Repository navigation
Simplify CMake and support older macOS deployment targets - #4
Conversation
Remove redundant configuration and custom source-consumer and timing checks. Preserve white-box and parity coverage, document project-specific additions, and encode Unicode test fixtures as UTF-8 bytes. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Use C-locale strtod_l on Apple platforms while retaining std::from_chars elsewhere. Preserve radix normalization and add coverage under a comma-decimal thread locale. Verified the full test suite with macOS deployment target 15.0; CI configuration is unchanged. Fixes #3 Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Replace duplicated public-header checks and license installation with cpp-library features, update dependency provenance, and retain the associated Scout knowledge. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
sean-parent
left a comment
There was a problem hiding this comment.
Talos review
Verdict: needs attention. One HIGH correctness finding survived the challenge pass.
Reviewed commit: 60cb220a872b1d870133eaa031d1b23fef85f6a0, full PR diff against main.
| Severity | Persona | Total |
|---|---|---|
| HIGH | 1 | 1 |
HIGH
| # | Location | Issue | Fix | Source |
|---|---|---|---|---|
| 1 | src/parser.cpp:183 |
Apple literal conversion follows the caller's rounding mode and can silently return the wrong binary64 value. | Use round-to-nearest during conversion and restore the caller's floating-point environment. | correctness |
Coverage
| Files | Reviewers | Candidates -> validated -> accepted | Posted | Tests |
|---|---|---|---|---|
| 16 | 6 baseline personas plus challenge | 1 -> 1 -> 1 | 1 of 1 inline | Formatting; 17/17 normal and 17/17 clang-tidy CTests; five public-header checks; 15/15 Python parity-tool tests; docs; default build/install and cpp-library artifact check |
Accepted: 1. Posted inline: 1. Duplicate recurrences: 0. Overflow: 0. Unanchored findings: 0. Rejected by challenge: 0. No automatic approval.
Persona routing: static roster (DCAP routing off; rerun with --dcap to enable).
Org knowledge: off (rerun with --knowledge).
Verification limitations
Local compiled checks used AppleClang with deployment target 15.0 on the current host, not execution on macOS 15. Hosted GCC/MSVC coverage is separate. The manually provisioned Rust-reference workflow, end-to-end C++/Rust parity, and its sanitizer/Cargo variants were not run locally. No custom installed-consumer test was added or run. The accepted finding was independently reproduced by the correctness and challenge reviewers with Darwin conversion probes; the existing passing tests do not cover nondefault floating-point rounding modes.
Scout
Used: risk MEDIUM, evidence LOW, status COMPLETE (scout 0.9.135, 1s).
Cross-file leads given to reviewers: 4 changed symbol(s) with uses outside the diff, 0 new value(s) not handled everywhere, 0 possibly unused symbol(s).
Scout rated its evidence LOW for this change; reviewers used it only as leads.
Scout evidence is incomplete (rows cut by the packet budget); reviewers treated missing evidence as unknown, not as proof.
Tests to run (Scout: 39 Must Run test(s)):
Scout left 39 affected-test entries out of its packet; for the full list run scout affected-tests -r '/Users/sean-parent/repos/github.com/stlab/cel-parser' -s compare -b origin/main.
Save and restore the caller's floating-point environment around C-locale conversion. Cover directed rounding, radix ties, range boundaries, and environment preservation with a red-green parser regression. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
sean-parent
left a comment
There was a problem hiding this comment.
Talos Review Complete
No new issues found in the full PR diff against main at cf85f7d236ac761075a41b637cc4cb1910c1059c. This is a comment review, not an approval.
The previous HIGH finding was reproduced with seven failing regression assertions, fixed in cf85f7d, and verified with the same rebuilt test. Its thread has a fix-and-verification reply and is resolved. The fix enforces round-to-nearest conversion on Apple platforms while preserving the caller's floating-point environment, including exception flags.
| Files | Reviewers | Candidates → validated → accepted | Posted | Tests |
|---|---|---|---|---|
| 17, including deletions | 6 personas, challenge | 0 → 0 → 0 | 0 of 0 | Local normal and clang-tidy: 17/17 CTests each; Python parity tooling: 15/15; formatting, docs, build/install and cpp-library package-artifact checks passed |
All six baseline personas completed: security, correctness, standards, architecture, devops, and documentation. Challenge completed with zero accepted and zero rejected observations. Dropped candidates: 0; duplicate findings: 0; inline overflow: 0. No new inline comments are needed, and the resolved finding is not duplicated.
