tool: cfg.Files; tool test: TestCpp_LLVMSystem

[xushiwei]

No description provided.

[codecov]

Codecov Report

❌ Patch coverage is 82.35294% with 3 lines in your changes missing coverage. Please review.
✅ Project coverage is 87.32%. Comparing base (c4406d2) to head (a91158e).
⚠️ Report is 21 commits behind head on main.

Files with missing lines	Patch %	Lines
tool/gen.go	82.35%	3 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #878      +/-   ##
==========================================
+ Coverage   86.65%   87.32%   +0.67%     
==========================================
  Files          22       22              
  Lines        1866     1878      +12     
==========================================
+ Hits         1617     1640      +23     
+ Misses        249      238      -11     

Flag	Coverage Δ
llgo-tests	87.32% <82.35%> (+0.67%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[fennoai]

Code Review: PR #878 — selectable header files via Files config

This PR adds a Files config field so a package config can select specific header files instead of scanning the whole Dir, and threads an explicit pkgName through NewPackage (with a backward-compatible fallback to path.Base(cfg.Name)). The core change is clean and the selected-files path is a net performance win over the full TopHeaders scan (no directory walk, no per-header include parsing).

A few small, non-blocking points are noted inline. No security or performance regressions were found — the path-traversal note is defense-in-depth only, since llcppg.cfg is a locally trusted, same-author input for this codegen tool.

I also checked the pkgName[len(pkgPrefix):] slice in gen_test.go: pkgPrefix is captured from conf.Name before Apply prepends the parent name, so for llvm + /system it correctly yields "system" with no out-of-range risk. Not a bug.

[fennoai]

listHeaderFiles passes each cfg.Files entry straight to filepath.Join(headerDir, selFile) with no validation or de-duplication. Two things differ from the pputil.TopHeaders path it replaces:

• No dedup: TopHeaders iterates a map and is inherently unique; here a duplicate entry in Files produces a duplicate top header fed into ParseSources/cl.NewPackage.
• No containment check: because filepath.Join cleans .. and strips a leading separator, entries like ../../foo.h or absolute paths resolve outside headerDir. For this local, trusted-config tool that's low risk, but rejecting absolute/escaping entries (and de-duping) would be a cheap hardening and keep the two code paths consistent.