[JJenkx]

What

scalbn and scalbnf now leave errno unchanged on every host, as the libc.prx that titles ship does.

Both call the host std::scalbn. On Windows that never sets errno. glibc's scalbn and scalbnf set ERANGE on overflow and on underflow to zero, as rooklz found while reviewing #12676. So on Linux both functions changed errno where the titles don't. They now save and restore errno around the host call, as scalbln, scalblnf and scalblnl already do with ErrnoGuard in MathExact.cpp, and as fmod does.

What the titles do, read natively from PPSA07923's libc.prx (a scratch loader calls its functions by NID), with errno preset to 7:

call	titles result	titles errno
scalbn(1, 2000)	inf	unchanged
scalbn(1, -2000), scalbn(1, -1075)	0	unchanged
scalbn(1, -1074)	0x1p-1074	unchanged
scalbnf(1, 200), scalbnf(1, 2000)	inf	unchanged
scalbnf(1, -200), scalbnf(1, -2000)	0	unchanged
scalbln/scalblnf at the same and at ±(2^47 − 1)	as above	unchanged
ldexp(1, 2000), ldexp(1, -2000), ldexp(1, -1075), for contrast	inf, 0, 0	ERANGE

ldexp (double) keeps setting ERANGE as before, which matches the titles. scalbnl is unchanged: I haven't measured the titles' long double behaviour.

Tested

Windows 11, WinLibs GCC 15.2.0 UCRT r7, Release, main 3a2b919816.

• Native comparison: on 100,000 random (x, n) pairs, scalbn and scalbnf match the titles' libc.prx on every value and errno, on PPSA07923 and PPSA05916, both on main and with this PR.
• Regression: guest_math gains CheckScalbnErrno: 10 rows from the table above. Each must give the titles' bits and leave errno unchanged.
• Negative control: Windows can't provide one. UCRT's scalbn doesn't set errno, so the test passes on Windows with main's Math.cpp too. The failure it guards against is on glibc. rooklz's standalone run of the same pattern for ldexpf (fix(libc): leave errno unchanged from ldexpf, as titles' libc.prx does #12676) showed errno = 34 there, but I haven't run this test on Linux.
• Test set: guest_math, guest_math_exact, guest_dinkumware_internals and public_headers pass 3/3. Conventions pass. tools/pr_overlap.py: 52 open PRs share a file. Two conflict in GuestMath.cpp at the list of checks in main() (line 2434), both mine: fix(libc): compute erff and erfcf with fdlibm's algorithm, as titles' libc.prx does #10226 and fix(libc): compute atan and atan2 with FreeBSD msun's algorithm, as titles' libc.prx does #10229. Each adds its own check call nearby, so whichever lands second needs a trivial rebase. Six more already conflict with main and couldn't be compared (feat(libc): export _FDsign/_Dsign like MSVC CRT (PPSA34015) #6085, feat(libc): implement the nextafter and nexttoward families #6295, feat(libc): implement fesetenv #8937, fix(libc): compute asin and acos with FreeBSD msun's algorithm, as titles' libc.prx does #10253, fix(libc): set EDOM in fmod and fmodf where titles' libc.prx does #10345, fix(libc): set ERANGE for decimal subnormals in wcstod as strtod does #11892); none of their diffs touches scalbn or scalbnf.

Not tested: a Linux or macOS build or run.

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 (Claude Code)

🤖 Generated with Claude Code

[rooklz]

Approving. Your Not-tested section said Linux and macOS weren't run, so I ran both at a561f4e.

Setup. A standalone build of the head scalbn_nid_postfix/scalbnf_nid_postfix bodies and the verbatim CheckScalbnErrno (GuestMath.cpp L656–686). I also built main's one-liners as the control. Each arm also ran a 1,000,000-pair xorshift sweep (random double/float bits, exponents in ±2100 / ±400). The sweep compared values against bare std::scalbn and counted errno changes with errno preset to 7.

host	arm	CheckScalbnErrno failures	sweep errno changes	sweep value diffs
Ubuntu 24.04 x86_64, glibc 2.39, GCC 13	head	0	0	0
same	main	7 (errno=34)	1,124,062	0
macOS arm64, Apple clang	head	0	0	0
same	main	0	0	0

So you get the Linux negative control you couldn't get on Windows: the new test is RED on glibc with main's Math.cpp, and green with this PR.

Mutants (glibc):

• Drop the restore in scalbn only: 3 rows fail.
• Drop the restore in scalbnf only: 4 rows fail.

Both are killed, so each function is covered on its own.

Other checks:

• The PR is based on 3a2b919816. git merge-tree is clean against main and pairwise against #12676.
• check_conventions --base upstream/main passes.
• The save/restore matches the existing fmod and exp2 pattern in the same file.

Note for a later PR (not blocking): glibc's scalbnl sets ERANGE too. std::scalbn(1.0L, ±20000) gives errno=34, so scalbnl still differs on Linux. Leaving it out until the titles' long-double behaviour is measured is the right call, per the rules.

[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

[babyhuey]

Windows run of this head a561f4e59176, merged onto main 3a2b9198165 (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 2908 tests: 2162 pass, 4 fail, 742 skip. No failures from this PR. guest_freebsd_clocks and guest_raise_exception also fail on 3a2b9198165 itself on this machine (the first because it expects the host time zone to be UTC, fixed by fix(tests): drop the guest_freebsd_clocks check that the host zone is UTC #12660; the second libkernel (Windows): guest_raise_exception slowed about 3x and sometimes aborts at "address waking" since 540364a24 (#10666) #11813). select_tests_SelectArgumentTests (30 s) and trace_digest_TraceDigestTests_test_cli_rejects_invalid_event_fields_without_partial_output (20 s) timed out under --parallel 4; alone on the same merged tree, rebuilt, they passed 3 of 3 in 13-17 s and 6.9-7.7 s.

AI-assisted: yes (Claude Code)

[adriadam10]

Checked the errno save/restore around scalbn/scalbnf against the existing ErrnoGuard pattern, the 10 CheckScalbnErrno rows, and the green Conventions/Windows jobs, with rooklz's Linux/macOS run as the glibc evidence; merged.