Arch Linux

Midori starts fine for me. Maybe you should rename/remove ~/.config/midori/ and try again.

It doesn't work. I deleted the directory, tried to restart, and a new empty folder is created as it segfaults. I tried to rebuild midori against this machine and the same thing. Here's the stack trace (snipped, of course):

(gdb) run
Starting program: /usr/bin/midori
warning: Lowest section in /usr/lib/libicudata.so.40 is .hash at 0000000000000120
[Thread debugging using libthread_db enabled]
[New Thread 0x7fa2439b8760 (LWP 18741)]
Program received signal SIGSEGV, Segmentation fault.
[Switching to Thread 0x7fa2439b8760 (LWP 18741)]
0x000000000042d12d in ?? ()
(gdb) (gdb) where
Undefined command: "". Try "help".
(gdb) #0 0x000000000042d12d in ?? ()
(gdb) #1 0x00000000004201d5 in ?? ()
(gdb) #2 0x00007fa243450a14 in g_type_create_instance ()
(gdb) from /usr/lib/libgobject-2.0.so.0
Undefined command: "from". Try "help".
(gdb) #3 0x00007fa2434363eb in g_object_constructor ()
(gdb) from /usr/lib/libgobject-2.0.so.0
Undefined command: "from". Try "help".
(gdb) #4 0x00007fa243436a33 in g_object_newv () from /usr/lib/libgobject-2.0.so.0
(gdb) #5 0x00007fa24343747b in g_object_new_valist ()
(gdb) from /usr/lib/libgobject-2.0.so.0
Undefined command: "from". Try "help".
(gdb) #6 0x00007fa2434376cc in g_object_new () from /usr/lib/libgobject-2.0.so.0
(gdb) #7 0x00000000004159e1 in ?? ()
(gdb) #8 0x00007fa2404053f6 in __libc_start_main () from /lib/libc.so.6
(gdb) #9 0x00000000004102f9 in ?? ()
(gdb) #10 0x00007fff4baf21d8 in ?? ()
(gdb) #11 0x000000000000001c in ?? ()
(gdb) #12 0x0000000000000001 in ?? ()
(gdb) #13 0x00007fff4baf29cf in ?? ()
(gdb) #14 0x0000000000000000 in ?? ()

Do you use the libwebkit from extra or a version from AUR? Any other changes from official repositories? So far I can't confirm this.

same problem here, using libwebkit from extra (and any other deps from official repositories)
I've tried to rebuilding libwebkit and midori based on the PKGBUILDs from arch-extra, but I still get segmentation fault immediately after starting midori.

Same problem here too. Upgraded to 0.0.20-1 and now Midori segfaults on starts up (it used to work fine in the previous version).

I'm using libwebkit 1.0.1-1 from extra.

I'm also using everything from extra. I don't have any special versions of much on this system, and most of what I have is unrelated to midori (FFmpeg, x264, etc).

Same issue

See the upstream bug http://www.twotoasts.de/bugs/index.php?do=details&task_id=97

It doesn't seem like an arch only issue.

I can't reproduce this and I really can't think of anything that I could do to fake that error.

Can one of you make a backtrace with full debugging symbols? It looks like it is somewhere in Glib.

Might be also worth checking out what optional libraries you have, specially GIO and/ or GVfs since that used to cause weird bugs in the past already.

I for one have Gtk/ Glib from svn trunk, it works - obviously - perfectly here, disabling any optional libraries doesn't change anything for me, related to that crash anyway.

Like on the upstream report, I get the exact same output with ultradebug enabled.

