Bug 384669 – app-text/ghostscript-gpl-9.04-r3 build failure - conflicting types for ‘recvmmsg’

• Product: Gentoo Linux
• Status: RESOLVED NEEDINFO

[Description Richard Freeman 2011-09-27 14:51:03 UTC]

When building ghostscript-gpl on stable amd64 I can't seem to figure out the cause of the attached build errors.  I get the same errors with just about any set of CFLAGS including -march=k8 -O2 -pipe.

As far as I can tell nothing in ghostscript-gpl actually defines recvmmsg, and grepping /usr/include I can't find anything else that defines it either.  

Note also that this error is generated during the install phase.  I'm not sure why it is even building anything in the install phase.

Reproducible: Always

[Comment 1 Richard Freeman 2011-09-27 14:52:58 UTC]

Created attachment 287937 [details]
build log

[Comment 2 Richard Freeman 2011-09-27 14:54:53 UTC]

Created attachment 287939 [details]
gzip build log

[Comment 3 Richard Freeman 2011-09-27 14:55:58 UTC]

Created attachment 287941 [details]
emerge --info

(Note that I can reproduce this with simplified CFLAGS.)

[Comment 4 Richard Freeman 2011-09-27 15:04:12 UTC]

Created attachment 287943 [details]
log with reduced CFLAGS

Error is same with reduced cflags.

[Comment 5 Richard Freeman 2011-10-15 19:46:40 UTC]

I suspect the issue lies with a package in @system.  During tests building from a stage3 I could reproduce this problem with my CFLAGS and USE flags.  If I drop the kerberos USE flag I can build with my CFLAGS, and if I reduce the CFLAGS to something more reasonable I can build with my full USE flags.  So, it is the combination of the kerberos USE flag and one of my CFLAGS that causes the problem when building ghostscript-gpl from a stage3.

The stage3 build fails on ghostscript-gpl but I'm pretty sure the problem originates from something in @system.  Simply building ghostscript-gpl without the kerberos USE flag doesn't fix the issue.  Neither does emerge -e ghostscript-gpl, which makes me suspect the problem isn't in a named dependency.

The other possibility is a non-system dependency that isn't named somewhere in the dependency graph.

In any case, this is obviously pretty obscure - up to you guys whether you want to track it down or close this...  In the end I just emerged it from a binary package from a stage3 chroot.

[Comment 6 Andreas K. Hüttel 2012-05-18 20:25:13 UTC]

Is this still a problem with 9.05?

[Comment 7 Richard Freeman 2012-05-18 23:41:00 UTC]

(In reply to comment #6)
> Is this still a problem with 9.05?

Yup.

Builds fine in a chroot.  The issue obviously creeps in from some dependency since nothing I do when building gimp seems to matter.

[Comment 8 Piotr Curious Slawinski 2012-10-27 03:41:20 UTC]

i also bumped against this bug. system is plain x86. 
it happens only on one of machines, but is very stubborn (reemerged world
several times since, and bug is still on).

[Comment 9 Piotr Curious Slawinski 2012-10-27 05:00:15 UTC]

tried to manually compile gstoraster.c and stripped various options. 
it seems it's gcc bug - removing -combine makes it compile.

it seems gcc is confused by two .c files given in single commandline , 
which are mutually interlinked (sharing various includes).

in simple words , just compile with CFLAGS="-Os" emerge ghostscript-gpl
until gcc will be fixed, and file gcc bug

[Comment 10 Andreas K. Hüttel 2013-03-24 20:18:15 UTC]

Maybe someone from toolchain could have a look.

[Comment 11 Jonas Stein 2016-10-30 00:25:18 UTC]

is the problem still present in 9.10 and 9.15, or can we close the ticked?

[Comment 12 SpanKY 2016-11-15 06:50:59 UTC]

if it's still present, can you get it to the point where you can go into the build dir and run the gcc command by hand and have it fail ?

[Comment 13 Andreas K. Hüttel 2017-02-18 12:54:56 UTC]

(In reply to Jonas Stein from comment #11)
> is the problem still present in 9.10 and 9.15, or can we close the ticked?