unsynchronized reads against concurrent writes; the sequence lock protocol detects and discards any data observed mid-update, so these races are benign. A sanitizer like TSAN can still flag them and the warnings are expected. A compile-time knob to disable the cache is available, if necessary.	2026-10-05 19:42:50 -07:00
Nik Bougalis	da2b5d5eaf	Remove beast::typeName Replace Beast's demangling wrapper (originally by @HowardHinnant) with calls to `boost::core::type_name`, eliminating the low-level dependency and manual memory management from the codebase.	2026-10-05 19:42:50 -07:00
Nik Bougalis	4de4902200	Rework and clean up the safe_cast framework: The SafeToCast concept transposed its signedness clause: it was written in <Dest, Src> order but it declared <Src, Dest>. This introduced what can be best described as a polarity bug. As a result: * Signed-to-wider-unsigned casts were wrongly deemed safe. * Unsigned-to-wider-signed casts were wrongly deemed unsafe. The root cause was drift caused by the safety condition being repeated in the concept and again as a static_assert in safeCast. The checking is now done only in the concept and is expressed in terms of range coverage, and not via a sizeof/signedness proxy. This change also improves the handling of bool (which previously compiled despite truncating) and eliminates the platform-dependent accept/reject behavior for same-size types. Additional fixes: * The pointer form of safeDowncast now performs the same static_cast in all builds. The dynamic_cast check is now entirely contained in the XRPL_ASSERT. * Extended integer types wider than intmax_t (e.g. __int128) are now excluded from SafeToCast instead of silently misevaluating through wrapped bounds; they remain expressible via unsafeCast. Cleanups: * Remove the single-parameter enum overload of safeCast: it had an explicitly-specified template argument binding to Src, where all sibling overload bound to Dest, silently yielding the underlying type from expressions that read as casts to an enum. * Add checkedCast for conversions guarded by runtime bounds checks where safety varies per instantiation. * Constrain safeDowncast to polymorphic sources and genuine public unambiguous downcasts, rejecting upcasts and unrelated types. * Route enum conversions through std::to_underlying; clarify the unsafeCast/checkedCast documentation split.	2026-10-05 19:42:49 -07:00
Nik Bougalis	2ea1f09dcb	Modernize beast::Zero for C++20 - Replace the twelve comparison operators with two `constexpr`-capable replacements: `==` and `<=>` and allow the compiler to synthesize the rest. - Constrain `==` and `<=>` using concepts and mark them as conditionally noexcept). - Remove the detail::zero_helper indirection which did not do what its comment claimed.	2026-10-05 19:30:38 -07:00
Nik Bougalis	68b4b894de	Make ApplyFlags a scoped enumeration: Convert ApplyFlags to a scoped enum, drop the now-redundant `Tap` prefix from its enumerators and replace the hand-written bitwise operators by opting in to enum_bitops. No functional change.	2026-10-05 19:30:38 -07:00
Nik Bougalis	defd83e5a6	Make STPathElement::Type a scoped enumeration: `STPathElement` stored its type as an `unsigned int`, returned it as a `std::uint32_t`, while the flags themselves were an unscoped `enum` (i.e. an `int`). The type field has always been serialized as a single byte. This commit turns `STPathElement::Type` into a scoped `enum` with a fixed sized that matches the serialized width, and opts it into the bitwise operators from `enum_bitops.h`: - Store the type as `Type`, and use `Type` for the constructor parameter and the return type of `getNodeType()`. - Keep the existing spellings such as `STPathElement::TypeAccount` valid through `using enum`. - Replace raw mask tests in the serializer, JSON output, strand construction and pathfinder with the existing predicates. - Validate the type byte once when deserializing, before any field is read, and restructure the loop around that. - Make `operator==` a hidden friend. - Mark the trivial accessors `noexcept`. No functional changes are intended. The one observable difference is the rejection error generated for a malformed path element: if the currency and MPT bits are both set and the input is truncated within the account field, the invalid bit combination is detected first; previously the short read would be detected. Such elements are rejected either way.	2026-10-05 19:30:38 -07:00
Nik Bougalis	452e31a05f	Introduce opt-in bitwise operators for scoped enumerations: Scoped enumerations do not support bitwise operators, so using an `enum class` as a set of flags requires extensive casting at call sites or a hand-written set of operators for each type. Add `enum_bitops.h`, which provides `&`, `\|`, `^`, `~` along with the compound assignment forms for any scoped enumeration that has an unsigned underlying type and is opted in: enum class Flags : std::uint8_t { a = 1, b = 2 }; // Opt into bitwise operations template <> struct enum_bitops::OptIn<Flags> : std::true_type {}; All synthesized operators are constexpr and noexcept, and preserve the enumeration type.	2026-10-05 19:30:37 -07:00