GNOME Bugzilla – Bug 734262
Shotwell sleeps on import
Last modified: 2014-09-07 05:27:05 UTC
I transferred some images to my system today and then ran shotwell. Shotwell is set to watch the directories and began to auto-import the images. However, before it had imported any, it just hung. Using `ps`, I see that the process is in "interruptible sleep". Eventually the "program is not responding" window came up - I have tried just waiting, but it doesn't appear to ever resume. I've run it multiple times and the result is the same. I've reset all the gsetting schemas, removed .local/share/shotwell but there isn't any change. I'm running it in gdb now, but it's still just sitting there. [asinha@ankur-laptop ~]$ ps aux | egrep shotwell asinha 25188 2.0 2.2 407120 182868 pts/8 S 16:44 0:03 gdb shotwell asinha 25214 6.5 1.5 1699812 126568 pts/8 Sl+ 16:44 0:09 /usr/bin/shotwell shotwell-0.18.0-2.fc20.x86_64 I'm aware that 0.18.1 has been released, but it hasn't hit the Fedora 20 Gnome 3.14 copr yet. I'll update whenever it's available. Any hints on the issue? Please let me know if there's any more data that I can provide you. My images are pretty much the same, just added a new set. Shotwell used to work just fine before. Thanks, Ankur
When I finally killed shotwell in gdb with a Ctrl C, it was stuck here: ^C Program received signal SIGINT, Interrupt. syscall () at ../sysdeps/unix/sysv/linux/x86_64/syscall.S:38 38 cmpq $-4095, %rax /* Check %rax for error. */ I'm attaching the complete multi-threaded stack trace.
Created attachment 282496 [details] Complete multithreaded stack trace
still happening: shotwell-0.19.0-1.fc21.x86_64 It's unusable at the moment. Can someone please take a look at this? Thanks, Warm regards, Ankur
I'm experiencing the same exact issue. I'm also happy to provide traces and do some debugging to help solve this issue.
This is going to be a blocking for Fedora 21, with its Beta at the end of the month, so marking as Urgent. Ankur, which was the latest version of shotwell to work correctly for you? Does compiling shotwell manually solve the problem?
0.18 didn't work, that's when I filed the bug. Fedora only saw 0.15 before that, so I think that would've worked. I'm not a 100% sure, though. I haven't tried to compile it by hand. I'll do that now and let you know. http://koji.fedoraproject.org/koji/packageinfo?packageID=9012 Thanks for looking into this, Bastien, Ankur
If you are looking for more data points, we have 0.18.1 for F19 and F20 here: http://http://copr.fedoraproject.org/coprs/thm/shotwell and 0.19.0 for F19 and F20 here: http://http://copr.fedoraproject.org/coprs/thm/shotwell-unstable
The bug was filed at shotwell-0.18.0-2.fc20.x86_64 (see comment 1). I haven't been able to use shotwell since. I'll see if I can try out 0.15 again. I think there was a libgexiv soname bump there though, so it may take me some time.
I proposed this as a Fedora 21 Final (not beta) blocker. It very clearly fits one of our final blocker criteria, so Shotwell will surely be dropped from the default install in F21 unless this is fixed before then. I noticed a possibly-relevant comment in another bug: https://bugzilla.gnome.org/show_bug.cgi?id=733500#c3 This makes me suspect you can fix this by simply using a newer valac to generate the next release.
Oh I should add: to reproduce this, you probably need to run with a mid/late glib 2.41 development release.
shotwell-0.19.0-1.fc1 has been compiled with vala-0.25.1, and the fix from bug 733500 is in 0.25.2. So, simply rebuilding shotwell should help? I can easily kick that.
(In reply to comment #11) > So, simply rebuilding shotwell should help? I can easily kick that. Yes, I think so. I didn't realize when I wrote comment #9 that Shotwell has a hand-written Makefile and does not distribute any C. So if the downstream rebuild fixes this, then this bug can be closed. (This is unconventional: Vala projects normally distribute C as well as Vala, so a new upstream release would be needed with the C compiled from a newer valac.)
Ok, here we go: http://koji.fedoraproject.org/koji/buildinfo?buildID=575633 (f21, built against vala 0.25.2) http://koji.fedoraproject.org/koji/buildinfo?buildID=575631 (f22, built against vala 0.25.3)
I tested the rawhide build and it indeed fixed the problem for me. This means nothing needs to be done upstream, yay! *** This bug has been marked as a duplicate of bug 733500 ***
+1 tested the new build on F21 and it works. Thanks a bunch! :D