[verusferro]

What

strftime's %G, %g and %V threw for a negative tm_yday, a tm_yday past the year and a tm_wday below -6, because no rule fitted the console's results. They now follow the rule of the SDK libc.prx that PPSA06416 ships in sce_module/ (title SDK 7.00), read from its disassembly (REVERSE_ENGINEERING, source 2). strftime (Av3zjWi64Kw, 0x6d720) passes the C-locale time table to _Strftime (0x3ce10), whose conversion switch is at 0x3d2e0: %G 0x3d41d, %V 0x3d6bb, %g 0x3d9a6, %U 0x3d697, %W 0x3d7fa.

The rule, all in 32-bit arithmetic with C remainders:

• 1 January falls on c = (tm_wday - tm_yday + 371) % 7 and the next one on n = (c + 365 + leap) % 7.

• When n is 2, 3 or 4 and tm_yday - leap >= 366 - n, the date is week 1 of the next year.

• Otherwise week = (tm_yday - (tm_wday + 6) % 7 + 7) / 7, plus 1 when c is 2, 3 or 4. A negative week is week 1 of the next year. Week 0 belongs to the previous year: %G prints that year, and %V prints its last week computed from its length, or 00 when its last days are week 1.

• The leap test (0x3e830) is not Gregorian. A year divisible by 4 and not by 100 is leap, and of the years divisible by 100 only those with tm_year / 400 equal to 100 or -300:

  3e857: lea  -0x9c40(%rdi),%ecx    tm_year - 40000
  3e85d: cmp  $0x190,%ecx; jb       below 400 unsigned: leap
  3e86a: test %edi,%edi; jns        otherwise non-negative: not leap
  3e86e: add  $0x1d64f,%edi         tm_year + 120399
  3e876: cmp  $0x190,%edi; setb %al
  

For every real date from year 1 to 9999 this still gives ISO 8601, so valid dates print as before. Fields that are each in range but disagree with each other can change through the leap test: tm_year 100, tm_wday 1, tm_yday 362 prints 2001|01|01 as the module does, where the FreeBSD loop gave 2000|00|53.

tm_year + 1 and the other sums wrap at 32 bits as in the module; the code reproduces that through uint32_t without signed overflow. %U and %W keep their formula, (tm_yday + 7 - (tm_wday + shift) % 7) / 7, but now also wrap at 32 bits as the module does, so a field near INT_MAX matches it. The debt entry now records where the rule comes from and what is not measured on a console.

Tested

Ubuntu 24.04 (WSL2) x86-64, GCC 13.3, Release, full build with BUILD_TESTING=ON:

• The title's libc.prx, converted by the relinker and loaded in a host process with ProcessParametersFixture, returns all 82 console results in guest_strftime (the zone conversions left out) and the console values the old debt entry listed: %V 01, 00, 00 and 52 for tm_yday -25, -13, -12 and -1; week 01 of 2027; 59; 53; 04.
• This library's strftime and the module's print the same %G|%g|%V|%U|%W|%Y|%y for all 675,076 inputs of three random runs: fields up to INT_MIN and INT_MAX, every tm_wday from -14 to 14 around year ends, and the years the leap test treats differently.
• guest_strftime gains the console's out-of-range cases, rows for each branch, the wrap at INT_MAX and INT_MIN, and two %U/%W rows. The old code throws on the first ISO row and prints 78|78 for tm_yday INT_MAX. A Gregorian leap test fails the tm_year 100 row.
• ctest -R '^(guest_strftime|guest_time_conversion|guest_time_extras|guest_wide_formatting|patched_libraries|conventions_unknown_nids)$': 6/6 pass.
• tools/progress.py: libc 1344 → 1345 of 1355 (libc_strftime implemented).
• python3 tools/check_conventions.py --base origin/main: passed. python3 tools/pr_overlap.py: no conflicting lines or same exports. Of the pull requests it could not compare, feat(libc): add the std::__sce_v2 thread start, timed wait, clock and processor count calls #9447 changes _Xtime_get_ticks in Time.cpp and fix(libc): write a 36-byte struct tm as the console does #10409 the GuestTm layout; neither touches the week code. The rest share only docs/dev/debt/libc.md, in other entries.

Windows: not run by me. This head passed in runs by @ycbuzi and @wachi35676; the second also compared every date from year 1 to 9999 against ISO 8601 with no mismatch.

Not tested: macOS, a console run of the cases the old debt entry did not list.

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: fix(libc): preserve partial wcsftime output on overflow #10895
• Fixes: none
• Unblocks: none
• GPUs tested: n/a
• AI-assisted: yes

[github-actions]

Progress report (e05ba6f → c9df854)

📈 System libraries: 49.64% (+0.01%, +1 implemented functions)

✅ 1 implemented

Library	Name
libc	libc_strftime_nid_postfix

Changed libraries (implemented exports)

Library	Before	After	Change
libc	1344/1355 (99.19%)	1345/1355 (99.26%)	+0.07 pp

[MarcZX10]

Reviewed the %G/%g/%V implementation and the cited console/SDK evidence. The new tests cover the reported negative and oversized fields, inconsistent year fields, and 32-bit boundary cases; the PR reports matching behavior against the SDK module over 675,076 inputs and six focused host tests. I independently verified the diff, conventions, and a local C++20 syntax check of Time.cpp. Hosted eligibility/conventions/report and non-product jobs pass; the PR does not claim a hosted full product build. Windows/macOS product tests and console measurements for untested cases remain outstanding as documented.