Hosted CI at this exact head passed Ubuntu GCC, Ubuntu Clang, macOS, Windows, and clang-tidy; each job's log confirms 17 passing CTests. The release-only hosted docs job was skipped; local docs passed.
Local configurations targeted macOS 15.0 on the current newer host; this is not execution on macOS 15. The CI matrix is unchanged. The manual provisioned-reference workflow was not run; no end-to-end Rust parity or sanitizer validation is claimed. No additional installed-package consumption tests were added or run beyond cpp-library's supplied checks.
Base-revision rule discovery found no Talos rule files, applicable GitHub custom instructions, or active OpenSpec changes. No reviewers failed.
Persona routing: static roster (DCAP routing off; rerun with --dcap to enable).
Org knowledge: off (rerun with --knowledge).
Scout
Used: risk MEDIUM, evidence LOW, status COMPLETE (scout 0.9.135, <1s).
Cross-file leads given to reviewers: 4 changed symbol(s) with uses outside the diff, 0 new value(s) not handled everywhere, 0 possibly unused symbol(s).
Scout rated its evidence LOW for this change; reviewers used it only as leads.
Scout evidence is incomplete (rows cut by the packet budget); reviewers treated missing evidence as unknown, not as proof.
Tests to run (Scout: 39 Must Run test(s)):
Scout left 39 affected-test entries out of its packet; for the full list run scout affected-tests -r '/Users/sean-parent/repos/github.com/stlab/cel-parser' -s compare -b origin/main.
Summary
strtod_lon Apple platforms to support deployment targets before macOS 26; preservestd::from_charselsewhere and add a locale regression test.Fixes #3.
Test plan
Fresh local validation on AppleClang 21 with CMake/CTest 4.5.0, Ninja 1.13.2, LLVM tools 23.1.5, Doxygen 1.18.0, and Python 3.9.9. All CMake configurations below used
-DCMAKE_OSX_DEPLOYMENT_TARGET=15.0. The test and clang-tidy configurations also used-DCEL_PARSER_ENABLE_PARITY=OFFand-DCMAKE_CXX_FLAGS="-Wall -Wextra -Wpedantic -Werror".git ls-files -z '*.cpp' '*.hpp' | xargs -0 clang-format --dry-run --Werror.cmake --preset=test,cmake --build --preset=test, andctest --preset=test --no-tests=error: 17/17 tests passed, including negative-compilation tests and parity-driver unit tests; all five public-header checks compiled.cmake --preset=clang-tidy,cmake --build --preset=clang-tidy, andctest --preset=clang-tidy --no-tests=error: clean build and 17/17 tests passed.cmake --preset=docsandcmake --build --preset=docs: documentation generated without warnings.cmake --preset=default,cmake --build --preset=default, andcmake --install build/default --prefix "$PWD/build/pr-open-install", followed bycmake "-DPACKAGE_TEST_PREFIX=$PWD/build/pr-open-install" -P build/default/cpp-library-test-package.cmake: toolkit-provided artifact check passed; both license notices installed. No custom installed-consumer test.python3 -m unittest discover -s tools/parity -p 'test_*.py': 15/15 tests passed.xcrun vtool -show-build build/test/tests: binary reportsminos 15.0. This is deployment-target validation on the current host, not execution on macOS 15.git diff --check.CIrun passed at60cb220a872b1d870133eaa031d1b23fef85f6a0: Ubuntu GCC, Ubuntu Clang, macOS, Windows, and clang-tidy each ran and passed all 17 CTests. Configure/build and package-artifact steps also passed for every compiler-matrix job.Provisioned reference validationwas not run: it requires a provisioned self-hosted runner and Rust reference checkout. End-to-end C++/Rust parity and the workflow's sanitizer/Cargo variants are not claimed here.The CI matrix is unchanged.
Generated with GitHub Copilot CLI