Bug 1779

Summary: gsettings-data-convert segfault on sdk image
Product: [Build System, Metadata & Runtime] OE-Core Reporter: Jiajun Xu <jiajun.xu>
Component: coreAssignee: Edwin Zhai <edwin.zhai>
Status: VERIFIED FIXED QA Contact:
Severity: normal    
Priority: Medium CC: dexuan.cui, hongnax.xu, jessica.zhang, meta.mr.watcher, meta.watcher, richard.purdie, sgw, shane.wang
Version: unspecified   
Target Milestone: 1.2 M1   
Hardware: x86   
OS: Multiple   
Whiteboard: patch out. Dec 29
OS type for building Yocto: --- Type of Regression: ---
Verified: Documentation change: ---

Description Jiajun Xu 2011-11-21 22:49:45 UTC 
Tree/Branch: Poky/master
Commit: f2d077255d9ddf9b87e9a781b989bcce69047cf5

With latest commit, gsteetings-data-convert would report segfault on sdk image. I find the issue on both qemux86 and qemux86-64 sato-sdk image. Meanwhile, there is no such issue on sato image.
Comment 1 Jiajun Xu 2011-12-05 17:31:37 UTC 
Sorry I make a mistake here, it should be "gsettings-data-convert", which will segfault on sdk image. It comes from gconf-3.2.3.
Comment 2 Jiajun Xu 2011-12-05 18:01:11 UTC 
I checked latest qemux86 sato-sdk images from autobuilder, commit e5b9ac3b71b6c16fb9a629f12e4d83103db2d5b0. The issue still exists.
Comment 3 Shane Wang 2011-12-13 08:07:29 UTC 
Jiajun, What is gsettings-data-convert and which recipe is it from?
Comment 4 Jiajun Xu 2011-12-13 16:41:05 UTC 
(In reply to comment #3)
> Jiajun, What is gsettings-data-convert and which recipe is it from?

It comes from gconf-3.2.3, which is used to migrate users' gconf database to gsettings.
Comment 5 Shane Wang 2011-12-15 19:27:39 UTC 
Give it to Edwin since Edwin is maintaining Gnome stuffs.
Comment 6 Shane Wang 2011-12-15 20:11:37 UTC 
Edwin, Dexuan got some ideas on it.
Comment 7 Dexuan Cui 2011-12-15 22:59:37 UTC 
(In reply to comment #6)
> Edwin, Dexuan got some ideas on it.

sato is ok because it doesn't have /usr/share/GConf/gsettings/ -- sato-sdk has this and gets this issue. Maybe there is something missing?
Comment 8 Jessica 2011-12-20 11:44:02 UTC 
Any update on the bug? It's still in the new state
Comment 9 Edwin Zhai 2011-12-26 18:56:44 UTC 
segmentation comes from NULL pointer reference.

---- gsettings-data-convert ----
main=>
handle_dir=>
handle_file=>
---- glib ----
g_settings_new
g_settings_delay(settings)=>
g_delayed_settings_backend_new(settings->priv->backend)

The issue here is the settings->priv->backend is NULL, so reference in g_delayed_settings_backend_new cause seg fault.

The root cause for the NULL is that schemas is not properly installed with error mesg from g_settings_new:
(gsettings-data-convert:1205): GLib-GIO-CRITICAL **: Settings schema 'org.gnome.crypto.cache' is not installed

To install schemas, we need file "gschemas.compiled" under /usr/share/glib-2.0/schemas. (see glib's source code initialise_schema_sources @ gio/gsettingsschema.c)

This file made by glib-compile-schemas from glib-utils.
Comment 11 Hongna Xu 2012-01-18 18:05:35 UTC 
The bug is not reproducible on crownbay_sato-sdk, qemux86 and qemux86-64 sato-sdk image,so set the status closed.If anyone encounter this bug, pls reopen it.

Commit information:

Tree/Branch: Poky/1.2_M2
Commit: 0f4d99d207b224bb9ce23de00a48f795ae20b3a0
Comment 12 Edwin Zhai 2012-01-18 18:18:19 UTC 
Hongna,
Fixed+Verified is more accurate than "not reproducible" here:)

(In reply to comment #11)
> The bug is not reproducible on crownbay_sato-sdk, qemux86 and qemux86-64
> sato-sdk image,so set the status closed.If anyone encounter this bug, pls
> reopen it.
> 
> Commit information:
> 
> Tree/Branch: Poky/1.2_M2
> Commit: 0f4d99d207b224bb9ce23de00a48f795ae20b3a0
Comment 13 Hongna Xu 2012-01-18 18:38:41 UTC 
(In reply to comment #12)
> Hongna,
> Fixed+Verified is more accurate than "not reproducible" here:)

I am sorry for that, I will take it more seriously next time :(