[tkcm01]

What

fflush_nid_postfix skips the host fflush for a stream that is not writing only on Linux (__fwriting, #4372). On macOS the host fflush of a stream that is reading seeks the descriptor back and drops the read buffer together with a pushed-back character, so after fgetc → 'a', ungetc('Z') and fflush on "abcdef\n", the next fread returned "\n" instead of "Zbcdef\n" (guest_standard_streams, CheckFlushWhileReading, for both "r" and "r+").

FreeBSD's __sflush returns without touching a stream whose flags lack __SWR. macOS's stdio keeps the same flag in FILE::_flags (<_stdio.h>), so the macOS branch tests it there.

The same two lines are part of the macOS port in #7201 (@marion909, credited as co-author); this splits them out as a small change, as for the rest of the macOS series on #1.

Tested

macOS 27.0.1, Apple M4, Apple clang 21, x86-64 build under Rosetta 2. The libraries do not link on macOS from main alone, so the run is on mugurc's ci/macos-series-2 (53db0ba, main 3fd1cc3) with #7524, the code commits of #7937 and the libSceSsl change from #7332, the setup described on #1. No Vulkan loader, so the AGC/GPU tests are skipped.

• Before: guest_standard_streams fails with fflush of a "r" stream that is reading changed the data read next: "\n" (1 bytes), expected "Zbcdef\n" (7 bytes).
• After: guest_standard_streams passes.
• Full ctest -j4: 1648 tests, 14 failed, all of which also fail without this change (15 without it, the 15th being guest_standard_streams); no new failure.
• Linux and Windows code is unchanged (#elif defined(__APPLE__)).
• python3 tools/check_conventions.py --base upstream/main: passed. tools/pr_overlap.py lists fix(libs): share Sony CRT handler and lconv state between libc and libSceAmpr #7081 and Macos port #7201 for FileAndHeap.cpp; fix(libs): share Sony CRT handler and lconv state between libc and libSceAmpr #7081 does not touch fflush_nid_postfix, Macos port #7201 contains this same change.

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

🤖 Generated with Claude Code

[MarcZX10]

The Apple-only _SWR guard is narrowly scoped and matches the read-stream behavior this test exercises. The described macOS before/after result is directly relevant, and current PR checks pass for conventions and the focused register-context jobs. I did not rerun the macOS test locally. One coordination point: #7201 currently contains the same guard; if this focused PR lands first, that hunk should be removed from #7201 during its rebase.

[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 7201	core/libs/prx/libc/src/FileAndHeap.cpp
PR 9526	core/libs/prx/libc/src/FileAndHeap.cpp
PR 9527	core/libs/prx/libc/src/FileAndHeap.cpp

[tkcm01]

Thanks for the review. Same note for #8658 (@iamthehimansh): it carries the same __SWR guard in fflush_nid_postfix, so if this one lands first that hunk can be dropped from #8658 on rebase, as for #7201.

[maximilianfv]

Independent test: Apple M1 Pro, macOS 27.0.1

Apple M1 Pro, macOS 27.0.1, Apple clang 21, x86-64 Release build under Rosetta 2, BUILD_TESTING=ON, no Vulkan loader. Base: mugurc's ci/macos-series-2 at 27e968974, with this pull request's commit cherry-picked (together with #9971 and #9972, which touch other files).

• guest_standard_streams: fails on the series alone (fflush of a "r" stream that is reading changed the data read next), and passes with this pull request in 2 of 2 runs.
• Full ctest -j 6 (mic tests excluded): 15 failures with this change against 20 without, and no test that passed before fails now. The other differences are precise_wait and guest_refused_connect, which are flaky here either way.

AI-assisted: yes (Claude Code ran the build and the tests).