Repository navigation
checkpoint/verify/rebuild_indexes on a thread holding a WriteTransaction deadlock instead of INT-001 #734
Description
Problem
Minigraf::checkpoint, verify and rebuild_indexes take inner.write_lock without the is_write_tx_active() check that execute, query, fact_log and begin_write make. Called on a thread that holds a WriteTransaction, they lock a std::sync::Mutex the same thread already holds: a deadlock (or a panic, as std leaves same-thread re-locking unspecified), instead of INT-001.
let db = Minigraf::open("x.graph")?;
let tx = db.begin_write()?;
db.checkpoint()?; // hangs- v3:
src/db.rscheckpoint_inner(~1102),verify_inner(~1138),rebuild_indexes_inner(~1183) - main (v2.x):
checkpoint_inner(src/db.rs:729);verify/rebuild_indexesare v3-only
Minigraf::prepare, PreparedQuery::execute and PreparedQuery::query also skip the check. They take no write lock, so they do not deadlock, but they are inconsistent with execute/query.
Fix
Add the is_write_tx_active() → INT-001 check to each, with a test per method. Found while reviewing the #180 spec (PR #730), whose re-entry rules depend on these guards.
🤖 Generated with Claude Code
Activity
docs(spec): #180 review: #734 guards, registry snapshot is new, uniff…
docs(plan): #734 branch starts after the plan merges
docs(#734): INT-001 names PreparedQuery's methods apart; note the blo…
test(#734): note the blocked thread on a deadlock
fix(db): checkpoint inside a WriteTransaction returns INT-001 instead…
Merge pull request #739 from project-minigraf/fix/734-tx-guards
adityamukho commentedon Oct 24, 2026
Fixed on
v3in #739 (7f90e95) and onmainin #740 (b3b3975). Stays open until the v2.x patch with the fix is published.release: v2.0.11
adityamukho commentedon Oct 26, 2026
Fixed in v2.0.11, now published: https://github.com/project-minigraf/minigraf/releases/tag/v2.0.11. The v2.x fix shipped in #740; v3.0.0 is also fixed in #739. Updated the pinned known-issues list (#421).