2:21 p.m.
Hello,
My attempt at building a wine update today failed for only the ARMv7 (32-bit) builds
across F26, F27, and Rawhide. I'm seeing the build log[1] show the following message:
error: variadic functions must use the base AAPCS variant
I'm unfamiliar with this type of failure. The code that is failing has not been
changed in years. The last build that succeeded was 2 weeks ago on the same gcc version.
What has changed in the last 2 weeks?
[1] https://koji.fedoraproject.org/koji/taskinfo?taskID=23367607
10:16 p.m.
My attempt at building a wine update today failed for only the ARMv7
(32-bit) builds across F26, F27, and Rawhide. I'm seeing the build log[1] show the
following message:
error: variadic functions must use the base AAPCS variant
Are those functions really variadic (take a different number of
actual arguments in different calls)? Does the declaration
specify variadic by using three periods "..." in the right-most
argument position?
A change in header files could account for differing behavior.
" gcc -E $< | grep '^# 1 ' " gives the list of files that were
processed.
What was the actual literal gcc version then versus now?
--
9:07 a.m.
On 11/25/2017 10:16 PM, John Reiser wrote:
Are those functions really variadic (take a different number of
actual arguments in different calls)? Does the declaration
specify variadic by using three periods "..." in the right-most
argument position?
Yes and yes.
A change in header files could account for differing behavior.
" gcc -E $< | grep '^# 1 ' " gives the list of files that were
processed.
What was the actual literal gcc version then versus now?
I don't have a 32-bit ARM box to run commands on.
7.2.1-1.fc28 versus 7.2.1-1.fc28
2:06 a.m.
On 11/25/2017 09:21 PM, Michael Cronenworth wrote:
Hello,
My attempt at building a wine update today failed for only the ARMv7
(32-bit) builds across F26, F27, and Rawhide. I'm seeing the build
log[1] show the following message:
error: variadic functions must use the base AAPCS variant
I'm unfamiliar with this type of failure. The code that is failing has
not been changed in years. The last build that succeeded was 2 weeks ago
on the same gcc version.
What has changed in the last 2 weeks?
This changed:
https://source.winehq.org/git/wine.git/commitdiff/644f497e87c51f1a1c62b26...
aapcs-vfp is incompatible with variadic functions on mainline GCC.
Maybe Wine upstream assumes that you use a patched GCC to build it?
You will have to ask them what the intent is behind this change. I
doubt that Fedora's Wine intends to run any Windows ARM binaries. So
you should be able to just drop the upstream patch.
Thanks,
Florian
8:52 a.m.
On 11/28/2017 02:06 AM, Florian Weimer wrote:
This changed:
https://source.winehq.org/git/wine.git/commitdiff/644f497e87c51f1a1c62b26...
aapcs-vfp is incompatible with variadic functions on mainline GCC. Maybe Wine
upstream assumes that you use a patched GCC to build it?
You will have to ask them what the intent is behind this change. I doubt that
Fedora's Wine intends to run any Windows ARM binaries. So you should be able to
just drop the upstream patch.
Thanks for the tip, Florian.
The native Wine source doesn't use variadic functions, but the Wine-staging patches
introduce them.
@Sebastian, have you received any reports for this issue?
Thanks,
Michael
2343
days inactive
2348
days old
4 comments
3 participants
participants (3)
-
Florian Weimer -
John Reiser -
Michael Cronenworth