utf_encodings.phpt fails on Windows 32-bit

[iluuu1994]

Description

https://github.com/php/php-src/actions/runs/5885623549/job/15962365817

========DIFF========
     == UTF-8 ==
002+ mb_check_encoding passed on bad UTF-8 string: (f4908080f090bc87)
003+ mb_check_encoding passed on bad UTF-8 string: (f7808080f0919a88)
004+ mb_check_encoding passed on bad UTF-8 string: (f7bfbfbfe0a196)
005+ mb_check_encoding passed on bad UTF-8 string: (eda080f3a08587)
     == UTF-16 ==
     == UTF-32 ==
     == UTF-7 ==
--
========DONE========
FAIL Torture test for UTF-{7,8,16,32} [D:\a\php-src\php-src\ext\mbstring\tests\utf_encodings.phpt] 

@alexdowad Do you know what's going on here?

PHP Version

master

Operating System

No response

[alexdowad]

Hi Iljia, the first thing to check is whether this is a SIMD-accelerated (SSE or AVX2) build of PHP or not. There may be some other issue with the SIMD code which wasn't discovered before. I will be more available to look at it next week.

…

On Thu, Aug 17, 2023 at 7:14 PM Ilija Tovilo ***@***.***> wrote: Description https://github.com/php/php-src/actions/runs/5885623549/job/15962365817 ========DIFF======== == UTF-8 == 002+ mb_check_encoding passed on bad UTF-8 string: (f4908080f090bc87) 003+ mb_check_encoding passed on bad UTF-8 string: (f7808080f0919a88) 004+ mb_check_encoding passed on bad UTF-8 string: (f7bfbfbfe0a196) 005+ mb_check_encoding passed on bad UTF-8 string: (eda080f3a08587) == UTF-16 == == UTF-32 == == UTF-7 == -- ========DONE======== FAIL Torture test for UTF-{7,8,16,32} [D:\a\php-src\php-src\ext\mbstring\tests\utf_encodings.phpt] @alexdowad <https://github.com/alexdowad> Do you know what's going on here? PHP Version master Operating System *No response* — Reply to this email directly, view it on GitHub <#11992>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAIESX5CL2QEX7UROUATIO3XVZGPRANCNFSM6AAAAAA3UMHKJI> . You are receiving this because you were mentioned.Message ID: ***@***.***>

[iluuu1994]

Hi @alexdowad! Thanks for your response. Since this only occurs in master there's no rush, so I'll let you handle it since I'm mostly unfamiliar with mbstring, and don't have a working Windows machine.

[alexdowad]

Looks like the build is failing on 32-bit Linux machines, not just 32-bit Windows.

[iluuu1994]

@alexdowad It looks fine on nightly (e.g. https://github.com/php/php-src/actions/runs/5933282918/job/16088514605), maybe there are multiple variables involved?

[alexdowad]

Hmm. We don't have a lot of clues here.

A really good first clue would be to know the values of the ZEND_INTRIN_AVX2_{NATIVE,RESOLVER} macros in this build.

I'm just trying to figure out if there is a good way to display the values of all the ZEND_INTRIN macros in the CI log... this would be good not just for debugging this problem, but even for debugging other similar problems in the future.

[iluuu1994]

@alexdowad Maybe lscpu and cat main/php_config.h could help. You can try to put them into a run step after configure in .github/workflows/push.yml. Edit: Oh wait, Windows of course ^^ I don't know what it's equivalent of lscpu is.

[alexdowad]

@alexdowad Maybe lscpu and cat main/php_config.h could help. You can try to put them into a run step after configure in .github/workflows/push.yml. Edit: Oh wait, Windows of course ^^ I don't know what it's equivalent of lscpu is.

It would be great to print various useful debugging info on Windows as well, but even printing on Linux only would be an improvement from the current state of things.

This mbstring issue seems to be occurring on both 32-bit Windows and 32-bit Linux hosts anyways.

[iluuu1994]

@alexdowad I understand you said this fails on Linux 32-bit as well. However, that's not the case on GitHub actions. So I've marked this test as xfail only for Windows + 32-bit for now so that nightly is green again. 6b74f1f

[nielsdos]

Looking at the last nightly this still seems to fail: https://github.com/php/php-src/actions/runs/6910901883/job/18804743373 The release of 8.3 is close now.
I suppose no one had a look yet at the cause?
I recall we had a similar strange issue before #10776

[nielsdos]

FWIW I'm investigating now.