Here's the backtrace:
warning: Lowest section in /usr/lib/libicudata.so.40 is .hash at 0000000000000120
[Thread debugging using libthread_db enabled]
[New Thread 0x7f3aeb722760 (LWP 1517)]

Program received signal SIGSEGV, Segmentation fault.
[Switching to Thread 0x7f3aeb722760 (LWP 1517)]
0x000000000042cffd in sokoke_gio_supports_http ()
Current language: auto; currently asm
(gdb) bt
#0 0x000000000042cffd in sokoke_gio_supports_http ()
#1 0x00000000004200a5 in midori_browser_init ()
#2 0x00007f3aeb1b6a14 in g_type_create_instance () from /usr/lib/libgobject-2.0.so.0
#3 0x00007f3aeb19c3eb in g_object_constructor () from /usr/lib/libgobject-2.0.so.0
#4 0x00007f3aeb19ca33 in g_object_newv () from /usr/lib/libgobject-2.0.so.0
#5 0x00007f3aeb19d47b in g_object_new_valist () from /usr/lib/libgobject-2.0.so.0
#6 0x00007f3aeb19d6cc in g_object_new () from /usr/lib/libgobject-2.0.so.0
#7 0x00000000004158b9 in main ()

So it seems to fail when trying to call GIO. What exact version of GLib is that? Maybe you could try updating Glib and see if that helps.

It's glib2-2.16.5-1 from core - so nothing to update. Maybe it's just a missing dependency?

oh, and I'm using the same package versions on i686, where it works fine. so it is x86_64 related in some way.

Unable to reproduce the problem whatsoever I removed the relevant code, it was a minor visual clue intended to hide the View Source feature when http was not supported, doesn't hurt much if that fixes the peculiar crash.

Anyone affected update Midori from git.xfce.org/kalikiana/midori please. The crash according to the backtrace won't happen anymore.

The git version actually solves the problem here. No crash anymore. Thanks for the quick fix.

Most happy to hear that. ^_^

The git version (built from AUR) does not solve the problem here.

skottish: I haven't used the AUR PKGBUILD, I just throw out the source-archive from the extra-PKGBUILD and added
kalikiana's git repository given above, and of course removed the ${pkgver} from the cd command, I've attached the resulting PKGBUILD

oh... and of course you could remove that CFLAGS="${CFLAGS} -ggdb}" - I've used that for verbose backtracing as in comment #32043

Ok, fixed PKGBUILD here

While you're at it..

> pkgname=midori
> pkgver=20080828
> pkgrel=1
> pkgdesc="A lightweight web browser based on Gtk WebKit"
> arch=('i686' 'x86_64')
> url="http://www.twotoasts.de/index.php?/pages/midori_summary.html"
> license=('GPL')

Midori is licensed under the LGPL.

> install=midori.install
> depends=('e2fsprogs' 'gtksourceview2' 'hicolor-icon-theme' 'libwebkit')
> _optdepends=('gvfs: support for source viewing')

You might mention favicons, too, since that's half of what GIO is good for.

> makedepends=('pkgconfig' 'libxml2' 'intltool')

And a 'python' here, it's a build dependency.

> source=()
> md5sums=()

I guess archlinux is probably going to release another PKGBUILD version anyway, the posted PKGBUILD does not use archlinux' guideline for git-PKGBUILDs. Just posted the PKGBUILD here to help skottish to build the fixed version.

Good to know about LGPL and python. Even better to know about 'favicon' - unfortunately my local midori doesn't support favicons yet, and there is no package called favicon in archlinux. Where can I find more information about it?

A favicon is just a term for these little icons you see on the tabs and in the location entry. What I mean is, GIO is what Midori uses to load these icons from the network. And more specifically, you should also have GVfs installed because that implements the actual HTTP functionality of GIO, which otherwise can only load local files.

going to close this bug as "upstream". any concerns?

Chris: looking at the git please tell me what commit I should apply to fix our 0.0.20 pkg or will see very soon 0.0.21?

Agreed on upstream. It's fine if it's closed to me.

Andreas, I just made a new release, 0.0.21, so you don't have to patch :)

http://www.twotoasts.de/index.php?/archives/8-Fast-and-fervent-follow-up.html

Confirmed working on 64 bit. Thanks midori devs for taking care of this.

One step closer to losing the other browsers. Hurray!