https://bugs.documentfoundation.org/show_bug.cgi?id=174268
--- Comment #2 from [email protected] --- I asked codex to investigate the root cause and suggest a simple fix. I don't expect it to meet your quality requirements, but I hope it saves you time. ----- Source analysis identifies a batching defect. lcl_probeLeftOrRightFGs() admits the active, already rejected E group while calculating the independent D group. Its speculative dependency probe reaches B4/C3 and re-enters the running A3 calculation, producing false Err:522. The proposed correction applies the original group’s cycle/disabled eligibility checks before admitting neighboring groups. Independent threaded calculation remains available. The batching path dates to 845e1cdca334 (2019-10-01; present in 6.4 source), not a binary-bisected introduction. Related: bugs 124270 and 132451; exact duplication remains unconfirmed. On locally patched 27.2.0.0.alpha0+ master 7528c0d40601, macOS 27.0.1 ARM64, all 500 cells pass three load/hard-recalculation cycles per threading setting. Both parallelism suites pass all 20 tests. Reverting restores the 392 incorrect cells. Patched Linux was not tested. --- a/sc/source/core/data/formulacell.cxx +++ b/sc/source/core/data/formulacell.cxx @@ -5379,6 +5379,10 @@ if (!xNGroup) break; + // Neighbor groups must satisfy the same eligibility checks as the entry group. + if (xNGroup->mbPartOfCycle || xNGroup->meCalcState == sc::GroupCalcDisabled) + break; + if (!pCell->GetCode()->IsEnabledForThreading()) break; -- You are receiving this mail because: You are the assignee for the bug.
