I'm suddenly getting segfaults when running tests on Windows. It works fine on Linux. I reduced it to a few lines (plus a dependency) with dustmite, but they don't really make sense[1]. Nevertheless they do trigger the segfault.

Can someone with Windows 10 and dmd 2.087.0 try the following and see if it crashes please?

> git clone https://github.com/zorael/tests -b wintestcrash && cd tests && dub test

If it ends with an assertion failure, okay, darn. If it "exited with code -1073741819" I'll have confirmed it happens for more than just me and AppVeyor, and thus can go ahead and file a bug report.

> Running custom 'unittest' configuration.
> Performing "unittest" build using C:\D\dmd2\windows\bin\dmd.exe for x86_64.
> cachetools 0.1.2: target for configuration "library" is up to date.
> requests 1.0.9: target for configuration "std" is up to date.
> test ~wintestcrash: building configuration "unittest"...
> Linking...
> To force a rebuild of up-to-date targets, run again with --force.
> Running .\test.exe
> Program exited with code -1073741819

[1]: https://github.com/zorael/tests/blob/wintestcrash/source/app.dx

On Friday, 12 July 2019 at 22:46:11 UTC, Anonymouse wrote:
> I'm suddenly getting segfaults when running tests on Windows. It works fine on Linux. I reduced it to a few lines (plus a dependency) with dustmite, but they don't really make sense[1]. Nevertheless they do trigger the segfault.
>
> Can someone with Windows 10 and dmd 2.087.0 try the following and see if it crashes please?
>
>> [...]
>
> If it ends with an assertion failure, okay, darn. If it "exited with code -1073741819" I'll have confirmed it happens for more than just me and AppVeyor, and thus can go ahead and file a bug report.
>
>> [...]
>
> [1]: https://github.com/zorael/tests/blob/wintestcrash/source/app.dx

maybe you haven't pushed?
I can't see it.

dmd v2.082 cannot reproduce

Needs to be tested on something newer perhaps?

On Friday, 12 July 2019 at 22:46:11 UTC, Anonymouse wrote:
> I'm suddenly getting segfaults when running tests on Windows. It works fine on Linux. I reduced it to a few lines (plus a dependency) with dustmite, but they don't really make sense[1]. Nevertheless they do trigger the segfault.
>
> Can someone with Windows 10 and dmd 2.087.0 try the following and see if it crashes please?
>
>> git clone https://github.com/zorael/tests -b wintestcrash && cd tests && dub test

I can reproduce it on Win10/x64 (qemu).
But it's really hard to debug on windows (at least with dmd and no V.Studio), I could only get a readable backtrace with "WinDbg Preview" debugger:

[0x0]   msvcr100!_output_l + 0x41e
[0x1]   msvcr100!printf + 0x7c
[0x2]   test!int core.runtime.runModuleUnitTests().__foreachbody1(object.ModuleInfo*) + 0xd1
[0x3]   test!int object.ModuleInfo.opApply(scope int delegate(object.ModuleInfo*)).__lambda2(immutable(object.ModuleInfo*)) + 0x27
[0x4]   test!int rt.minfo.moduleinfos_apply(scope int delegate(immutable(object.ModuleInfo*))).__foreachbody2(ref rt.sections_win64.SectionGroup) + 0x54
[0x5]   test!int rt.minfo.moduleinfos_apply(scope int delegate(immutable(object.ModuleInfo*))) + 0x1f
[0x6]   test!int object.ModuleInfo.opApply(scope int delegate(object.ModuleInfo*)) + 0x27
[0x7]   test!runModuleUnitTests + 0xfe
...
msvcr100!_output_l+0x41e:
00000000`62e0346e 443811   cmp  byte ptr [rcx],r10b ds:00000000`00000010=??
Registers are: rcx = 0x0000000000000010, r10 = 0x0000000000000000
it seems like ( str[i] == '\0')

So the program try to print from an invalid pointer and the StackTrace/Throwable object is somewhat related.

On Friday, 12 July 2019 at 23:05:41 UTC, Stefan Koch wrote:
>> [1]: https://github.com/zorael/tests/blob/wintestcrash/source/app.dx
>
> maybe you haven't pushed?
> I can't see it.

Sorry, typo in the link.

https://github.com/zorael/tests/blob/wintestcrash/source/app.d

On Saturday, 13 July 2019 at 06:36:06 UTC, Boris Carvajal wrote:
> On Friday, 12 July 2019 at 22:46:11 UTC, Anonymouse wrote:
>> I'm suddenly getting segfaults when running tests on Windows. It works fine on Linux. I reduced it to a few lines (plus a dependency) with dustmite, but they don't really make sense[1]. Nevertheless they do trigger the segfault.
>>
>> Can someone with Windows 10 and dmd 2.087.0 try the following and see if it crashes please?
>>
>>> git clone https://github.com/zorael/tests -b wintestcrash && cd tests && dub test
>
> I can reproduce it on Win10/x64 (qemu).
> But it's really hard to debug on windows (at least with dmd and no V.Studio), I could only get a readable backtrace with "WinDbg Preview" debugger:

Thank you!

Filed as https://issues.dlang.org/show_bug.cgi?id=20048.

On Saturday, 13 July 2019 at 16:39:51 UTC, Anonymouse wrote:

> Thank you!
>
> Filed as https://issues.dlang.org/show_bug.cgi?id=20048.

https://github.com/dlang/druntime/pull/2675

On Sunday, 14 July 2019 at 08:49:43 UTC, Boris Carvajal wrote:
> On Saturday, 13 July 2019 at 16:39:51 UTC, Anonymouse wrote:
>
>> Thank you!
>>
>> Filed as https://issues.dlang.org/show_bug.cgi?id=20048.
>
> https://github.com/dlang/druntime/pull/2675

Excellent, thanks!

I got OPTLINK errors while trying to confirm that it worked with digger, so I'll just trust it and wait for 2.088.