[SOLVED] libreoffice fails to start on a -current clean install

Alien Bob · 12-12-2020, 10:31 AM

People with "application error" can you check /etc/profile.d/libreoffice.sh and tell me if there's any line that is UN-commented? Here on my Slackware-current running Plasma5, I do not have any line un-commented.

_peter · 12-12-2020, 12:58 PM

here it seems there are no lines that are UN-commented.

Quote:

$more /etc/profile.d/libreoffice.sh
#!/bin/sh

# To force the use of a certain VCL UI interface, use one of these envvars.
#export SAL_USE_VCLPLUGIN=gen
#export SAL_USE_VCLPLUGIN=gtk3
#export SAL_USE_VCLPLUGIN=gtk3_kde5
#export SAL_USE_VCLPLUGIN=kf5
#export SAL_USE_VCLPLUGIN=qt5

sairum · 12-13-2020, 04:16 AM

Nope

Code:

bash-5.1$ cat /etc/profile.d/libreoffice.sh
#!/bin/sh

# To force the use of a certain VCL UI interface, use one of these envvars.
#export SAL_USE_VCLPLUGIN=gen
#export SAL_USE_VCLPLUGIN=gtk3
#export SAL_USE_VCLPLUGIN=gtk3_kde5
#export SAL_USE_VCLPLUGIN=kf5
#export SAL_USE_VCLPLUGIN=qt5
bash-5.1$

sairum · 12-13-2020, 04:22 AM

_peter I assume that you are experiencing the exact same problema as I am. Can you run libreoffice through gdb just to check why it is failing?

Code:

gdb /usr/lib64/libreoffice/program/soffice.bin

Just type 'run' when prompted

Code:

...
For help, type "help".
Type "apropos word" to search for commands related to "word"...
Reading symbols from /usr/lib64/libreoffice/program/soffice.bin...
(No debugging symbols found in /usr/lib64/libreoffice/program/soffice.bin)
(gdb) run

_peter · 12-13-2020, 05:19 AM

ok sairum,

Code:

gdb /usr/lib64/libreoffice/program/soffice.bin
GNU gdb (GDB) 10.1
Copyright (C) 2020 Free Software Foundation, Inc.
License GPLv3+: GNU GPL version 3 or later <http://gnu.org/licenses/gpl.html>
This is free software: you are free to change and redistribute it.
There is NO WARRANTY, to the extent permitted by law.
Type "show copying" and "show warranty" for details.
This GDB was configured as "x86_64-slackware-linux".
Type "show configuration" for configuration details.
For bug reporting instructions, please see:
<https://www.gnu.org/software/gdb/bugs/>.
Find the GDB manual and other documentation resources online at:
    <http://www.gnu.org/software/gdb/documentation/>.

For help, type "help".
Type "apropos word" to search for commands related to "word"...
Reading symbols from /usr/lib64/libreoffice/program/soffice.bin...
(No debugging symbols found in /usr/lib64/libreoffice/program/soffice.bin)
(gdb) run
Starting program: /usr/lib64/libreoffice/program/soffice.bin 
[Thread debugging using libthread_db enabled]
Using host libthread_db library "/lib64/libthread_db.so.1".
[Detaching after fork from child process 25501]
[New Thread 0x7fffeacd4700 (LWP 25506)]
[New Thread 0x7fffea4d3700 (LWP 25507)]
[New Thread 0x7fffe99b5700 (LWP 25508)]
[New Thread 0x7fffe91b4700 (LWP 25509)]
[New Thread 0x7fffdbfff700 (LWP 25510)]
[New Thread 0x7fffdb7fe700 (LWP 25511)]
[Thread 0x7fffdb7fe700 (LWP 25511) exited]
[Thread 0x7fffeacd4700 (LWP 25506) exited]
[New Thread 0x7fffeacd4700 (LWP 25515)]
[New Thread 0x7fffdb7fe700 (LWP 25516)]
[Thread 0x7fffdb7fe700 (LWP 25516) exited]
[New Thread 0x7fffdb7fe700 (LWP 25517)]
[Thread 0x7fffdb7fe700 (LWP 25517) exited]
[Thread 0x7fffeacd4700 (LWP 25515) exited]

Thread 1 "soffice.bin" received signal SIGSEGV, Segmentation fault.
0x00007fffc66f1ec9 in ?? () from /usr/lib64/gallium-pipe/pipe_iris.so
(gdb)

sairum · 12-13-2020, 05:47 AM

Quote:

Originally Posted by _peter

ok sairum,

Code:

gdb /usr/lib64/libreoffice/program/soffice.bin
[...]
[Thread 0x7fffeacd4700 (LWP 25515) exited]

Thread 1 "soffice.bin" received signal SIGSEGV, Segmentation fault.
0x00007fffc66f1ec9 in ?? () from /usr/lib64/gallium-pipe/pipe_iris.so
(gdb)

So you do have exactly the same problem! What's your GPU (graphics card). Mine is an intel HD Graphics 630

Code:

bash-5.1# lspci 
[...]
00:02.0 VGA compatible controller: Intel Corporation HD Graphics 630 (rev 04)

Because it's a weekend I don't have access to the machine (it's switched off). But tomorrow I may try to recompile mesa (which is where the pipe_iris.so lives) or downgrade it to a previous package to see what happens.

