[SizzleUnrlsd]

What and why

Updates coretrace-stack-analyzer from v0.22.0 to v0.23.1. v0.22.0 takes 30 minutes to analyze CoreTrace's own code in an unoptimized build, and v0.23.1 takes 5. The gain comes from CoreTrace/coretrace-stack-analyzer#187, which computes the ownership summaries with fewer solves and rounds; the measurements are in CoreTrace/coretrace-stack-analyzer#227.

v0.23.1 is v0.23.0 plus one fix. v0.23.0 does not build with AppleClang 15, the compiler of the macos-14 runner: the first push of this pull request failed there (CoreTrace/coretrace-stack-analyzer#228). v0.23.1 is cut from the analyzer's release/v0.23 branch. Its main also updates coretrace-compiler to v0.10.0, and CoreTrace refuses to configure with it while the concurrency analyzer stays on 866fa76 (cmake/concurrencyAnalyzer.cmake:45).

The bridge (src/Process/Tools/StackAnalyzerToolImplementation.cpp) needs no change: the report format is unchanged, and rule identifiers pass through as they are.

Closes #169

How it was tested

ctrace --compile-commands <CoreTrace's 22 translation units> --invoke ctrace_stack_analyzer, with a fresh cache and one run at a time:

Build	v0.22.0	v0.23.x
Linux x86_64, no build type (the README's commands)	29 min 39 s	5 min 18 s
Linux x86_64, Release	3 min 16 s	1 min 33 s
macOS arm64, Release, -Werror removed from the database	—	1 min 13 s

These runs used the analyzer at 0668fb8 (v0.23.0); v0.23.1 only changes how one lambda captures a variable. On macOS, -Werror has to be removed from the database: otherwise both versions stop on src/App/Config.cpp, on a deprecation warning that only the analyzer's libc++ emits (CoreTrace/coretrace-stack-analyzer#214, not fixed by this update).

With v0.23.1, on macOS arm64 (AppleClang 17, LLVM 20), Release:

• ctest --test-dir build --output-on-failure: 77/77;
• F7 harness: passes;
• ./scripts/format-check.sh (clang-format 20) and scripts/release/check-pinned-dependencies.sh pass.

With v0.23.0, ctest (77/77) and the F7 harness also pass on Linux. AppleClang 15 is checked by CI's macos-14 job.

On CoreTrace's own code, the diagnostics go from 70 infos and 3 warnings to 67 infos and 7 warnings:

• the 5 false positives on string_view parameters passed by value are gone;
• 2 new infos appear, one of them Take the LoadContext of findStackAnalyzerSection by const reference #177;
• 4 new StackPointerEscape warnings appear at include/Process/Tools/ToolsInvoker.hpp:300. They are a false positive, tracked in False positive: StackPointerEscape on the caller of a function that passes the address to a virtual call coretrace-stack-analyzer#226.

There is no new test: the defect was the running time, and the test policy keeps tests independent of timing. The existing tests cover the bridge.

Checklist

• The title is an English Conventional Commit, such as fix(cli): handle missing input.
• Tests cover the change (see the test policy). No new test: see above.
• ctest --test-dir build --output-on-failure passes.
• ./scripts/format-check.sh passes.
• The README, docs/ or the help text follow the change, if users see it.