fix(#341/#363): OwnAutoNamingScope — per-instance override, not a global mutex — v1.15.15

[gsdali]

Summary

• Follow-up to NCollection race under parallel execution: documented downstream as 'run --no-parallel', never filed or characterised upstream — needs the #298 TSan protocol #341 (v1.15.5) after upstream review on OCCT#1388. Maintainer gkv311 pointed out AutoNamingScope's recursive_mutex only serialized the three known override call sites against each other — every other read of theAutoNaming in XCAFDoc_ShapeTool.cxx (AddShape, MakeReference, SetSHUO) stayed outside any scope, so an unrelated unscoped caller on another thread could still observe another thread's temporary override. Making the flag atomic closed the memory-safety gap, not the logical one.
• Root design issue: theAutoNaming was never meant to express per-document intent — the three overriding call sites each want to suppress naming for their own document's build, and XCAFDoc_ShapeTool is already one instance per document, so the override belongs there.
• Fix: XCAFDoc_ShapeTool::OwnAutoNamingScope replaces AutoNamingScope, saving/restoring a per-instance myOwnAutonaming field instead of the shared flag — no locking needed, and XCAFDoc_Editor::Expand()'s self-recursion still composes correctly since each nested scope restores exactly the override state it observed on entry, not an unconditional reset. Patch 0011 updated in place (same fix, corrected design, not a new patch number).
• Verified: same TSan stress as the original fix, zero theAutoNaming races. New isolation scenario (Scripts/repro/363-own-autonaming/) checks the property the mutex fix couldn't guarantee: half the threads override locally while the other half do unscoped AddShape() on independent documents, concurrently — 3000 operations, zero leaks.
• Full production OCCT.xcframework rebuild (macOS, iOS device, iOS simulator). OCCTBridge.xcframework also rebuilt — ABI-only, XCAFDoc_ShapeTool grew a field so the previously-compiled bridge static library needed relinking against the new kernel layout, even though no bridge source changed.
• Docs: CLAUDE.md Known OCCT Bugs, docs/CHANGELOG.md, docs/thread-safety.md, Scripts/patches/README.md, okf/references/carried-occt-patches.md, Scripts/repro/341-meshcaf/README.md.
• Upstream: OCCT#1388 updated to the new design and re-reviewed — CI green across all 3 platforms, every build/GTest/regression/test job.

Test plan

• swift build clean (source bridge, rebuilt kernel)
• OCCTSWIFT_BRIDGE_PREBUILT=1 swift build clean (rebuilt bridge + rebuilt kernel)
• Issue341MeshCafThreadSafetyTests/Issue361SharedSingletonThreadSafetyTests/Issue359 pass on both build paths
• Full swift test — 4427 tests / 1284 suites, clean

Closes #363

🤖 Generated with Claude Code