`ext\spl\tests\bug77263.phpt` test is causing sigfault on Windows 32-bit OS

[mvorisek]

Description

code https://github.com/php/php-src/blob/9b8f76bfa29ff234ff45a27b8171e13c9b8d2928/ext/spl/tests/bug77263.phpt
CI: https://github.com/mvorisek/php-src/runs/6201706203?check_suite_focus=true#step:8:3851

========DIFF========
001+ Can't initialize heap: [0x000001e7] Attempt to access invalid address

001- OK

002+ 

003+ Termsig=-1073741819
========DONE========
FAIL Bug #77263 (Segfault when using 2 RecursiveFilterIterator) [D:\a\php-src\php-src\ext\spl\tests\bug77263.phpt] 

for some reasons, this issue is not always present

PHP Version

8.x, probably master too

Operating System

Windows 32-bit

[cmb69]

I haven't been able to reproduce this locally, and I have no idea what exactly is going there.

[mvorisek]

it is failing randomly

sapi/phpdbg/tests/watch_001.phpt - sapi/phpdbg/tests/watch_004.phpt tests have never failed for me
but sapi/phpdbg/tests/watch_005.phpt - sapi/phpdbg/tests/watch_006.phpt is failing consistently with the same Termsig=-1073741819 error code, not sure if this is related, but can you reproduce the segfault for these tests on Windows 32-bit?

[cmb69]

it is failing randomly

Yeah, I know, but I ran several times (x86 ZTS, with and without JIT), and it never failed. :(

Neither sapi/phpdbg/tests/watch_005.phpt nor sapi/phpdbg/tests/watch_006.phpt is failing for me. However, -1073741819 is basically 0xc0000005 which is the code for an access violation (aka. segfault), and these tests mention "intentionally causes segfaults"; these should be caught, though, but for some reason they're not (or an uncaught segfault happens). Anyway, hard to debug if not reproducible.

[mvorisek]

I do not know how to debug this, but here is repro code: https://github.com/mvorisek/php-src/tree/repro_win32bit_segfault

the sapi/phpdbg/tests/watch_005.phpt and sapi/phpdbg/tests/watch_006.phpt tests always fail [1], the env is Microsoft Windows Server 2019 Datacenter 64 bit OS & php-src compiled for x86 platform

is it possible to update the repro code to dump the C stack trace to understand the issue just from the CI output?

[1] https://github.com/mvorisek/php-src/runs/6213041816?check_suite_focus=true#step:5:84

[cmb69]

is it possible to update the repro code to dump the C stack trace to understand the issue just from the CI output?

Hmm, that would require to install and register a post mortem debugger, which than would create a (huge) dump file. That dump file would need to be uploaded as artifact, and could then be downloaded for inspection. Not sure if that's worth the trouble for phpdbg issues.

[mvorisek]

duplicate of #9650