[Bug 56956] New: MSVC cl.exe 19.* fails to flush intermediate file

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=56956

Bug ID: 56956
           Summary: MSVC cl.exe 19.* fails to flush intermediate file
           Product: Wine
           Version: 9.11
          Hardware: x86-64
                OS: Linux
            Status: UNCONFIRMED
          Severity: normal
          Priority: P2
         Component: msvcrt
          Assignee: [e-mail]
          Reporter: [e-mail]
      Distribution: ---

Created attachment 76772
  --> https://bugs.winehq.org/attachment.cgi?id=76772
a trivial C program to reproduce the issue

When using any MSVC 19.* version (I've tested several) compilation of an object
now fails with the following message:

c1: fatal error C1088: Cannot flush compiler intermediate file:
'C:\users\lmiddlebrook\Temp_CL_c8f4abecex': Bad file descriptor

I used git-bisect and found that the regression commit is:

commit 1163e399376ab74eb38a6068daba69497b6b14be (HEAD)
Author: Shengdun Wang 
Date:   Wed Jun 12 13:01:31 2024 +0200

ucrtbase: Fix FILE _flag values.

dlls/msvcrt/msvcrt.h       | 8 ++++++++
 dlls/ucrtbase/tests/file.c | 2 +-
 include/msvcrt/stdio.h     | 2 ++
 3 files changed, 11 insertions(+), 1 deletion(-)

I've attached my example C program (a basic hello world). I use the following
command-line to repro:

cl.exe -c z:/tmp/test.c

[Bug 56956] MSVC cl.exe 19.* fails to flush intermediate file

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=56956

Liam Middlebrook  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
    Regression SHA1|                            |1163e399376ab74eb38a6068dab
                   |                            |a69497b6b14be

[Bug 56956] MSVC cl.exe 19.* fails to flush intermediate file

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=56956

Fabian Maurer  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |[e-mail]
           Keywords|                            |regression

--- Comment #1 from Fabian Maurer  ---
How do you get the cl.exe for testing?

[Bug 56956] MSVC cl.exe 19.* fails to flush intermediate file

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=56956

Fabian Maurer  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |[e-mail]

--- Comment #2 from Fabian Maurer  ---
CCing author, can you please take a look?

[Bug 56956] MSVC cl.exe 19.* fails to flush intermediate file

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=56956

--- Comment #3 from cqwrteur  ---
Hi. I haven't checked what's causing regression for msvc.

Can you first use this with clang for cross compilation? My
windows-msvc-sysroot project provides msvc libraries for clang to compile your
code instead of msvc on linux.

https://github.com/trcrsired/windows-msvc-sysroot/tree/main

Also, please support me in this LLVM PR.
https://github.com/llvm/llvm-project/pull/96417

[Bug 56956] MSVC cl.exe 19.* fails to flush intermediate file

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=56956

--- Comment #4 from Liam Middlebrook  ---
(In reply to Fabian Maurer from comment #1)
> How do you get the cl.exe for testing?

It comes with Microsoft Visual Studio. I'm not sure if it's available as a
separate command-line distribution. It looks like the following link may be for
that but I'm not familiar with that, sorry:
https://visualstudio.microsoft.com/downloads/#build-tools-for-visual-studio-...

(In reply to cqwrteur from comment #3)
> Hi. I haven't checked what's causing regression for msvc.
>
> Can you first use this with clang for cross compilation? My
> windows-msvc-sysroot project provides msvc libraries for clang to compile
> your code instead of msvc on linux.

No. I am not going to change my build toolchain due to a regression in Wine.
I've locally reverted the regressing commit in my build so that my work is not
blocked.

[Bug 56956] MSVC cl.exe 19.* fails to flush intermediate file

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=56956

--- Comment #5 from cqwrteur  ---
(In reply to Fabian Maurer from comment #2)
> CCing author, can you please take a look?

