buildcheck tests: write fixtures with newline="" so the suite passes on a CRLF checkout (4-repo sync)

[bernardladenthin]

Summary

• One test of the shared buildcheck suite has failed on every Windows checkout with core.autocrlf since the suite exists: test_sharedfiles.RepoEntryTest.test_the_repository_name_is_replaced_by_a_placeholder compares b'...{repo}/issues\r\n' against b'...\n'. Linux (CI) is green, so this is developer experience, not release safety -- but a permanently red suite behind a documented command (python3 -m unittest discover -s .github/buildcheck/tests -t .github in CLAUDE.md) teaches people to ignore failures, the opposite of what every other guard here is for.
• The checker is not at fault, so it is not touched. sharedfiles.content() reads with "rb" and write_manifest() writes with newline="\n", both byte-exact on purpose: a CRLF difference between repositories is exactly what the checksum must find, and normalising line endings in content() would hide it. The fixture writers in the tests omit newline=, so Python translates \n to \r\n on Windows while the assertions compare LF bytes. Every text-mode writer in the tests now passes newline="": the three in test_sharedfiles.py (lines 40, 117, 160) and, in this repository only, the one in test_nativedeps.py (line 210). They are the only four text-mode writers in the suite (checked with a grep over tests/*.py); every other fixture is written "wb".
• Cross-repo by construction. test_sharedfiles.py is line 22 of .github/shared-files.sha256, so the change goes to all four repositories with the manifest rewritten by check-shared-files.py --write in each -- which is how this was found: a change in one repository alone fails test_this_repository_matches_its_manifest and reds the shared-files job, i.e. the gate worked. Sibling PRs on the same branch name: bernardladenthin/BitcoinAddressFinder#381, bernardladenthin/srcmorph#218, bernardladenthin/streambuffer#174. Until all four are merged, the shared-files job of each warns that a sibling's default branch lists the file with a different hash; that is the documented behaviour of a sync landing in four steps, not a failure.

Test plan

• python3 -m unittest discover -s .github/buildcheck/tests -t .github green in all four repositories (here 103, the CRLF-sensitive test included; it is the Linux half -- the Windows half is the reporter's checkout, where the suite went 102/103 before).
• python3 .github/check-shared-files.py reports nothing in each repository against the local siblings, and the four copies of test_sharedfiles.py are byte-identical (sha256 fe36042f…).
• The only other change per repository is the one manifest line.
• CI green on this branch (the shared-files job may warn about the siblings until they land).

Related issues / PRs

Sibling PRs with the same change: bernardladenthin/BitcoinAddressFinder#381, bernardladenthin/srcmorph#218, bernardladenthin/streambuffer#174.

Checklist

• I have read CONTRIBUTING.md and CODE_OF_CONDUCT.md
• My commits follow Conventional Commits. They use the repository's usual imperative style instead.
• No security-sensitive changes.

🤖 Generated with Claude Code

https://claude.ai/code/session_01AytmJF9faEiQEVt6eetQS2

[sonarqubecloud]

Quality Gate passed

Issues
0 New issues
0 Accepted issues

Measures
0 Security Hotspots
0.0% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud