[Bug 57782] New: gdiplus:brush test crashes in _controlfp_s on intel mac

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=57782

Bug ID: 57782
           Summary: gdiplus:brush test crashes in _controlfp_s on intel
                    mac
           Product: Wine
           Version: 10.0
          Hardware: x86-64
                OS: MacOS
            Status: NEW
          Severity: normal
          Priority: P2
         Component: msvcrt
          Assignee: [e-mail]
          Reporter: [e-mail]

On my intel mac, the gdiplus:brush crashes with:
0024:err:seh:call_seh_handlers invalid frame 00000001000FF790
(0000000000022000-0000000000220000)
0024:err:seh:NtRaiseException Exception frame is not in stack limits => unable
to dispatch exception.

If I comment out the line calling _controlfp_s in brush.c, the test succeeds.

[Bug 57782] gdiplus:brush test crashes in _controlfp_s on intel mac

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=57782

Esme Povirk  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |[e-mail]
           Keywords|                            |download, source, testcase

[Bug 57782] gdiplus:brush test crashes in _controlfp_s on intel mac

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=57782

Piotr Caban  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
          Component|msvcrt                      |-unknown
                 CC|                            |[e-mail]

--- Comment #1 from Piotr Caban  ---
It's not an msvcrt bug - setting component to unknown.

The same exception can be seen on Linux when Wine is compiled with Clang. We're
getting _EM_INVALID exception in monitor_get_dpi function when computing
scale_x and scale_y. I didn't look what exactly is happening there. I'm not
sure what's the correct fix - maybe we should mask some floating point
exceptions in syscalls.

There are more exceptions if monitor_get_dpi is hacked to workaround this
issue. I didn't look on the next one.

[Bug 57782] gdiplus:brush test crashes in _controlfp_s on intel mac

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=57782

--- Comment #2 from Piotr Caban  ---
In case of monitor_get_dpi it should be possible to avoid the problem with
MulDiv function.

Next problem is in GdipInvertMatrix. In this case code generated by Clang also
throws EM_INVALID exception. I have worked around it with a hack and all the
remaining tests passed. So these seems to be the only 2 functions that are
affected.

Probably the next step is to determine if compiler is correct creating code
that can emit such exception.