[xdCloudy]

What

Normalize _nid_postfix and _nid_disambigN export names in PR overlap comparisons so aliases that resolve to the same NID are detected.

Tested

Windows local Python: python -m unittest tools.tests.test_pr_overlap_exports tools.tests.test_pr_overlap_wrapped_exports passed (10 tests). git diff --check passed. The normalized export check covered 813 open PRs; 3 unrelated diffs could not be retrieved.

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; silent stubs are documented (not applicable)
• No notes, investigation .md files or images added to the repository
• New third-party code is a submodule built from source (not applicable)
• Generic behaviour, not specific to one title
• Depends on: none
• Fixes: none
• Unblocks: none
• GPUs tested: n/a
• AI-assisted: yes

[MarcZX10]

The current Conventions job fails for two fixable reasons: the three-line explanatory comment above the canonicalization in tools/pr_overlap.py violates the repository rule that code comments belong only to technical debt, and the PR body is missing the required ### What, ### Tested, and ### Checklist sections. The code change itself is focused; moving the rationale into the PR description and filling the template should clear these findings. relinker-asan and the Ubuntu relinker job pass; macOS/Windows relinker jobs are still pending.

[xdCloudy]

I found a gap in my overlap checker: it compared export names literally and did not normalize _nid_postfix / _nid_disambigN aliases, which can resolve to the same NID as a plain name. I updated the checker to compare normalized names and checked this PR against current main and the available open PR diffs; no remaining normalized-name overlap turned up for this PR.

[MarcZX10]

I independently ran the focused tests from exact head 3a1acb576b: python3 -m unittest tools.tests.test_pr_overlap_exports tools.tests.test_pr_overlap_wrapped_exports passes (10 tests). The remaining red check is the conventions/template failure above; I used an isolated temporary worktree and removed it afterward.

[MarcZX10]

The author removed the disallowed implementation comment and completed the required PR sections; the latest Conventions check passes. I independently ran the focused overlap tests on the previous head and all 10 passed; the implementation changes since then are documentation/comment cleanup. The cross-platform relinker jobs are still pending, so this approval reflects the focused code and convention review, not those unfinished jobs.

[MarcZX10]

Thanks for fixing the template and comment issue. The current Conventions check is green. My 10-test focused run was on the preceding code-equivalent head; the remaining platform jobs are still running.

[oneandonlydean]

@xdCloudy thanks, the alias check reads right and its normalisation matches tools/nid_names.py. A non-blocking idea for a follow-up, not this PR: pr_overlap.py compares open pull requests with each other, but the duplicates we have been finding in the stub PRs are against main itself, where a plain-name stub exports the same NID as an existing _nid_postfix implementation. A check of each PR's new exports against main's (including APS5_ASM_FUNCTION and raw APS5_EXPORT NIDs) would catch those automatically. Full CI on this head is next.

[xdCloudy]

Thanks, agreed. This PR stays focused on normalizing aliases between open PRs. I separately checked this batch against current main, including APS5_VABI definitions, APS5_ASM_FUNCTION exports, and named APS5_EXPORT aliases; no matches turned up. An automated PR-versus-main export inventory check would be a useful separate follow-up.

[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 988	tools/pr_overlap.py

[xdCloudy]

I manually reviewed #988's change to tools/pr_overlap.py. It adds end-of-run summary logging near the script entry point; this PR changes export-name normalization in exports(). The edits are independent, and a direct three-way merge check found no conflict in tools/pr_overlap.py. The bot warning reflects #988's conflict with its base branch, so this file should combine cleanly after the branches are rebased.

[THEMANnt]

Reviewed at 39b0c5cb4, with both commits cherry-picked onto main 44e3389c8 on Windows.

• Full test_pr_overlap*.py suite: 61 passed, 4 platform-specific skips (65 run).
• Compared 36 alias/suffix cases against the production C++ Nid::Internal::StripNidPostfix, compiled with GCC 15.2; all matched.
• Three overlaps() probes reproduce missed collisions on main and detect them with this change. Distinct names and no-patch helpers remain excluded.
• Conventions and whitespace checks pass.

No actionable findings. Linux-only cases and a live exhaustive PR audit were not run.

AI-assisted: yes (Codex); results above were run locally.