- Add Doxygen documentation to new methods, members, constants, and
macros introduced by the TSAN/ASAN PR (Database::startReadThreads,
BasicApp::DeferStart, ResourceManager::start, XRPL_SANITIZER_ACTIVE,
coroStackSize, yieldStackSize)
- Add @note thread-safety tags where atomics were introduced
- Update Database constructor docs to reflect deferred thread startup
- Remove temporary CI filter that restricted builds to sanitizer-only
variants (must run full matrix before merge)
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Clang doesn't have the 2GB shadow memory collision that GCC has,
so keep ASAN+UBSAN and TSAN+UBSAN combined for Clang-20. Only GCC-13
needs separate ASAN, TSAN, and UBSAN jobs.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Combined ASAN+UBSAN instrumentation inflates data sections past 2GB,
colliding with ASAN's fixed shadow memory layout at 0x7fff8000+.
Split into independent ASAN, TSAN, and UBSAN jobs per compiler to
keep each binary within the shadow memory limit.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Change -mcmodel=large to -mcmodel=medium for GCC ASAN builds. The large
model inflates code size with 64-bit absolute addresses, pushing the
binary past the 2GB limit where GCC's pre-compiled CRT startup code
(crtstuff.c) can't reach data sections with 32-bit relocations.
Medium model keeps code compact (CRT-compatible) while allowing data
beyond 2GB.
- Add nproc_subtract=20 for TSAN builds to reduce build parallelism.
TSAN instrumentation significantly increases per-compilation-unit memory
usage, causing OOM on CI runners with high parallelism.
- Plumb nproc_subtract from the CI matrix through to the build workflow
so individual configurations can control their parallelism.
- Temporarily restrict CI matrix to only the two previously-failing
configs (gcc-13 asan-ubsan, clang-20 tsan-ubsan) to validate fixes
without burning CI resources on passing configs.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Fix BasicApp::startIOThreads() member corruption: replace
while(numberOfThreads_--) with a for loop to preserve the member value.
- Fix Coro.ipp non-sanitizer stack size: increase default from 1MB to 2MB
to match yield_to.h and prevent stack overflows in deep call chains.
- Remove stale "TSAN deactivated" comment and dead activate_tsan variable
from CI matrix generator.
- Clarify Application.cpp setup() comment to distinguish io_context threads
from subsystem-specific threads.
- Use explicit load(std::memory_order_relaxed) at all SHAMap::ledgerSeq_
read sites for consistency with the atomic store.
- Extract sanitizer detection into XRPL_SANITIZER_ACTIVE macro in
sanitizers.h, replacing duplicated preprocessor blocks in Coro.ipp
and yield_to.h.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Make Journal::Sink::thresh_ atomic to fix data race between
concurrent log threshold reads (hot path) and writes (RPC/startup)
- Make SHAMap::ledgerSeq_ atomic to fix data race between concurrent
node fetch reads and ledger acquisition writes
- Suppress known lock-order-inversion between getMasterMutex() and
RCLConsensus::mutex_ (requires larger lock-ordering redesign)
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
This change deletes the `SecretKey` equality/inequality operators from the public library header and moves the comparison logic into test-only code.
Specifically, the `operator==` and `operator!=` free functions on `SecretKey` have been removed from `include/xrpl/protocol/SecretKey.h` and have been replaced with explicitly deleted member functions to prevent accidental use in production code. A named `test::equal()` helper has also been added in `src/test/unit_test/utils.h` for test assertions that need to compare secret keys.
The rdb module was not properly designed, which is fixed in this change. The module had three classes:
1) The abstract class `RelationalDB`.
2) The abstract class `SQLiteDatabase`, which inherited from `RelationalDB` and added some pure virtual methods.
3) The concrete class `SQLiteDatabaseImp`, which inherited from `SQLiteDatabase` and implemented all methods.
The updated code simplifies this as follows:
* The `SQLiteDatabaseImp` has become `SQLiteDatabase`, and
* The former `SQLiteDatabase `has merged with `RelationalDatabase`.
This change modularizes the `WalletDB` and `Manifest`. Note that the wallet db has nothing to do with account wallets and it stores node configuration, which is why it depends on the manifest code.
In certain cases, such as when modifying headers used by many compilation units, performing a unity build is slower than when performing a regular build with `ccache` enabled. There is also a benefit to a unity build in that it can detect things such as macro redefinitions within the group of files that are compiled together as a unit. This change therefore restores the ability to perform unity builds. However, instead of running every configuration with and without unity enabled, it is now only enabled for a single configuration to maintain lower computational use.
As part of restoring the code, it became clear that currently two configurations have coverage enabled, since the check doesn't focus specifically on Debian Bookworm so it also applies to Debian Trixie. This has been fixed too in this change.
Unity builds were intended to speed up builds, by bundling multiple files into compilation units. However, now that ccache is available on all platforms, there is no need for unity builds anymore, as ccache stores compiled individual build objects for reuse. This change therefore removes the ability to make unity builds.
Currently we're passing the `Application` object around, whereby the `Application` class acts more like a service registry that gives other classes access to other services. In order to allow modularization, we should replace `Application` with a service registry class so that modules depending on `Application` for other services can be moved easily. This change adds the `ServiceRegistry` class.
`PeerImp` processes `TMGetObjectByHash` queries with an unbounded per-request loop, which performs a `NodeStore` fetch and then appends retrieved data to the reply for each queried object without a local count cap or reply-byte budget. However, the `Nodestore` fetches are expensive when high in numbers, which might slow down the process overall. Hence this code change adds an upper cap on the response size.
This change removes the `master` branch as a trigger for the CI pipelines, and updates comments accordingly. It also fixes the pre-commit workflow, so it will run on all release branches.
This change adds support for sanitizer build options in CI builds workflow. Currently `asan+ubsan` is enabled, while `tsan+ubsan` is left disabled as more changes are required.
This change fixes the last of the spelling issues, and enables the pre-commit (and CI) check for spelling. There are no functionality changes, but it does rename some enum values.
Right now, each pipeline invocation builds the source code from scratch. Although compiled Conan dependencies are cached in a remote server, the source build objects are not. We are able to further speed up our builds by leveraging `ccache`. This change enables caching of build objects using `ccache` on Linux, macOS, and Windows.
This change renames all occurrences of `rippled.cfg` to `xrpld.cfg`. It also provides a script to allow developers to replicate the changes in their local branch or fork to avoid conflicts. For the time being it maintains support for `rippled.cfg` as config file, if `xrpld.cfg` does not exist.
This change renames all occurrences of `namespace ripple` and `ripple::` to `namespace xrpl` and `xrpl::`, respectively, as well as the names of test suites. It also provides a script to allow developers to replicate the changes in their local branch or fork to avoid conflicts.
