[AlexFigas]

What

__fpclassifyf returned FreeBSD's <math.h> codes: FP_INFINITE 0x01, FP_NAN 0x02, FP_NORMAL 0x04, FP_SUBNORMAL 0x08 and FP_ZERO 0x10. #1987 took them from FreeBSD's fpclassify.c. The console's own __fpclassifyf returns the Dinkumware codes instead: 0 zero, −2 subnormal, −1 normal, 1 infinite and 2 NaN. Those are the codes that AnyPS5's __fpclassifyd, __fpclassifyl and _FDtest already return.

So a title's fpclassify on a float gave the wrong answer for every finite value, while the same call on a double was right. Only infinities and NaNs agreed, because FreeBSD and Dinkumware happen to share 1 and 2 there.

The evidence is the libc.prx bundled with PPSA21564, which exports __fpclassifyf (z7aCCd9hMsI) at 0x54790. That code is the library's own:

vmovd ecx, xmm0
eax = ecx >> 23; if al == 0xff: return (ecx & 0x7fffff) != 0 ? 2 : 1
if al != 0: return -1
return (ecx & 0x7fffff) == 0 ? 0 : -2

This is _FDtest for every input, so __fpclassifyf now forwards to it.

Tested

On Ubuntu 24.04 under WSL2 (GCC 13.3, Release):

• Native comparison. I used the harness from fix(libc): compute acosl with FreeBSD msun's ld80 code as the bundled libc.prx does #11260, with that libc.prx's segments mapped into a Linux process. It runs the binary's __fpclassifyf and AnyPS5's on all 4,294,967,296 float bit patterns:

  • with this PR: 0 mismatches;
  • on main: every pattern that is not an infinity or NaN differs.

  The same run finds no mismatch for __fpclassifyd or __fpclassifyl on 20,000,000 random encodings each, which confirms that the double and long double versions already used these codes.

• guest_math passes. Its _FDtest case table covers both zeros, the extreme subnormals, normals, both infinities, and quiet and signalling NaNs. It used to require the FreeBSD constants from __fpclassifyf; it now requires the same code as _FDtest.

• Conventions. python3 tools/check_conventions.py --base origin/main passes.

• Overlap. python3 tools/pr_overlap.py --branch fix/libc-fpclassifyf-dinkumware-codes reports no line overlap. It could not compare 12 PRs that conflict with main (feat(libc): implement the nextafter and nexttoward families #6295, Macos port #7201, fix(libc): match lgamma's signgam for -0 and its errno at poles as measured on a PS5 #7544, feat(libc): implement tgamma, lgamma, and gamma math functions #8417, fix(libc): compute asinf with FreeBSD msun's algorithm, as titles' libc.prx does #9771, fix(libc): compute log10f and log2f with FreeBSD msun 9.0's algorithms, as titles' libc.prx does #9873, fix(libc): follow msun for log1pf #10007, fix(libc): compute expf with FreeBSD msun's algorithm, as titles' libc.prx does #10113, fix(libc): compute erff and erfcf with fdlibm's algorithm, as titles' libc.prx does #10226, fix(libc): preserve tgammaf underflow sign #10467, fix(libc): preserve acoshf domain errno and exception flags #10610 and feat(libc): implement _LCosh and _LSinh #10742). For each of them, git merge-tree against main and against this branch gives the same conflict count in Math.cpp and GuestMath.cpp, so this branch adds no conflict.

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

[MarcZX10]

Reviewed exact head 40619ba1cc683093d75fd6450e052a258673522c. The one-line forwarding change makes __fpclassifyf return the same Dinkumware codes as the existing _FDtest helper, and the updated guest test checks that mapping. The PR reports an exhaustive comparison against the bundled libc.prx across all 2^32 float encodings; I did not independently rerun that harness. No blocking code finding. Policy and conventions checks pass; full product builds were skipped under the PR path policy, and my documented manual dispatch attempt was rejected by GitHub because this account lacks repository admin rights.

[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 6085	core/libs/prx/libc/src/Math.cpp
PR 6295	core/libs/prx/libc/src/Math.cpp core/libs/tests/GuestMath.cpp
PR 8417	core/libs/prx/libc/src/Math.cpp core/libs/tests/GuestMath.cpp
PR 9867	core/libs/prx/libc/src/Math.cpp core/libs/tests/GuestMath.cpp
PR 9873	core/libs/prx/libc/src/Math.cpp core/libs/tests/GuestMath.cpp
PR 10113	core/libs/prx/libc/src/Math.cpp core/libs/tests/GuestMath.cpp
PR 10222	core/libs/prx/libc/src/Math.cpp core/libs/tests/GuestMath.cpp
PR 10226	core/libs/prx/libc/src/Math.cpp core/libs/tests/GuestMath.cpp
PR 10229	core/libs/prx/libc/src/Math.cpp core/libs/tests/GuestMath.cpp
PR 10245	core/libs/prx/libc/src/Math.cpp core/libs/tests/GuestMath.cpp
PR 10247	core/libs/prx/libc/src/Math.cpp core/libs/tests/GuestMath.cpp
PR 10250	core/libs/prx/libc/src/Math.cpp core/libs/tests/GuestMath.cpp
PR 10253	core/libs/prx/libc/src/Math.cpp core/libs/tests/GuestMath.cpp
PR 10265	core/libs/prx/libc/src/Math.cpp core/libs/tests/GuestMath.cpp
PR 10274	core/libs/prx/libc/src/Math.cpp core/libs/tests/GuestMath.cpp
PR 10283	core/libs/prx/libc/src/Math.cpp core/libs/tests/GuestMath.cpp
PR 10328	core/libs/prx/libc/src/Math.cpp core/libs/tests/GuestMath.cpp
PR 10345	core/libs/prx/libc/src/Math.cpp core/libs/tests/GuestMath.cpp
PR 10349	core/libs/prx/libc/src/Math.cpp core/libs/tests/GuestMath.cpp
PR 10510	core/libs/prx/libc/src/Math.cpp core/libs/tests/GuestMath.cpp
PR 10529	core/libs/prx/libc/src/Math.cpp core/libs/tests/GuestMath.cpp
PR 11109	core/libs/prx/libc/src/Math.cpp core/libs/tests/GuestMath.cpp

[wachi35676]

Windows run of head 40619ba1c merged onto main 0f34ebb85, WinLibs GCC 15.2.0, Release. The description's runs are under WSL.

• guest_math passes on Windows. With main's Math.cpp it aborts with Guest __fpclassifyf failed for 00000000.
• fix(libc): return the sign mask from __signbit, __signbitf and __signbitl #11278 merges cleanly on top of this one, and fix(libc): compute exp2l with FreeBSD msun's ld80 code as the bundled libc.prx does #11266 merges cleanly in either order.

AI-assisted: yes (Claude Code); results above were run on this machine.

[babyhuey]

Windows run of this head 40619ba1cc6, merged onto main 0a570f617d8 (clean merge, local only), with the selection CI's skipped Windows product job uses: Windows 11, Ryzen 7 5800X, RTX 3080, WinLibs GCC 15.2, Release, BUILD_TESTING=ON. The tree was configured fresh after the merge, then select_tests.py select / ninja -f select.ninja selected / --target libs / select_tests.py ctest --parallel 4 --timeout 120.

• Selected 2547 tests: 1838 pass, 1 fails, guest_raise_exception, which also fails on 0a570f617d8 itself on this machine (crash or timeout under --parallel 4; it passed on 27ad7244d22, and I'm narrowing that down separately), 708 skip. No failures from this PR.

AI-assisted: yes (Claude Code)