Bug 30155 - libreoffice crashes signal 11
Summary: libreoffice crashes signal 11
Status: RESOLVED WORKSFORME
Alias: None
Product: Mageia
Classification: Unclassified
Component: RPM Packages (show other bugs)
Version: Cauldron
Hardware: All Linux
Priority: Normal critical
Target Milestone: ---
Assignee: Thierry Vignaud
QA Contact:
URL:
Whiteboard:
Keywords:
Depends on:
Blocks:
 
Reported: 2022-03-12 05:44 CET by Pierre Fortin
Modified: 2022-10-27 16:40 CEST (History)
1 user (show)

See Also:
Source RPM: libreoffice-7.3.1.3-1.mga9
CVE:
Status comment:


Attachments
strace oocalc (89.27 KB, text/plain)
2022-03-12 05:45 CET, Pierre Fortin
Details
konsole output from oocalc (5.85 KB, text/plain)
2022-03-12 05:46 CET, Pierre Fortin
Details
konsole output from oowriter (5.86 KB, text/plain)
2022-03-12 05:46 CET, Pierre Fortin
Details
a slightly better strace (last 1000 lines) (115.75 KB, text/plain)
2022-03-12 16:18 CET, Pierre Fortin
Details
crash stack with [physical_address] removed (4.57 KB, text/plain)
2022-05-13 04:48 CEST, Pierre Fortin
Details

Description Pierre Fortin 2022-03-12 05:44:37 CET
Description of problem:  I was making extensive use of oocalc a day or two ago. Since I'm on Cauldron, I apply updates as quickly as possible. Today, oocalc crashes with Signal 11 

Version-Release number of selected component (if applicable):  7.3.1.3


How reproducible: Always


Steps to Reproduce:
1. oocalc &
2. Click Start on Document Recovery dialog (9 documents: 2 ods, 7 csv)
3. Error dialog reports one document does not exist.
4. Click OK
5. Get another dialog containing:
   Due to an error, LibreOffice crashed. All the files you were working on will  now be saved. The next time LibreOffice is launched, you files will be recovered automatically.
The following files will be recovered:
[list of 9 files -- including the one that no longer exists]

oowriter crashes similarly.

Did glibc get updated twice in the past 24-48 hours? Pretty sure I had to reboot twice for glibc...

Tried gdb:
"/usr/bin/oocalc": not in executable format: file format not recognized
(gdb) run
Starting program:  
No executable file specified.
Use the "file" or "exec-file" command.
(gdb) 

So I resorted to strace.
Comment 1 Pierre Fortin 2022-03-12 05:45:47 CET
Created attachment 13169 [details]
strace oocalc
Comment 2 Pierre Fortin 2022-03-12 05:46:25 CET
Created attachment 13170 [details]
konsole output from oocalc
Comment 3 Pierre Fortin 2022-03-12 05:46:58 CET
Created attachment 13171 [details]
konsole output from oowriter
Comment 4 Pierre Fortin 2022-03-12 06:02:05 CET
Argh! Uninstalled 7.3.1.3 with intention to install 7.3.0.3; but now neither package can be selected.  Just applied more updates, including kernel; so will reboot now...
Comment 5 Pierre Fortin 2022-03-12 06:35:26 CET
Rebooted. All up to date.  Still can't select either 7.3.0.3 or 7.3.1.3 to install.  Just my luck; I have a zoom call scheduled in the morning. Maybe I can limp along with my mga8 laptop...  ;p
Pierre Fortin 2022-03-12 15:10:45 CET

Summary: libreoffice crashes signal 11 => libreoffice crashes signal 11 (uninstalled; now can't select 7.3.0.3 or 7.3.1.3 to install)

Comment 6 sturmvogel 2022-03-12 15:25:35 CET
Normally there is an additional information why you can't select a package. Please provide this information.
Comment 7 Pierre Fortin 2022-03-12 15:39:56 CET
restarting mcc "solved" the 'unable to select issue'

Summary: libreoffice crashes signal 11 (uninstalled; now can't select 7.3.0.3 or 7.3.1.3 to install) => libreoffice crashes signal 11

Comment 8 Pierre Fortin 2022-03-12 15:42:30 CET
re-installed 7.3.1.3 and still crashes.  Must be an update in the past 48 hours.
Comment 9 Pierre Fortin 2022-03-12 16:18:55 CET
Created attachment 13173 [details]
a slightly better strace (last 1000 lines)

added -f option to strace
Comment 10 Marja Van Waes 2022-03-12 17:31:15 CET
(In reply to Pierre Fortin from comment #9)
> Created attachment 13173 [details]
> a slightly better strace (last 1000 lines)
> 
> added -f option to strace

Assigning to the LO maintainer, who understand strace outputs, I still need to learn that :-(

I don't manage to reproduce it, but then I didn't have any documents to be recovered from before updating.

