[Dapper] qgo segmentation fault

Bug #4048 reported by Loïc Martin
18
Affects Status Importance Assigned to Milestone
qgo (Ubuntu)
Fix Released
Medium
MOTU
scim (Ubuntu)
Invalid
Medium
Unassigned

Bug Description

In current Dapper (RC) qgo segfaults at startup

(In Ubuntu 5.10 Breezy qgo version 1.0.0-r2.2 (and and qgo-1.03-R2) exited at launch reporting a segmentation fault too)

Changed in qgo:
assignee: nobody → motu
Revision history for this message
Loïc Martin (loic-martin3) wrote :

I can confirm segfault is caused by scim.

Not just a scim install (qgo stills works here with scim installed on a clean Ubuntu 5.10 Breezy install) but when configuring scim (that is, configuring chinese locale and getting scim-gtk2-immodule configured + configuring scim to work with non gtk2 apps like openoffice).

That should make it easier to track the bug (at least it's easy to reproduce).

description: updated
description: updated
Revision history for this message
Loïc Martin (loic-martin3) wrote :

Just to correct a possible mistake. scim is not necessarily the one causing the problem, it could be related to the configuration necessary to get scim running on breezy (locales? or maybe these lines :
export GTK_IM_MODULE=scim
export XMODIFIERS=@im=scim
)

Revision history for this message
Ming Hua (minghua) wrote :

As noted in malone bug #2246, GTK_IM_MODULE=scim can cause problems if scim is loaded in an app linked to libstdc++5. GTK_IM_MODULE=xim would be a workaround.

Also XMODIFIERS=@im=scim is WRONG and useless. The correct way to specify XMODIFIERS for scim is XMODIFIERS=@im=SCIM, uppercase. (This is clearly documented in /usr/share/doc/scim/README.Debian.gz by the way, whatever doc you get this setting from should be fixed.)

Revision history for this message
Loïc Martin (loic-martin3) wrote :

Thanks for the information.

However, with GTK_IM_MODULE=xim I can't input chinese in Open Office (and other apps like Thunderbird). So qgo will still segfault.

Is there a way to start qgo wihle export GTK_IM_MODULE=scim configured in ./.gnomerc?

If so, we could close the bug

Revision history for this message
Loïc Martin (loic-martin3) wrote :

Hi,
On a clean breezy install, with scim installed from http://svn.ubuntu.org.cn/ubuntu-cn/d...n/binary-i386/
then configured following your advices, that is

gedit ~/.gnomerc

then adding the following 3 lines:

export XMODIFIERS="@im=SCIM"
export GTK_IM_MODULE=scim
scim -d

(the only thing I changed is the second line, where scim replaces "xim"
that was necessary to get scim input in open office and thunderbird)

I don't even need to start gqo using
GTK_IM_MODULE=xim qgo

just qgo on the command line is enough.

On the other hand, that means qgo would still segfault. It does indeed do so on my previous (the third) breezy install - still on my hd if somebody (possibly a qgo developper) wants to investigate.

So the bug could be closed, but it's still there. It might have something to do with either :
- me running :
export GTK_IM_MODULE=scim
export XMODIFIERS=@im=scim

- or me running dpkg-reconfigure locales and selecting english, french and chinese.

These are the last 2 actions I did before qgo segfaulted (plus installing scim but now we know that's not the problem).

I know it looks hopeless to understand, so there's few hope it will be fixed.

Since you are packaging scim for Ubuntu, maybe the ubuntu documentation list would listen to you more if you propose setting up a howto on how to configure scim along with acroread and qgo in breezy.
I can write it (where to get the packages, how to install them and add a .gnomerc file - even though that method is not the best for KDE users - and modify /usr/bin/acroread)
then I can send it to you for proof reading.

