tests: four tests and directory_named still make their own "a directory named with a byte that is not UTF-8, or skip", on any OSError, beside the shared helper #663 adds, and the test that pins make_file's UTF-8 name fails under an ASCII file-name encoding #757

Goal: every test that needs a directory named with a byte that is not UTF-8 asks the one helper whether this system can carry it, so a full disk or a permission error fails the test instead of skipping it, and a test about a skip helper is held on the encodings CI does not have.

Filed by lane openRepoShape-5, from the writer's checklist candidates of #663's PR (#723, w663, "Four tests make their own ...") and from the fix round of that PR (w663b and rv663b, "optional, v1.1.1"). Both are tests only.

Lands after: lane 5's #663 (PR #723, fix/663-non-utf8-name-tests, head 6ba8c5d2). It adds skip_where_a_name_cannot_be_not_utf_8 to tests/conftest.py (at 6ba8c5d2:tests/conftest.py:904), which the first line below calls; main has it only in tests/test_adopt_pins.py:893 today.

Lines (path:line at main 66b230ed, except the last, which is at #723's head 6ba8c5d2)

• Four tests and directory_named make the directory themselves and skip on ANY OSError. Read from source, not measured. Each one builds a name from bytes, calls mkdir, and turns OSError and UnicodeError into pytest.skip: tests/test_stdout_encoding.py:209-215 (b"b\xffd", os.mkdir, then a listing check), tests/test_adopt_git_reads.py:465-469 (caf and byte 0xE9, parent.mkdir(), then :471-473 skips again where git will not work there), tests/test_adopt_template_values.py:296-300, tests/test_adopt_plan_escapes.py:402-406 (the same, then :407-410) and tests/test_clone_url_bytes.py:136-152 (directory_named, six callers at :188, :204, :232, :245, :292 and :307). tests: one skip helper for names and branches that are not UTF-8, Windows guards on make_file and the --branch a\udcffb test by emulation, and the UTF-8 promise and the leg finding's role pinned (#663) #723's helper answers the same question in two steps and skips only on the two answers that mean "this system cannot carry the name": os.fsencode first (Windows) and EILSEQ second (macOS). Any other OSError is the test's own and is raised. These five skip on EACCES, ENOSPC and the rest, so a disk that fills up, or a tmp_path this user cannot write in, reads as SKIPPED where the helper's tests read it as a failure. Direction: call the helper for the question and keep each test's own second skip (git will not work under the name; the listing that shows a rewritten name). For directory_named, whose raw is any bytes, call the helper for the names that are not UTF-8 and keep its own Windows skip for a name holding a control character. tests/test_adopt_plan_escapes.py:462-468 asks the same question of a FILE (caf and byte 0xE9 plus .txt, write_bytes); it is the file twin, and tests: one skip helper for names and branches that are not UTF-8, Windows guards on make_file and the --branch a\udcffb test by emulation, and the UTF-8 promise and the leg finding's role pinned (#663) #723's make_file skip in tests/test_non_utf8_git_reads.py is its model.
• test_make_file_writes_a_name_that_is_utf8_where_fsencode_makes_no_byte fails where the file-name encoding is ASCII (tests/test_non_utf8_git_reads.py:120-127 at 6ba8c5d2). Measured twice, by w663b at 0b1d1ddb and by rv663b at 6ba8c5d2, under PYTHONUTF8=0 LC_ALL=C PYTHONCOERCECLOCALE=0 (1 failed, 6 passed, 2 skipped): make_file cannot write caf and U+00E9 plus .yml, skips with "this filesystem cannot hold a file name that is not UTF-8: 'ascii' codec can't encode character", and the test's assert skip_message(make_file, target, ...) is None fails at :126. The neighbouring control, :130-142, already ends if reason is not None and "filesystem" in reason: pytest.skip(reason). Direction: the same two lines in this test. CI never has that encoding, so nothing is red today; a person who runs the suite under a C locale sees one failure that is not a defect.

Verify: tests/test_stdout_encoding.py, tests/test_adopt_git_reads.py, tests/test_adopt_template_values.py, tests/test_adopt_plan_escapes.py, tests/test_clone_url_bytes.py and tests/test_non_utf8_git_reads.py pass on Linux, and tests/test_non_utf8_git_reads.py passes under PYTHONUTF8=0 LC_ALL=C PYTHONCOERCECLOCALE=0. For the first line, one test per file under monkeypatch.setattr(os, "mkdir", <raises PermissionError>) that asserts a failure and not a skip, as #723's tests/test_non_utf8_name_helpers.py does for the helper itself. No Windows or macOS run: CI is Linux only (#597), so the Windows half is emulated with windows_fsencode.

Overlaps. Open branches that touch these files: fix/663-non-utf8-name-tests (#723) edits tests/conftest.py and tests/test_non_utf8_git_reads.py; fix/672-bump-leg-branch-and-push-line (#746) also edits tests/test_non_utf8_git_reads.py (three dirty-list expectations); fix/693-local-remote-dir (96ebe450) edits tests/test_clone_url_bytes.py. Nothing on origin touches tests/test_stdout_encoding.py, tests/test_adopt_git_reads.py, tests/test_adopt_template_values.py or tests/test_adopt_plan_escapes.py. A line on #459 (issuecomment 6214823552, R433 O5) says directory_named skips every bad-name test on Windows, reasoned and not run; it is a different question about the same function, so decide it with the first line here. No pinned file changes.

Parent: #461.

Milestone: v1.1.1. Tests only; nothing a user runs changes.

🤖 Generated with Claude Code