If I create one file to be recovered (by starting a new file and then killing LO), recovery of that one file works fine. If I do the same and then rename the file, then recovery doesn't work (of course), I get a message that I don't have access to the file. However, LO doesn't crash and just starts normally.

What happens when you remove the not existing document from the Document Recovery dialog before continuing with the recovery?

Assignee: bugsquad => thierry.vignaud
CC: (none) => marja11
Source RPM: (none) => libreoffice-7.3.1.3-1.mga9

Comment 11 Pierre Fortin 2022-03-13 02:31:17 CET
There's no way to solve the issue via the startup process.  
Looks like LO does not know how to handle recovery files that no longer exist.
I did "touch <missing_file>" and startup tried to recover an empty file and moved on to the next missing file. touch'ed that filename and then I could start LO successfully after that.  I did not see any reference to these missing files in $HOME/.config/libreoffice so not sure where that info is kept.

LO uses several scripts during startup:

/usr/bin/oocalc  simply calls:  exec libreoffice --calc "$@"

/usr/bin/libreoffice  chooses between /usr/lib and /usr/lib64; then calls libreoffice/program/soffice "$@" in the appropriate lib

/usr/lib64/libreoffice/program/soffice "$@"  sets up some debugging and locates the installation directory; then calls /usr/lib64/libreoffice/program/oosplash

Now, I'll have to let the LO folk figure it out from here...  :)
Comment 12 Pierre Fortin 2022-05-11 15:19:34 CEST
Not seeing crashes on LO 7.3.3.2
Closing.

Status: NEW => RESOLVED
Resolution: (none) => WORKSFORME

Comment 13 Pierre Fortin 2022-05-13 04:21:51 CEST
Re-opening:  after system lockup (bug 30405), LO crashes every time I try to open oocalc...
- displays splash quickly
- Document Recovery (16 spreadsheets)
- Start
- one sheet "does not exist"
- OK
- Document Recovery: "Due to an error, LibreOffice crashed. All the files you were working on will now be saved. The next time LibreOffice is launched, your files will be recovered automatically.
The following files will be recovered:"
  - same 16 files including the non-existent one.
- OK
- crash 

LO does not remove the missing file from its list, so the crashes are identical.

Status: RESOLVED => UNCONFIRMED
Ever confirmed: 1 => 0
Resolution: WORKSFORME => (none)

Comment 14 Pierre Fortin 2022-05-13 04:39:16 CEST
Re-opening:  after system lockup (bug 30405), LO crashes every time I try to open oocalc...
- displays splash quickly
- Document Recovery (16 spreadsheets)
- Start
- one sheet "does not exist"
- OK
- Document Recovery: "Due to an error, LibreOffice crashed. All the files you were working on will now be saved. The next time LibreOffice is launched, your files will be recovered automatically.
The following files will be recovered:"
  - same 16 files including the non-existent one.
- OK

If I remove the physical memory address (in brackets) at the end of each record in the crash stack, the crashes are identical.
Comment 15 Pierre Fortin 2022-05-13 04:48:36 CEST
Created attachment 13247 [details]
crash stack with [physical_address] removed

LOL  Priceless...  rather than fix a bug; LO documented it:

--norestore
              Disables restart and file recovery after a system crash. It is possible that LO will try to restore a file it keeps crashing on, if that happens --norestore is the only way to start LO.

At least, the File->Recent Documents list is intact...
Comment 17 Pierre Fortin 2022-05-15 19:23:58 CEST
Posted a really crappy workaround at https://bugs.documentfoundation.org/show_bug.cgi?id=148431#c2
Comment 18 Pierre Fortin 2022-10-27 16:40:53 CEST
Closing. Not seen for months.
Now using both current and 7.5.0 and reporting directly to documentfoundation.

Resolution: (none) => WORKSFORME
Status: UNCONFIRMED => RESOLVED


Note You need to log in before you can comment on or make changes to this bug.