I'd rather that than starting a new thread in the forums, cause I would run the possibility to mess up things for other people with wrong advices.
Thank you for your work. I really enjoy scim in Ubuntu, and couldn't do without it.
Loïc

Revision history for this message
ZhengPeng Hou (zhengpeng-hou) wrote : Re: [Bug 4048] qgo segmentation fault

在 星期二 27 十二月 2005 00:12,您写道:
> Public bug report changed:
> https://launchpad.net/malone/bugs/4048
>
> Comment:
> Hi,
> On a clean breezy install, with scim installed from
> http://svn.ubuntu.org.cn/ubuntu-cn/d...n/binary-i386/ then configured
> following your advices, that is
>
> gedit ~/.gnomerc
>
> then adding the following 3 lines:
>
> export XMODIFIERS="@im=SCIM"
> export GTK_IM_MODULE=scim
> scim -d
>
if you use gdm or kdm , we recommend you put them in follow:
/etc/X11/Xsession.d/95xinput
and do remmember make it excuteable
if you use startx , put them in $HOME/.xinitrc
> (the only thing I changed is the second line, where scim replaces "xim"
> that was necessary to get scim input in open office and thunderbird)
>
> I don't even need to start gqo using
> GTK_IM_MODULE=xim qgo
>
> just qgo on the command line is enough.
>
>
> On the other hand, that means qgo would still segfault. It does indeed do
> so on my previous (the third) breezy install - still on my hd if somebody
> (possibly a qgo developper) wants to investigate.
>
> So the bug could be closed, but it's still there. It might have something
> to do with either : - me running :
> export GTK_IM_MODULE=scim
> export XMODIFIERS=@im=scim
>
> - or me running dpkg-reconfigure locales and selecting english, french
> and chinese.
>
> These are the last 2 actions I did before qgo segfaulted (plus
> installing scim but now we know that's not the problem).
>
> I know it looks hopeless to understand, so there's few hope it will be
> fixed.
>
> Since you are packaging scim for Ubuntu, maybe the ubuntu documentation
> list would listen to you more if you propose setting up a howto on how to
> configure scim along with acroread and qgo in breezy. I can write it (where
> to get the packages, how to install them and add a .gnomerc file - even
> though that method is not the best for KDE users - and modify
> /usr/bin/acroread) then I can send it to you for proof reading.
>
> I'd rather that than starting a new thread in the forums, cause I would
> run the possibility to mess up things for other people with wrong advices.
> Thank you for your work. I really enjoy scim in Ubuntu, and couldn't do
> without it. Loïc

Revision history for this message
Ming Hua (minghua) wrote :

On Tue, Dec 27, 2005 at 12:30:04AM -0000, freeflying wrote:
> 在 星期二 27 十二月 2005 00:12,您写道:
> > Hi,
> > On a clean breezy install, with scim installed from
> > http://svn.ubuntu.org.cn/ubuntu-cn/d...n/binary-i386/ then configured
> > following your advices, that is
> >
> > gedit ~/.gnomerc
> >
> > then adding the following 3 lines:
> >
> > export XMODIFIERS="@im=SCIM"
> > export GTK_IM_MODULE=scim
> > scim -d
> >
> if you use gdm or kdm , we recommend you put them in follow:
> /etc/X11/Xsession.d/95xinput

I am not sure using /etc/X11/Xsession.d/ is a good idea because (1) it's
a system-wide file and will affect every user; and (2) written poorly
(without extra checking) it may interfere with other settings from
m17n-env, im-switch and possible future auto-start support from scim
itself.

> and do remmember make it excuteable

And I see absolutely no reason to make the files in /etc/X11/Xsession.d/
executable.

> if you use startx , put them in $HOME/.xinitrc
[long quote snipped]

Ming
2005.12.30

Revision history for this message
Loïc Martin (loic-martin3) wrote : Re: qgo segmentation fault

An update to the situation :
qGo 1.0.2 was really unstable (frequent crashes when observing games, to the extent that I didn't even dare to start a game).
Then, without any (apparent) change in the system it stopped working and reverted to the situation I had on my previous install (segfault at launch).
Following the advices from qGo's only developper (upstream) I'm trying to compile qGo with debug enabled - still unsuccessful atm, (never really was anyways).
The last version (1.0.4) solves a bug which could be the one at fault in my case. OTOH, it's only available in Dapper.

