Bug 389931 – dev-lang/vala fails tests on portage-multilib, ignoring CFLAGS when calling VALAC

• Status: RESOLVED FIXED
• Product: Gentoo Linux

[Description Nathan Phillip Brink (binki) (RETIRED) 2011-11-09 05:39:30 UTC]

Created attachment 291981 [details, diff]
vala-0.12.1-multilib-tests.patch

The tests always fail for ABI=x86 when running under portage-multilib on an amd64 machine. The reason is that -m32 is not being passed to the computer by the VALAC.

The particular build error in the build.log is caused by a mismatch between glib headers and gcc's compilation mode:

ohnobinki@ohnopublishing ~ $ ABI=amd64 pkg-config --cflags glib-2.0
-I/usr/include/glib-2.0 -I/usr/lib64/glib-2.0/include  
ohnobinki@ohnopublishing ~ $ ABI=x86 pkg-config --cflags glib-2.0
-I/usr/include/glib-2.0 -I/usr/lib32/glib-2.0/include  

ohnobinki@ohnopublishing ~ $ grep -R -e typedef.\*gsize /usr/lib*/glib-2.0/include
/usr/lib32/glib-2.0/include/glibconfig.h:typedef unsigned int gsize;
/usr/lib64/glib-2.0/include/glibconfig.h:typedef unsigned long gsize;

The negative array reference is when sizeof(gsize) != sizeof(gpointer) in the definition of g_atomic_pointer_get() by the glib/gatomic.h header. It is apparently a sanity check of some sort. Of course, when -m32 is used sizeof(int) == sizeof(void*). But when -m32 is forgotten, the error condition of sizeof(int) != sizeof(void*) is encountered.

The attached patch incorporates CFLAGS from the environment and fixes this problem.

[Comment 1 Nathan Phillip Brink (binki) (RETIRED) 2011-11-09 05:40:27 UTC]

Created attachment 291983 [details]
vala-0.12.1-build.log

[Comment 2 Nathan Phillip Brink (binki) (RETIRED) 2011-11-09 05:40:47 UTC]

Created attachment 291985 [details]
emerge--info.txt

[Comment 3 Pacho Ramos 2011-11-09 10:32:31 UTC]

I would send this to upstream also: bugzilla.gnome.org

[Comment 4 Pacho Ramos 2011-11-22 01:34:50 UTC]

(In reply to comment #3)
> I would send this to upstream also: bugzilla.gnome.org

Did you reported this?

[Comment 5 Nathan Phillip Brink (binki) (RETIRED) 2011-11-22 21:49:42 UTC]

(In reply to comment #4)
> (In reply to comment #3)
> > I would send this to upstream also: bugzilla.gnome.org
> 
> Did you reported this?

No. I might if I have time, unfortunately I might forget to do so altogether :-/.

[Comment 6 Nathan Phillip Brink (binki) (RETIRED) 2011-11-23 03:04:17 UTC]

Upstream bug filed at https://bugzilla.gnome.org/show_bug.cgi?id=664606, thanks for the poking ;-).

[Comment 7 Pacho Ramos 2012-01-30 11:06:43 UTC]

(In reply to comment #6)
> Upstream bug filed at https://bugzilla.gnome.org/show_bug.cgi?id=664606, thanks
> for the poking ;-).

Looks like upstream think honor CPPFLAGS wouldn't be needed, could you reply to him there please? Thanks

[Comment 8 Nathan Phillip Brink (binki) (RETIRED) 2012-02-01 15:48:35 UTC]

Thanks again for the poke.

Upstream has committed the fix to their git repo, in case if that is justification enough to start applying the upstreamed patch to gentoo-x86 ;-).

[Comment 9 Pacho Ramos 2012-02-02 10:18:10 UTC]

From my point of view, feel free to apply it to latest vala version if you need the fix already, if not, we will probably wait for next version ;)

[Comment 10 Pacho Ramos 2012-04-01 17:32:21 UTC]

vala-0.16 was added