GNOME Bugzilla – Bug 669319
Some cards are misrendered
Last modified: 2012-03-16 16:27:01 UTC
Created attachment 206688 [details] Screenshot showing bonded deck Some cards are no longer (this means: it worked for 3.2.x) rendered correctly in klondike when using aisleriot from git master. In *ang*lo themes cards numbered 10 and 5 are affected, in bonded e.g. 2,3,4,5.
Created attachment 206689 [details] Screenshot showing gnomeangelo deck
Not reproducible here, on aisleriot git master, with librsvg git master. Which versions do you have?
Git master versions, too. I'm running a complete self-compiled gnome 3.3.x stack (no package older than one week) on debian sid/experimental.
Addition: The builds I'm using are comparable to jhbuild ones.
Hmm. A brand-new compiler maybe? Strange CFLAGS ? Some exotic (ie any non-x86/x86_64) arch? Would be good if you could try librsvg from git master, and see if it shows in the rsvg-view-3 programme run on aisleriot/cards/bonded.svg --- it should, since aisleriot is just using librsvg. Then, when that reproduces, you should git bisect between the 3.2.1 tag (assuming you mean that version when you said that 2.32 worked) and git master, to see which commit it is that introduces the problem for you.
Sorry, correction: "... between the 2.32.1 tag... " obviously.
I'm on i386, using gcc-4.6.2. CFLAGS used for librsvg are -g -O2 -fstack-protector --param=ssp-buffer-size=4 -Wformat -Wformat-security -Werror=format-security rsvg-view-3, version 2.35.1-26-ga3c9aaa produced with these option renders both, gnomeangelo and bonded.svg{,z} fine. Aisleriot is compiled against this version of librsvg.
ga3c9aaa isn't a revision in git.gnome.org's librsvg repo... This is very odd. Maybe it's because aisleriot only renders part of the image; does rsvg-view --id "#diamond_10" (or "#club_5", etc) on these files show the problem, maybe?
Created attachment 206718 [details] Screenshot from rsvg-view-3 > ga3c9aaa isn't a revision in git.gnome.org's librsvg repo... It is, but without the leading g added by git describe. > does rsvg-view --id "#diamond_10" (or "#club_5", etc) on these files show the > problem, maybe? No. The cards are rendered fine.
No idea what's wrong, then.
You could try running aisleriot under valgrind and attach the log here. See https://live.gnome.org/Valgrind for which options to use. Do use --num-callers=30 (at least!) and --track-origins=yes, too.
I am experiencing the same issues along with card #7 is also distorted.
>You could try running aisleriot under valgrind I tried it, unfortunately it segfaults when running under valgrind, so that there is not much in the logs. I even did a clean jhbuild build of aisleriot (with a bootstrapped guile), it still segfaults when running under valgrind. I've tried to pass different options, no difference. Other applications do not show this behaviour when using valgrind, so most likely no problem with the valgrind binary. Anyway, there are some news concerning this bug. I'm using intel graphics and recently I've tried out sna enabled drivers and what shall I say - here aisleriot renders the cards correctly. Maybe it is a bug in intel driver (uxa), though it is strange, that aisleriot should be the only application where one can observe such errors ...
(In reply to comment #13) > >You could try running aisleriot under valgrind > I tried it, unfortunately it segfaults when running under valgrind, I think you may need to upgrade your libgc (I remember having had this problem before, too, and updating & recompiling the gc library solved it.) > Anyway, there are some news concerning this bug. I'm using intel graphics and > recently I've tried out sna enabled drivers and what shall I say - here > aisleriot renders the cards correctly. Maybe it is a bug in intel driver (uxa), > though it is strange, that aisleriot should be the only application where one > can observe such errors ... Yeah, a graphics driver bug sounds like a good explanation.
Created attachment 206824 [details] Valgrind output Yeah, libgc needed an update. For completeness I'm attaching the valgrind output as requested.
Created attachment 206826 [details] Maybe a better log
Log doesn't show anything interesting... the driver wouldn't show up in there, though.
*** Bug 669600 has been marked as a duplicate of this bug. ***
Looks like a driver issue, see https://bugzilla.redhat.com/show_bug.cgi?id=796652#c3 -> NOTGNOME.
Debian bug for this: http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=656800