mirror of
https://github.com/XRPLF/rippled.git
synced 2026-09-27 15:28:03 +00:00
504138dd9c
The acquire span opens three phase children -- header, account-state tree and transaction tree -- and the contract documented all three parentings while checking none of them. astree had an entry marked skipped; header and txtree had no entry at all. The skip reason was that the parent is optional, because a healthy 5-node cluster agreeing from genesis rarely back-fills history, so the hierarchy check would fail against a parent with no traces. Run 32969481032 refutes the premise: ledger.acquire reported 5 traces and so did each of the three children, one per node. The parenting was never the uncertain part -- beginPhaseSpan() parents through the acquire span's own captured SpanContext rather than the ambient thread context, so it holds whichever worker opens a phase. Asserting these matters because of what the phases are for. A fresh sync is dominated by the account-state tree, and the flat parent span cannot separate that from the much smaller transaction tree or from the header wait that gates both. If a phase stops nesting under the acquire it still emits, still carries its missing-node count and its timeout flag, and nothing else in this harness notices -- but the trace stops answering which phase the sync is stuck in, which is the whole reason these spans exist. Routed here rather than to phase-10 because phase-10 has no ledger.acquire.header or ledger.acquire.txtree span at all; the phase children were introduced on this branch. Verification: JSON parses; 10 relationships, 6 asserted and 4 skipped, no duplicates, every non-wildcard child resolves to a declared span entry; counters unchanged at 48 span types and 74 unique attributes; otel-naming exits 0; pre-commit clean. Edited by surgical text replacement -- a first attempt used a json.dumps round-trip and reflowed the whole file, 148 insertions against 47 deletions with unrelated compact arrays expanded and unrelated notes rewritten; that was reverted and redone, and the churn is now 11 against 3. Whether a child is findable INSIDE the parent's fetched trace is what CI will decide.