[MarcZX10]

What

Windows module names are case-insensitive, but the audit only matched .sprx, .suprx and .debug_prx aliases when their suffix was lowercase. For example, libscea.SPRX was reported as a missing library and as a provider mismatch even when LibSceA.prx was present.

Fold the suffix case before alias matching and include .prx itself so filenames with differently cased extensions resolve to the built PE provider. The regression covers lower and upper case for all four recognized suffixes, in both registry imports and module dependencies.

Tested

Linux, synthetic PE and ELF fixtures:

• The new regression failed before the fix with exit code 1 for libscea.SPRX.
• python3 -m unittest tools.tests.test_import_audit: 151 tests passed, 1 skipped.
• python3 tools/check_conventions.py --base upstream/main: passed.
• git diff --check: passed.
• Manually reviewed the shared-file PR overlap results for fix(tools): recognize guest module identities in import audit #6171 and fix(tools): resolve $ORIGIN-prefixed converted module dependencies in import_audit #7454; their changed regions do not overlap this matching logic or regression test.

Checklist

• Based on current main; no other open PR implements the same functions
• One topic per PR; follow-ups go in a new PR
• No comments in code except technical debt
• Unimplemented paths throw (NotImplemented_nid_no_patch); silent stubs are listed under Silent stubs in their library's technical debt file
• No notes, investigation .md files or images added to the repository (attach them to this PR)
• New third-party code is a submodule built from source
• Generic behaviour, not specific to one title
• Depends on: none
• Fixes: none
• Unblocks: none
• GPUs tested: n/a
• AI-assisted: yes

[github-actions]

Some comparisons could not be completed because a pull request conflicts with the base branch. Review these shared files manually before concluding that there is no overlap.

Pull request	Shared files not compared
PR 6171	tools/import_audit.py tools/tests/test_import_audit.py
PR 7454	tools/import_audit.py tools/tests/test_import_audit.py

[MarcZX10]

Synced the branch onto current main. The branch is now 2 commits ahead and 0 behind, the merge preview is clean, and current-main conventions and diff checks pass. I reran python3 -m unittest tools.tests.test_import_audit: 151 passed, 1 skipped. The current overlap scan completed successfully; the earlier manual check of #6171/#7454 remains unchanged.

[MarcZX10]

Rebased again onto current main 48c160c after it advanced. The branch is 2 commits ahead and 0 behind; current-main sync, conventions, diff check, and merge preview pass. I reran python3 -m unittest tools.tests.test_import_audit: 151 passed, 1 skipped.

The overlap scan still cannot compare #6171 and #7454 because their branches conflict with main. I inspected both current diffs: #6171 changes ELF module identity parsing and read_modules; #7454 handles $ORIGIN dependencies in module classification/auditing. This PR changes only Windows suffix normalization in system_library and its regression test. I found no shared behavior or dependency. The hosted checks have restarted for this head; relinker matrix/ASan are pending.

[roboterspinne]

Reviewed 7d802ef merged onto main 570089a (2026-10-26); the merge is clean. The rebased head 2f2247e has the same changes, and main 88795ed hasn't touched these files or ImportModule.hpp since. Windows 11, Python 3.13.9.

Results

• test_import_audit.py: 151 tests OK, 3 skipped.
• Counter-check: with main's import_audit.py, test_windows_prx_suffix_lookup_folds_the_suffix_case fails in 5 of 8 subtests (.SPRX, .SUPRX, .DEBUG_PRX, .prx, .PRX). .sprx, .suprx and .debug_prx already pass on main.

Should fix: see the inline comment. Folding the console suffixes hides a library the Windows output cannot load.

Nits

• The lowercase .prx subtest was red on main because of the stem case (libscea against LibSceA), not the extension case: main folds the stem only for lowercase .sprx, .suprx and .debug_prx. That part of the fix is right for Windows, and the What could say so.
• Deriving normalized from folded instead of name (line 726) passes the whole suite. It would turn a match into a provider mismatch when two built providers differ only in case, for example LibSceA.prx in one --libs directory and libscea.prx in another. The exact-name preference at line 728 has no test, also not on main.

Note: nothing changes for non-Windows outputs, since windows_names is empty there.

Not checked: Linux, and whether any title has uppercase suffixes in its needed libraries or import modules.

For disclosure: none of my open PRs touches these files.

Written with AI assistance (Claude Code). I ran the tests, the counter-check and the probes above on my Windows 11 machine.

[roboterspinne]

Should fix: folding .SPRX, .SUPRX and .DEBUG_PRX makes the audit report a library as available that the Windows output can't load. The relinker maps those suffixes with a case-sensitive ends_with in Domain::NeededLibraryName (core/relinker/domain/include/domain/ImportModule.hpp:14-23): libscea.sprx becomes libscea.prx, but libscea.SPRX stays libscea.SPRX. That is the name the Windows output loads (WindowsImportBuilder.cpp:47, LoadLibraryExW in WindowsDependencyStubBuilder.cpp:251).

With only LibSceA.prx in a directory, LoadLibrary finds it as libscea.prx, libscea.PRX and LIBSCEA.PRX, and fails with error 126 for libscea.SPRX, libscea.SUPRX and libscea.DEBUG_PRX. So main's exit 1 for libscea.SPRX was right, the example in What describes correct behaviour as a bug, and the subtests for .SPRX, .SUPRX and .DEBUG_PRX assert the wrong outcome.

Folding only .prx keeps the half that is right:

                if (folded if suffix == ".prx" else name).endswith(suffix):

That fails exactly those three subtests and passes everything else, so they would need to expect a missing library (or go). If uppercase console suffixes should load, NeededLibraryName has to fold them first, in its own PR.

[THEMANnt]

Reviewed 2f2247eb8b043e9e55668d98ad6c8aaac722bb14. Approve; no actionable finding in this focused review.

Windows system-library matching folds the suffix as well as the lookup key, covering uppercase PRX, SPRX, SUPRX and DEBUG_PRX while preserving exact provider choices and unresolved-library reporting.

Validation: Windows/Python 3.11. Candidate: test_import_audit: 151 run, 3 skip reports; no failures. Comparison base: test_import_audit: 150 run, 3 skip reports. Old production fails five suffix-resolution assertions.

Scope: Synthetic PE/library resolution; no Windows loader or guest execution. Three candidate platform cases skip. Current head and effective hosted checks are verified again before submission.

[oneandonlydean]

@MarcZX10 Please fold only the .prx suffix (roboterspinne's one-line change at import_audit.py:724). The relinker maps .sprx/.suprx/.debug_prx case-sensitively in Domain::NeededLibraryName, so libscea.SPRX stays unmapped and Windows can't load it. Main's report of a missing library for it was correct. The .SPRX/.SUPRX/.DEBUG_PRX subtests should then expect a missing library (or go). If uppercase console suffixes should load, that needs NeededLibraryName to fold them first, in its own PR.

[MarcZX10]

Thanks for the correction, @oneandonlydean. I narrowed case folding to .prx filenames only; lowercase .sprx, .suprx, and .debug_prx aliases still use the relinker’s existing case-sensitive mapping, and uppercase forms remain reported as missing. I updated the regression cases and pushed commit d4a9619fdd. The full import-audit suite passes (154 tests, 1 skipped), and current-main conventions and sync checks pass.

[oneandonlydean]

@MarcZX10 Thanks, d4a9619 folds only the .prx suffix, and the uppercase .SPRX/.SUPRX/.DEBUG_PRX cases now expect a missing library. We'll run CI on this head.