[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 5021	core/libs/prx/libc/src/Time.cpp docs/dev/debt/libc.md
PR 5303	docs/dev/debt/libc.md
PR 6713	docs/dev/debt/libc.md
PR 7081	docs/dev/debt/libc.md
PR 7201	core/libs/prx/libc/src/Time.cpp
PR 7386	docs/dev/debt/libc.md
PR 7447	docs/dev/debt/libc.md
PR 7544	docs/dev/debt/libc.md
PR 7648	docs/dev/debt/libc.md
PR 7715	docs/dev/debt/libc.md
PR 8150	docs/dev/debt/libc.md
PR 9371	docs/dev/debt/libc.md
PR 9447	core/libs/prx/libc/src/Time.cpp docs/dev/debt/libc.md
PR 9526	docs/dev/debt/libc.md
PR 9772	docs/dev/debt/libc.md
PR 9863	docs/dev/debt/libc.md
PR 9929	docs/dev/debt/libc.md
PR 9953	docs/dev/debt/libc.md
PR 9964	docs/dev/debt/libc.md
PR 10002	docs/dev/debt/libc.md
PR 10007	docs/dev/debt/libc.md
PR 10158	docs/dev/debt/libc.md
PR 10200	docs/dev/debt/libc.md
PR 10234	docs/dev/debt/libc.md
PR 10315	docs/dev/debt/libc.md
PR 10317	docs/dev/debt/libc.md
PR 10321	docs/dev/debt/libc.md
PR 10325	docs/dev/debt/libc.md
PR 10328	docs/dev/debt/libc.md
PR 10345	docs/dev/debt/libc.md
PR 10349	docs/dev/debt/libc.md
PR 10405	docs/dev/debt/libc.md
PR 10409	core/libs/prx/libc/src/Time.cpp core/libs/tests/GuestStrftime.cpp
PR 10467	docs/dev/debt/libc.md
PR 10506	docs/dev/debt/libc.md
PR 10510	docs/dev/debt/libc.md
PR 10512	docs/dev/debt/libc.md
PR 10529	docs/dev/debt/libc.md
PR 10565	docs/dev/debt/libc.md
PR 10749	docs/dev/debt/libc.md
PR 11332	docs/dev/debt/libc.md
PR 11341	docs/dev/debt/libc.md

[ycbuzi]

A Windows run of this head: Windows 11 Home 10.0.26200, WinLibs GCC 15.2.0 r7 (MinGW-w64 UCRT, posix-seh), CMake 4.3.1, Ninja 1.13.2, Release. c9df854687 merged onto main d2b3ee3bcc with no conflicts.

• guest_strftime_tests builds with no warnings or errors.
• guest_strftime passes both runs of ctest --repeat until-fail:2.
• Negative control: libc/src/Time.cpp taken from main, with this test. It fails with strftime: ISO week of a negative tm_yday not implemented.

Not run: Linux, macOS, the comparison against the title's libc.prx.

AI-assisted: yes (Claude Code)

[wachi35676]

First Windows run of head c9df85468795 merged onto main 90f918fa7. Windows 11 10.0.26300, WinLibs GCC 15.2.0, Release.

• guest_strftime passes 3/3. With Time.cpp from main, it aborts with strftime: ISO week of a negative tm_yday not implemented.
• I checked the "every real date still gives ISO 8601" claim independently. For every day of years 1 to 9999 (3,652,059 dates, proleptic Gregorian, 1 Jan 0001 a Monday), I compared strftime("%G|%g|%V|%U|%W") against an ISO 8601 / C99 week calculation of my own. There are 0 mismatches once %G is printed with four digits, as the module does (0001). So the non-Gregorian leap test only affects inconsistent field combinations, never real dates.

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

[StoneDC]

@verusferro Thanks for the careful work and reference-module evidence in this PR! I maintain #10895, which fixes partial output from wcsftime on overflow. Your week-number fix addresses a separate issue, but I compared our current heads (920903db9 and c9df854687) and confirmed merge conflicts in Time.cpp and docs/dev/debt/libc.md.

Would you be happy with #10895 landing first, followed by rebasing #11004 onto the updated main and recording Depends on: #10895 in your checklist? #10895 already has approvals, and we have agreed on that same first step with the author of #10987.

When rebasing, we would need to preserve the templated FormatBuffer declaration from #10895 alongside your week-number changes, retain both regression test sets, and keep both debt-entry updates. Rerunning guest_strftime and guest_wide_formatting after integration would help check the combined behavior. I have checked the textual conflicts, but have not run the combined code.

Thanks for coordinating! Please let me know if you would prefer a different order; the final merge decision remains with the maintainers.

[verusferro]

@StoneDC Agreed: #10895 first. I've set Depends on: #10895, and I'll rebase #11004 once #10895 is on main.

I tried that rebase locally, replaying c9df854687 onto your head 920903db93. Both conflicts are adjacent edits:

• Time.cpp: this PR deletes isLeapYearSum (nothing else calls it after the week rewrite), and fix(libc): preserve partial wcsftime output on overflow #10895 adds template<typename TChar> on the next line. The resolution drops the function and keeps the templated FormatBuffer.
• libc.md: this PR changes only the strftime entry and fix(libc): preserve partial wcsftime output on overflow #10895 only the wcsftime entry below it, so the resolution keeps both.

Both test sets merge without conflicts. On that combined tree, guest_strftime, guest_wide_formatting, guest_time_conversion, guest_time_extras, patched_libraries and conventions_unknown_nids pass (6/6, Linux), and so does check_conventions.py.