[SOLVED]Latest additions to testing crashes darktable

spsf64 · 2018-03-16 16:21:56

Running XFCE here.
I just enable testing repo to check the gnome software 3.28
Everything seems to be working fine, except for darktable.
I think it is related to the latest GTK?
It crashes as soon it is launched, the msgs from terminal are:

>> darktable
[defaults] found a 64-bit system with 32868608 kb ram and 8 cores (0 atom based)
[defaults] setting very high quality defaults

(darktable:7358): Gtk-WARNING **: 12:19:56.724: Theme parsing error: gtk.css:38:0: expected '}' after declarations
free(): invalid pointer
Aborted (core dumped)

Anyone else experiencing this?
TIA

Last edited by spsf64 (2018-03-17 10:44:06)

loqs · 2018-03-16 16:35:11

What is the backtrace from the coredump? Core_dump#Examining_a_core_dump

spsf64 · 2018-03-16 16:42:27

loqs wrote:

What is the backtrace from the coredump? Core_dump#Examining_a_core_dump

I'm no pro, lol! Is this right?
Here you go:

me on Fri Mar 16 at 12:39 PM in ~/Desktop
>> coredumpctl info 7358
PID: 7358 (darktable)
UID: 1001 (me)
GID: 1001 (me)
Signal: 6 (ABRT)
Timestamp: Fri 2018-03-16 12:19:57 EDT (19min ago)
Command Line: darktable
Executable: /usr/bin/darktable
Control Group: /user.slice/user-1001.slice/session-c2.scope
Unit: session-c2.scope
Slice: user-1001.slice
Session: c2
Owner UID: 1001 (me)
Boot ID: 87c67ff3878b4b47857b74e4ab25314d
Machine ID: "removed by me"
Hostname: me-pc
Storage: /var/lib/systemd/coredump/core.darktable.1001.87c67ff3878b4b478>
Message: Process 7358 (darktable) of user 1001 dumped core.

Stack trace of thread 7358:
#0 0x00007f4d1fef1860 raise (libc.so.6)
#1 0x00007f4d1fef2ec9 abort (libc.so.6)
#2 0x00007f4d1ff34437 __libc_message (libc.so.6)
#3 0x00007f4d1ff3a78b malloc_printerr (libc.so.6)
lines 1-23

loqs · 2018-03-16 17:03:21

It looks like some of the strack trace is missing. Please also use code tags for commands and their output.

spsf64 · 2018-03-16 17:09:10

loqs wrote:

It looks like some of the strack trace is missing. Please also use code tags for commands and their output.

Sorry, I didn't check the whole thing

https://pastebin.com/6uP5s0XA

loqs · 2018-03-16 17:23:52

                #0  0x00007f4d1fef1860 raise (libc.so.6)
                #1  0x00007f4d1fef2ec9 abort (libc.so.6)
                #2  0x00007f4d1ff34437 __libc_message (libc.so.6)
                #3  0x00007f4d1ff3a78b malloc_printerr (libc.so.6)
                #4  0x00007f4d1ff3c24c _int_free (libc.so.6)
                #5  0x00007f4d1e59f10c n/a (libcairo.so.2)
                #6  0x00007f4d1e509fd6 cairo_font_face_destroy (libcairo.so.2)
                #7  0x00007f4d1e55e434 n/a (libcairo.so.2)
                #8  0x00007f4d1e509fd6 cairo_font_face_destroy (libcairo.so.2)
                #9  0x00007f4d1e50c0ea n/a (libcairo.so.2)
                #10 0x00007f4d1e50c547 n/a (libcairo.so.2)
                #11 0x00007f4d1e565882 cairo_restore (libcairo.so.2)
                #12 0x00007f4d20645228 dt_view_manager_expose (libdarktable.so)

Looking at call #6-#1 and the error message "free(): invalid pointer" cairo_font_face_destroy or whatever call #5 is passes an invalid pointer to free causing the coredump.
Might be there is an earlier issue that causes corruption but upstream is best suited to deal with that. Would be helpful to upstream to provide a backtrace with debug symbols
they appear to be missing from both the cairo and gtk3 packages.

spsf64 · 2018-03-16 17:54:36

@loqs, thanks.
I'll try to report upstream

progandy · 2018-03-16 17:58:32

Did you try different themes? Maybe this problem is only visible with some gtk themes. You should still report it since a theme should not be able to cause a crash.

spsf64 · 2018-03-16 18:02:43

@progandy, yes, also checked with adwaita (the latest from testing) theme, same problem...