_peter · 12-13-2020, 06:14 AM

Ha ! a corner case then, thanks.

here :

Code:

00:02.0 VGA compatible controller: Intel Corporation HD Graphics 5500 (rev 09)

sairum · 12-14-2020, 07:21 AM

As I suspected, I can now confirm that downgrading mesa to version 20.2.3 (found on https://slackware.uk/cumulative/slac...3-x86_64-1.txz) brings libreoffice alive again! Marking this SOLVED again, for now...

pchristy · 12-14-2020, 09:44 AM

Very strange! I'm running mesa-20.3.0 on three machines (two Intel, one AMD gfx) and LibreOffice-7.0.3 works fine on all of them (since Eric's last update)!

--
Pete

sairum · 12-14-2020, 10:25 AM

Quote:

Originally Posted by pchristy

Very strange! I'm running mesa-20.3.0 on three machines (two Intel, one AMD gfx) and LibreOffice-7.0.3 works fine on all of them (since Eric's last update)!

--
Pete

Strange it is. I also have many other machines all running slackware -current without issues. My main workstation is now at home due to COVID19 and it has an i5 4300 with a strange xeon intel graphics card. Works like a charm! The problem only reveals itself on this particular machine at may office, which has an intel UHD 630 (built into an i5 7400). I suspect that _peter, who has an intel 5500, has the exact same problem. Let's see if he downgrades mesa and solves the problem for now! Meanwhile I'll do some more research in my spare time. This issues really make me mad (I mean, failing software that leaves no trace of why it fails).

NOTE: rebuilding mesa-20.3.0 using Slackbuild in source also didn't solve the problem...

_peter · 12-14-2020, 02:40 PM

ok thanks sairum, for verification sake, same behavior here libreoffice starts after reverting to mesa-20.2.3 from mesa-20.3.0.

pchristy · 12-15-2020, 06:19 AM

OK, so this is getting even stranger! My main desktop machine is an i7 on a Gigabyte motherboard with Intel integrated graphics: Intel Xeon E3-1200 rev06. Libreoffice works perfectly.
My laptop is an Intel i5, again with Intel integrated graphics: Intel UHD 620 rev 30. Libreoffice works perfectly.

Most interestingly, from your perspective, is an old Dell laptop that I was given because it has a faulty connection between the motherboard and LCD (I would replace the cable but some of the captive bolts are no longer captive, meaning I can't dismantle it!). Usually a wiggle of the lid fixes it for a while.

The Dell is an i5, Intel integrated graphics: Intel HD 5500 rev 09 - which is the one I think is giving you problems! I've put a clean install of 64-current on it, and Libreoffice works perfectly!

This probably isn't what you wanted to hear, but sometimes even a negative result can be useful.

Let me know if there's anything you want me to check on it, and I'll see if I can help.

--
Pete

sairum · 12-15-2020, 06:57 AM

Thanks for the info pchristy!

In my case (OP), the "faulty" GPU is an Intel Corporation HD Graphics 630 (rev 04), not the 5500 (that's from _peter). For your information, I've tried everything, from a standard -current install with multilib (this is the "norm" for me), to a clean -current install without any extras, "clean" (new) account, deleting libreoffice config folder... in all cases libreoffice fails with the "Application error" message and nothing more. Using gdb I traced the problem to pipe_iris.so. Again, I don't need libreoffice for what I normally do (although I like to have it around) but it bothers me when something breaks in slackware and I can't understand why. Right now, in this particular machine, downgrading mesa to version 20.2 was a fix. I do have lots of other machines (more than 10!) where this problem does not occur, including some like yours with an Intel Xeon E3-1200 or an UHD 620. In all them AlienBob's libreoffice is working flawlessly.

Yesterday I tried to recompile mesa from stock -current and it didn't help. I will now try to recompile mesa with some debugging turned on to understand where exactly it fails. The gdb's message is itself not very informative as it cannot signal the exact place where the segmentation violation occurred (??).

Code:

Thread 1 "soffice.bin" received signal SIGSEGV, Segmentation fault.
0x00007fffa8cccec9 in ?? () from /usr/lib64/gallium-pipe/pipe_iris.so
(gdb)

Maybe it's a dying CPU/GPU?

mumahendras3 · 12-17-2020, 03:58 AM

It seems I'm also got hit by this problem...

Code:

gdb /usr/lib64/libreoffice/program/soffice.bin

Code:

Thread 1 "soffice.bin" received signal SIGSEGV, Segmentation fault.
0x00007fffad871039 in ?? () from /usr/lib64/gallium-pipe/pipe_iris.so

My graphics info:

Code:

00:02.0 VGA compatible controller: Intel Corporation HD Graphics 530 (rev 06)

I'm going to try reverting to mesa 20.2.3 and see if that works.

EDIT: Yep, reverting to mesa 20.2.3 fixed it for me. I'll try upgrading to version 20.2.6 and see if there are any problems.
EDIT2: Upgrading to mesa 20.2.6 still works fine for me.

sairum · 12-17-2020, 05:23 AM

Can you try the new mesa-20.3.1-x86_64-1.txz that got out yesterday? Today, I don't have access to my machine with the intel 630, but I'm curious if this has not been fixed in this new version of mesa. There is a suspicious bug fixed with the commit string "Storing pointer to temporary value inside the Iris driver". Check it if you can!