I wasn't the actual author. I wrote my patch, and then the wine developers
rewrote the code. I have trouble tracing the changes

[Bug 56956] MSVC cl.exe 19.* fails to flush intermediate file

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=56956

--- Comment #6 from cqwrteur  ---
(In reply to Liam Middlebrook from comment #4)
> (In reply to Fabian Maurer from comment #1)
> > How do you get the cl.exe for testing?
>
> It comes with Microsoft Visual Studio. I'm not sure if it's available as a
> separate command-line distribution. It looks like the following link may be
> for that but I'm not familiar with that, sorry:
> https://visualstudio.microsoft.com/downloads/#build-tools-for-visual-studio-
> 2022
>
> (In reply to cqwrteur from comment #3)
> > Hi. I haven't checked what's causing regression for msvc.
> >
> > Can you first use this with clang for cross compilation? My
> > windows-msvc-sysroot project provides msvc libraries for clang to compile
> > your code instead of msvc on linux.
>
> No. I am not going to change my build toolchain due to a regression in Wine.
> I've locally reverted the regressing commit in my build so that my work is
> not blocked.

The abi before the commit was just broken. UCRT abi in the microsoft's
toolchain and windows is different from the ABI in UCRT. If you pass a FILE*
from the static linked program into DLL, it would just crash.

[Bug 56956] MSVC cl.exe 19.* fails to flush intermediate file

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=56956

--- Comment #7 from cqwrteur  ---
(In reply to Liam Middlebrook from comment #4)
> (In reply to Fabian Maurer from comment #1)
> > How do you get the cl.exe for testing?
>
> It comes with Microsoft Visual Studio. I'm not sure if it's available as a
> separate command-line distribution. It looks like the following link may be
> for that but I'm not familiar with that, sorry:
> https://visualstudio.microsoft.com/downloads/#build-tools-for-visual-studio-
> 2022
>
> (In reply to cqwrteur from comment #3)
> > Hi. I haven't checked what's causing regression for msvc.
> >
> > Can you first use this with clang for cross compilation? My
> > windows-msvc-sysroot project provides msvc libraries for clang to compile
> > your code instead of msvc on linux.
>
> No. I am not going to change my build toolchain due to a regression in Wine.
> I've locally reverted the regressing commit in my build so that my work is
> not blocked.

I am trying to trace what breaks. However, I suggest you not to use msvc any
more. In fact, I would argue Microsoft should just give Visual C++ like they
gave up EdgeHtml.
https://github.com/cppfastio/fast_io/issues/770

clang is perfectly fine and better than msvc in every possible way and it works
for cross compiling.

[Bug 56956] MSVC cl.exe 19.* fails to flush intermediate file

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=56956

Fabian Maurer  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |[e-mail]

--- Comment #8 from Fabian Maurer  ---
> The abi before the commit was just broken. UCRT abi in the microsoft's toolchain and windows is different from the ABI in UCRT. If you pass a FILE* from the static linked program into DLL, it would just crash.

Well now it seems broken in a different way.
> I suggest you not to use msvc any more.

That is not an option, msvc has keep working.
> I wasn't the actual author. I wrote my patch, and then the wine developers rewrote the code. I have trouble tracing the changes

CCing reviewer then. Piotr, mind taking a look here?

[Bug 56956] MSVC cl.exe 19.* fails to flush intermediate file

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=56956

Piotr Caban  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |[e-mail]

--- Comment #9 from Piotr Caban  ---
(In reply to Fabian Maurer from comment #8)
> CCing reviewer then. Piotr, mind taking a look here?

Sure, I will try to reproduce the bug on Monday.

