ssindex crashes (no crash reports needed anymore)

[Jauco]

ssindex crashes upon indexing the attached csv file

[Jauco (spam-jauco)]

crash report

[Jauco (spam-jauco)]

The file that made it crash

[Emmanuel Touzery (emmanuel-touzery)]

i also got a crash of ssindex, but i was doing something completely independant on the computer, so i'm 99% certain that beagle called it on one CSV file on my HD while indexing, and then it crashed.

i'll attach the crash report, it's probably the same bug than the original reporter found.

otherwise, the fun thing: in my case the .csv file was 0 bytes.. so ssindex crashes on empty files.

[Emmanuel Touzery (emmanuel-touzery)]

to crash ssindex, run
ssindex -i <empty file.csv>

[Eric van der Vlist (vdv)]

Another auto generated crash report for ssindex while indexing a csv file.

[Eric van der Vlist (vdv)]

Hmmm...

I have found these lines in kern.log:

Oct 9 08:40:41 localhost kernel: [17212488.128000] Out of Memory: Kill process 7889 (ssindex) score 68668 and children.
Oct 9 08:40:41 localhost kernel: [17212488.128000] Out of memory: Killed process 7889 (ssindex).

I my case, the crash seems to have been caused by an hibernation which resume hasn't worked, causing the system to start without and swap and, in turn, to kill the ssindex process for lack of memory.

Sorry for the noise.

Eric

[edschofield (schofield)]

I can confirm this bug. I have 1.5GB of memory free and I've just seen a crash message from ssindex caused by beagled. My kern.log shows nothing untoward.

Crash report attached. The file comm_jar_includes.csv being indexed is from the Thunderbird source code tree. I'll attached this file too below.

[edschofield (schofield)]

Here is the file comm_jar_includes.csv that causes ssindex to crash. It's a plain text file, not comma separated data.

Running:

user@host$ ssindex -i comm_jar_includes.csv

gives:

Reading file:///home/schofield/mozilla-thunderbird-1.5.0.7/debian/comm_jar_includes.csv
Segmentation fault (core dumped)

[Emmanuel Touzery (emmanuel-touzery)]

i've reported "my" crash (crash on empty file) upstream ( http://bugzilla.gnome.org/show_bug.cgi?id=361391 ), and linked the bug in this bugreport. but even if/when "my" crash will be fixed upstream, it doesn't mean the others are fixed too.

[Emmanuel Touzery (emmanuel-touzery)]

see the comment from upstream. the crash with empty CSV file was fixed in Gnumeric 1.7.1.

for the other crashes, it should be checked as well whether 1.7.1 fixes them.

so at least part of this bug will be fixed by upgrading the gnumeric in ubuntu from 1.7.0 to 1.7.1.

[Emmanuel Touzery (emmanuel-touzery)]

(the comment from upstream that I talk about is http://bugzilla.gnome.org/show_bug.cgi?id=361391#c3 )

[Emmanuel Touzery (emmanuel-touzery)]

another thing: in the upstream bug report, the developer:

1. explained that a whole class of such problems were fixed in 1.7.1, so hopefully all the crashes found in this bug were fixed

2. gives a small (15 lines or so) patch to fix the issue in 1.7.0 if it was too late now to move ubuntu to 1.7.1

so i think now ubuntu should either go to 1.7.1, or if the freeze is too deep and it's too late, the small patch suggested and adopted by upstream which fixes the issue should be applied to ubuntu's gnumeric package.

[Jani Monoses (jani)]

It would be helpful if you could check or coordinate with upstream to confirm whether other gnuneric busg in ubuntu would be fixed by this. A strong argument is needed in favor of upgrading this late to 1.7.1, and if the upgrade fixed many of the current LP bugs that would be one such argument.

[edschofield (schofield)]

There's a whole raft of bug fixes in 1.7.1:

http://www.gnome.org/projects/gnumeric/announcements/1.7/gnumeric-1.7.1.html

I've emailed mdz and cjwatson to request an exception to the upstream version freeze.

[edschofield (schofield)]

Sorry Jani, didn't see your comment before. I can't see any other Edgy-specific bug reports in LP that are fixed by 1.7.1. But 1.7.1 does fix a long list of bugs reported in bugzilla.gnome.org, as the release notes describe.

If we keep 1.7.0 in for Edgy, it's only a matter of time before more of those bugs get re-discovered by Edgy users. Then the fixes that 1.7.1 comprises will need to be applied individually, in response to specific reports about bugs that we know exist, wasting the time of users and developers.

The only packages that depend on gnumeric are

      gnumeric-doc
      gnumeric-plugins-extra
      beagle

so the risk of breakage seems very limited. :)

[edschofield (schofield)]

Matt Zimmerman has said that he has already discussed a potential upgrade to 1.7.1 privately with Gauvain Pocentek and Jani Monoses and decided against it because it would require updates to libraries used by other packages (presumably libgsf 1.14.2 etc., described in http://ftp.gnome.org/pub/GNOME/sources/gnumeric/1.7/gnumeric-1.7.1.changes)

[Emmanuel Touzery (emmanuel-touzery)]

OK, if we can't get 1.7.1 then it would be great to apply that small patch suggested by the developer.

1. the crash will be experienced by anybody having installed gnumeric and having .csv on his hard disk which gnumeric does not like
2. the patch is very small
3. the patch is approved by upstream
4. it's easy to check whether the patch does fix the crash: we have in this bugreport all the files which triggered crashes so far. run the files through 1.7.0 with the patch and check that they indeed don't crash gnumeric

The problem is that apport makes this problem very visible: you are working normally on your computer, and suddendly you get a window that "ssindex", a program you never heard of, crashed. I'm sure many people will think "VIRUS!!!", plus I don't know how this affects beagle indexing.

[Jani Monoses (jani)]

I have not checked but is it certain that 1.7.1 requires the latest gsf? If it were the single updated package it would be less of a risk. Anyway I agree we should get this patch in,
and maybe others that are reported in LP or upstream considers them critical, like other crashers or data loss ones.

[Danny Haak (d-j-t-haak)]

Another crash report.

[Markus Kienast (elias1884)]

mine certainly crashed on beagle calling it. I let the machine run at night with nothing running to make it idle for beagle to index my hdd.

[Andrew Frank (frank-geoinfo)]

i am somewhat surprised that 4 month later this problem has not found a fix. i have on a regularly updated machine gnumeric 1.7.0
and it crashes regularly when beagle does an index over night.

andrew

[Jauco (spam-jauco)]

The bug has been fixed, but the new version not uploaded to the apt
repositories if I'm not mistaken.

On 2/15/07, Andrew Frank <email address hidden> wrote:
>
> i am somewhat surprised that 4 month later this problem has not found a
> fix. i have on a regularly updated machine gnumeric 1.7.0
> and it crashes regularly when beagle does an index over night.
>
> andrew
>
> --
> ssindex crashes (no crash reports needed anymore)
> https://launchpad.net/bugs/63297
>

--
groeten,
     Jauco Noordzij

[William Grant (wgrant)]

The fix has been in Feisty for a while now.