Last edited by spsf64 (2018-03-16 18:40:55)

loqs · 2018-03-16 18:37:10

@progandy which upstream would you report it to?

spsf64 · 2018-03-16 18:42:29

@loqs, just found this page for bugs
https://redmine.darktable.org/projects/ … acker_id=1
Nothing on github page
We have to register to report...

Last edited by spsf64 (2018-03-16 18:43:43)

loqs · 2018-03-16 19:13:25

I would report it to either https://gitlab.gnome.org/GNOME/gtk/issues as a gtk issue or https://bugs.freedesktop.org/describeco … duct=cairo but not sure which component
Edit:
https://bugs.freedesktop.org/show_bug.cgi?id=104616 looks somewhat similar not sure if the commit is in range.

https://git.archlinux.org/svntogit/pack … 87517983d9
This seems odd the commit id has not changed

Edit:
could also be https://bugs.freedesktop.org/show_bug.cgi?id=105084 building cairo with https://cgit.freedesktop.org/cairo/comm … a15517fa3b would test that.
Edit2:
Missed the intention of the commit was only to fix the pkgver despite the comment.

Last edited by loqs (2018-03-16 19:50:53)

progandy · 2018-03-16 19:37:43

loqs wrote:

I would report it to either https://gitlab.gnome.org/GNOME/gtk/issues as a gtk issue or https://bugs.freedesktop.org/describeco … duct=cairo but not sure which component

I'd start with cairo since there seem to be quite a few memory leak fixes and other changes in font code. If these changes are valid, this might have exposed a bug (double-free?) in another component like pango, or even darktable)

This seems odd the commit id has not changed

Same commit, but a tag has been added to an older commit so that the version is calculated correctly

Edit: darktable is at least starting here with gtk3 3.22.28-1, cairo 1.15.10-1. I haven't done the upgrade yet and I use i3 and not gnome.

Last edited by progandy (2018-03-16 19:44:24)

loqs · 2018-03-16 19:50:00

@spsf64 this should build an updated cairo package with a fix for one of the possible causes

pacman -S devtools
git clone git://git.archlinux.org/svntogit/packages.git --single-branch --branch "packages/cairo"
cd packages/trunk/ #change _commit to _commit=1ed124ace201946177eab962d3b07bb2fedf9026
gnome-unstable-x86_64-build
pacman -U cairo-1.15.10+54+g1ed124ace-1-x86_64.pkg.tar.xz

Edit:
You could also try downgrading cairo to 1.15.10-1 the soname has not been bumped so it should be compatible if it is not you can restore the package with pacman -S cairo.
Edit:
Possibly also https://bugs.archlinux.org/task/57867

Last edited by loqs (2018-03-16 22:11:17)

spsf64 · 2018-03-17 10:45:44

@loqs Thanks, the problem was cairo!
Updated to the latest version from testing and darktable is working perfectly!

Last edited by spsf64 (2018-03-17 10:46:08)

Arch Linux

#1 2018-03-16 16:21:56

[SOLVED]Latest additions to testing crashes darktable

#2 2018-03-16 16:35:11

Re: [SOLVED]Latest additions to testing crashes darktable

#3 2018-03-16 16:42:27

Re: [SOLVED]Latest additions to testing crashes darktable

#4 2018-03-16 17:03:21

Re: [SOLVED]Latest additions to testing crashes darktable

#5 2018-03-16 17:09:10

Re: [SOLVED]Latest additions to testing crashes darktable

#6 2018-03-16 17:23:52

Re: [SOLVED]Latest additions to testing crashes darktable

#7 2018-03-16 17:54:36

Re: [SOLVED]Latest additions to testing crashes darktable

#8 2018-03-16 17:58:32

Re: [SOLVED]Latest additions to testing crashes darktable

#9 2018-03-16 18:02:43

Re: [SOLVED]Latest additions to testing crashes darktable

#10 2018-03-16 18:37:10

Re: [SOLVED]Latest additions to testing crashes darktable

#11 2018-03-16 18:42:29

Re: [SOLVED]Latest additions to testing crashes darktable

#12 2018-03-16 19:13:25

Re: [SOLVED]Latest additions to testing crashes darktable

#13 2018-03-16 19:37:43

Re: [SOLVED]Latest additions to testing crashes darktable

#14 2018-03-16 19:50:00

Re: [SOLVED]Latest additions to testing crashes darktable

#15 2018-03-17 10:45:44

Re: [SOLVED]Latest additions to testing crashes darktable

Board footer