Two solutions :
- can I be provided a binary with debug enabled so I can send the relevant information to upstream?
- could 1.0.4 be backported so I can test it and make sure the bug is solved?

On the other hand I'll still try to build the binary myself during that time, but I'm not so optimistic.
Loïc

Phil Bull (philbull)
Changed in qgo:
status: Unconfirmed → Needs Info
Revision history for this message
Loïc Martin (loic-martin3) wrote :

Thanks to upstream advices, I was able to compile a debug enabled version of last version of qGo. Same problem, the program crashes at startup. I sent upstream every information I could get (using Kdevelop) but it turned out it's not sufficient enough. I would have to learn programming and debug the program myself, which isn't going to happen anytime soon.

The Status is "Need Info". Since all the relevant info a normal user could give has been given to upstream, I vote for changing it from "Need Info" to "Need Programmer"

Revision history for this message
Reinhard Tartler (siretart) wrote :

seems to me a scim related problem, following the bugtrail. I just created a bugtask for scim therefore

Revision history for this message
Loïc Martin (loic-martin3) wrote :

Hi,
I'm sorry the bugtrail looks like that - I realise I haven't been clear enough when I said it might not be scim's fault. I've discussed it with upstream, and qGo continues segfaulting even after I remove scim (which is not the case with scim related bugs).
There is also a misconception about scim (I was also mistaken when my system became instable upgrading from hoary to breezy) : the "bugs" where not scim's fault, but libstdc++ ABI change between version 5 and 6 (AFAIK). BTW, qGo doesn't compile against any libstdc++.

According to upstream, it looks like a bug in qGo's code :
- the function MyMessageHandler (need a look at the call stack)
- a variable (need to run the program step by step and put stop points to see which variable is badly initialised and make everything crash - according to upstream it might be due to the *msg chain)

At the moment, qGo's code would need to be looked at by somebody who nows a bit of programming (I don't). However, upstream is just one man and he doesn't use Ubuntu nor Debian, and can't reproduce the bug.

Revision history for this message
Ming Hua (minghua) wrote :

From the reporter's comments, it seems scim is not related to this bug. Rejecting on the scim side.

Changed in scim:
status: Unconfirmed → Rejected
Revision history for this message
atie (atie-at-matrix) wrote :

I can run Dapper's qgo 1.0.4-r2-1 without any problem so closing this bug as Fix Released.

Loïc, please upgrade to the most recent version, and open a new bug if the problem persists.

Changed in qgo:
status: Needs Info → Fix Released
Revision history for this message
Loïc Martin (loic-martin3) wrote :

Thanks, I already have qgo working in Dapper. However, since scim configuration is not complete, I can't really say if the pb won't reappear when (hopefully) scim default configuration will allow me to use it to input Chinese in Open Office under a non-Chinese session.
Once scim default configuration allows that, I'll add a comment here to update the situation of this bug. Atm, I don't really know if it's not going to be there in Dapper final (the bug seemed more related to Ubuntu locale configuration than scim program, and that is still undergoing changes in Dapper).

Revision history for this message
Loïc Martin (loic-martin3) wrote :

Tested using Ubuntu 6.06 Dapper Drake RC in both hard disk install and live cd : with a default english system (just booting the Live CD, then installing qgo) qgo segfaults at startup
This has been confirmed by another user and is reproductible : see Bug #46271

Marked as critical because the program don't work at all (segfault at startup)

Changed in qgo:
status: Fix Released → Confirmed
description: updated
Revision history for this message
Daniel Robitaille (robitaille) wrote :

Lowering severity. A critical bug is one that "has a severe impact on a large portion of Ubuntu users" (https://wiki.ubuntu.com/Bugs/CommonTasks). qgo is not widely in use in the Ubuntu users base. It is still an annoying bug that should be fixed, but it is not a critical one.