[Bug 56956] MSVC cl.exe 19.* fails to flush intermediate file

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=56956

--- Comment #10 from Piotr Caban  ---
I've created MR that addresses the bug:
https://gitlab.winehq.org/wine/wine/-/merge_requests/6077

[Bug 56956] MSVC cl.exe 19.* fails to flush intermediate file

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=56956

Liam Middlebrook  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |RESOLVED
      Fixed by SHA1|                            |40a4809c1c6bda1c6dd5b64a4dd
                   |                            |52393d5ac965f
         Resolution|---                         |FIXED

--- Comment #11 from Liam Middlebrook  ---
Confirming that MSVC is working again as of
40a4809c1c6bda1c6dd5b64a4dd52393d5ac965f (Piotr's commit merged into master).

Tested on Release 9.13 as well as the fix commit (and verified broken on its
parent).

Thanks for the quick investigation on this!

[Bug 56956] MSVC cl.exe 19.* fails to flush intermediate file

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=56956

--- Comment #12 from cqwrteur  ---
(In reply to Liam Middlebrook from comment #11)
> Confirming that MSVC is working again as of
> 40a4809c1c6bda1c6dd5b64a4dd52393d5ac965f (Piotr's commit merged into
> master).
>
> Tested on Release 9.13 as well as the fix commit (and verified broken on its
> parent).
>
> Thanks for the quick investigation on this!

Thanks everyone for helping fix it.

BTW pitor can you check any other flags may be broken? UCRT changes all the
flags value and i changed it but you did not want to break headers. I guess
that is the reason for this bug? Can you check whether any other flags are
broken? Thanks

[Bug 56956] MSVC cl.exe 19.* fails to flush intermediate file

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=56956

--- Comment #13 from cqwrteur  ---
(In reply to Piotr Caban from comment #9)
> (In reply to Fabian Maurer from comment #8)
> > CCing reviewer then. Piotr, mind taking a look here?
>
> Sure, I will try to reproduce the bug on Monday.

_IOMYBUF is the IO_USERBUF? If I remember correctly?

[Bug 56956] MSVC cl.exe 19.* fails to flush intermediate file

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=56956

--- Comment #14 from cqwrteur  ---
(In reply to cqwrteur from comment #13)
> (In reply to Piotr Caban from comment #9)
> > (In reply to Fabian Maurer from comment #8)
> > > CCing reviewer then. Piotr, mind taking a look here?
> >
> > Sure, I will try to reproduce the bug on Monday.
>
> _IOMYBUF is the IO_USERBUF? If I remember correctly?

ok it is CRTBUF

https://bugs.winehq.org/attachment.cgi?id=73552

Here was the old patch i put for the meaning of these flags value

[Bug 56956] MSVC cl.exe 19.* fails to flush intermediate file

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=56956

--- Comment #15 from Piotr Caban  ---
(In reply to cqwrteur from comment #12)
> BTW pitor can you check any other flags may be broken? UCRT changes all the
> flags value and i changed it but you did not want to break headers. I guess
> that is the reason for this bug? Can you check whether any other flags are
> broken?

I have changed all the flags that are used in wine, adding more internal
constants want change anything unless you also fix the code to use them.

Note that this regression was also present in your version of the patch.

[Bug 56956] MSVC cl.exe 19.* fails to flush intermediate file

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=56956

--- Comment #16 from cqwrteur  ---
(In reply to Piotr Caban from comment #15)
> (In reply to cqwrteur from comment #12)
> > BTW pitor can you check any other flags may be broken? UCRT changes all the
> > flags value and i changed it but you did not want to break headers. I guess
> > that is the reason for this bug? Can you check whether any other flags are
> > broken?
>
> I have changed all the flags that are used in wine, adding more internal
> constants want change anything unless you also fix the code to use them.
>
> Note that this regression was also present in your version of the patch.

yeah i think so since NOBUF wasn't part of code.

Thank you.

[Bug 56956] MSVC cl.exe 19.* fails to flush intermediate file

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=56956

Alexandre Julliard  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|RESOLVED                    |CLOSED

--- Comment #17 from Alexandre Julliard  ---
Closing bugs fixed in 9.14.