Revision history for this message
Loïc Martin (loic-martin3) wrote :

I've made a wiki page related to the bug that could be usefull, especially for non-ubuntu programmers :
https://wiki.ubuntu.com/QgoSegmentationFault

The Live CD offers gdb for debugging purposes :)

Revision history for this message
Loïc Martin (loic-martin3) wrote :

Sorry for the severity, I thought it was only related to the program (not to the distribution). https://wiki.ubuntu.com/Bugs/CommonTasks has proven an interesting read, thank you.

Revision history for this message
Loïc Martin (loic-martin3) wrote :

Qgo's developper (Emmanuel Beranger) just fixed the bug upstream, after having been able to confirm it/fix it using Dapper's Live cd.

He's asking if qgo package maintainer could help me test the CVS version. If it works, the bug fix should be in qgo's next version.

Thanks

Revision history for this message
Mazirian (bowman-mazirian) wrote :

Has there been amy progress on this. The package in dappers univers repo still suffers from the esgfault at startup problem.

Revision history for this message
Loïc Martin (loic-martin3) wrote :

I'll have no internet starting from tomorrow. If the packages maintainer reads this, here is a summary :

Ubuntu qgo packages are useless at the moment (they segfault on Dapper, Ubuntu or Xubuntu - and it should be the same for Kubuntu). They might (according to EB) even segfault on Debian.

So there's no point compiling them

qgo's last version as for today (the one in Edgy), that is 1.5r1 or 2, still has the bug (so it's useless).

The bug is fixed in the CVS version, so there's 2 possibilities:
-either hope qgo releases a new version before Edgy freezes, or get an UVF (highly unlikely for something like a go program :)
or
- get the fix into Ubuntu's 1.5 packages.

According to qgo's developper, the fix is really small (uninitialased variable, only happens on Debian systems)

Cheers, and thanks for the one that will address the problem!

Revision history for this message
StuartYeates (stuart-yeates) wrote :

I've attempted to compile the CVS version to confirm whether the fix works, but failed because of autoconf errors. The current error message is:

cd . && /bin/sh /tmp/qgo/qgo/admin/missing --run aclocal-1.6
/tmp/qgo/qgo/admin/missing: line 46: aclocal-1.6: command not found
WARNING: `aclocal-1.6' is needed, and you do not seem to have it handy on your
         system. You might have modified some files without having the
         proper tools for further handling them. Check the `README' file,
         it often tells you about the needed prerequirements for installing
         this package. You may also peek at any GNU archive site, in case
         some other package would contain this missing `aclocal-1.6' program.
make: *** [aclocal.m4] Error 1

but this is almost certainly a symtom raterh than a cause.

Revision history for this message
Loïc Martin (loic-martin3) wrote :

New qgo version fixes the bug (that also appeared in Debian). That version (1.5.1) appeared in July, 13, and is present in Debian but Ubuntu only has a 1.5-r1 version.

If someone could merge Ubuntu's version with Debian one, or tell me how I could request a merge, that would be dreatly appreciated. Thank you in advance.

Revision history for this message
Ming Hua (minghua) wrote :

Sync requested at bug #62101.

Revision history for this message
Ming Hua (minghua) wrote :

qgo 1.5.1-1 has been synced from Debian unstable and been built for all arches now on edgy. Please test the qgo on edgy to make sure this bug is indeed fixed.

If you want this bug also fixed in dapper, please contact the backport team through Launchpad (read the instructions in http://ubuntuforums.org/showthread.php?t=153402) for a dapper backport.

Revision history for this message
Loïc Martin (loic-martin3) wrote :

Tested with Edgy beta : qgo 1.5.1-1 works without segfault. Marking this bug fixed.
Thanks a lot.

Changed in qgo:
status: Confirmed → Fix Released
Revision history for this message
Loïc Martin (loic-martin3) wrote :

For anyone using Dapper : qgo main site offer a dirty compiled package of qgo 1.5.1 for Dapper.

To post a comment you must log in.
This report contains Public information  
Everyone can see this information.

Duplicates of this bug

Other bug subscribers

Remote bug watches

Bug watches keep track of this bug in other bug trackers.