Jump to content

Photo

[Solved] Moksha Segfault


Best Answer Jeff, 27 January 2017 - 06:39 PM

Out of curiosity - does disabling the window geometry text in Window Display settings make the crashes stop for you?

oNKMVxs.png

I can't produce a lock on my system since disabling these. Going to look into fixing it, but for now this seems like a good work around.

Go to the full post


  • Please log in to reply
69 replies to this topic

#1 Astroboy

Astroboy

    Member

  • Members
  • 339 posts
  • LocationZacatecas, Mexico

Posted 12 December 2016 - 04:26 PM

I would suggest to look closely at the thread http://forums.bodhil...sh-and-restart/

 

Since the start of that thread we already installed Escuelas Linux 5.0 (based in Bodhi 4.0) in a fairly good number of computers, and we can confirm that in some of them Moksha ocasionally restarts without any apparent cause.

 

In a computer I own, this issue is almost always reproducible just by dragging some window. I said "almost" because sometimes I can drag a window without having Moksha restarting.





A big thank you to everyone who contributes to Bodhi Linux


#2 The waiter

The waiter

    Module Master

  • Developer
  • 1659 posts
  • LocationBanska Bystrica, Slovakia

Posted 12 December 2016 - 05:36 PM

Astroboy: did you backup /home folder and then return it back, or did you do a fresh install with empty /home? Some problems here emerged because of this approach.



#3 Astroboy

Astroboy

    Member

  • Members
  • 339 posts
  • LocationZacatecas, Mexico

Posted 12 December 2016 - 06:13 PM

Oh, we always do some sort of backup. As part of the Escuelas Linux installation, we provide a user account with all the configuration files for our included programs already set up, in order to provide our final users with an environment ready to be used at its max potential for its intended use in education or schools. As such, yes, we do provide since some years ago a custom E/Moksha configuration, included in the .e/ directory.

 

If including a custom .e/ directory for the user account is the source of the restart graphical issue in Bodhi 4.0, is there a way to prevent it without having to withdraw that custom configuration? In the thread we pointed previously was recommended to erase all profile files but the number one in .e/e/config, and we found that it alleviates the issue, but it shortly reappears as the profile files populate automatically that directory...
 



#4 The waiter

The waiter

    Module Master

  • Developer
  • 1659 posts
  • LocationBanska Bystrica, Slovakia

Posted 12 December 2016 - 06:20 PM

Hmm, understand. I would pick up one or two PCs where the issue is noticeable, and in the first phase erase whole .e/ folder. If does not help, continue with the phase two: erase all config data in /home. Just an idea. Perhaps worth trying.



#5 Astroboy

Astroboy

    Member

  • Members
  • 339 posts
  • LocationZacatecas, Mexico

Posted 12 December 2016 - 08:07 PM

Ok. Will do that test this evening :-)



#6 Jeff

Jeff

    Lead Developer

  • Developer
  • 12524 posts
  • LocationBloomington, IL

Posted 12 December 2016 - 10:21 PM

Yea - I use Moksha probably 6+ hours daily and don't see regular segfaults. Anything we can't reproduce on command is near impossible to chase down. We will be testing newer EFL - hopefully it is an underlying issue in that which gets resolved in this update or just a configuration issue like the waiter indicates. Start with a fresh / default config and let us know if your crashes happen under that. 



#7 Astroboy

Astroboy

    Member

  • Members
  • 339 posts
  • LocationZacatecas, Mexico

Posted 13 December 2016 - 12:33 AM

Ok, these are the results of the test in a notebook where is fairly reproducible the issue described: the infamous Dell M3800 -yes, the same one I talked about a year ago in http://forums.bodhil...upport/?p=94997, although we had observed the Moksha restart issue sporadically on some other computers (desktop and notebooks).

 

Background

The .e/ directory in all the computers is the custom configuration included in Escuelas Linux 5.0. That custom configuration was created from scratch in the previous 3.2.0 Bodhi version, in which there were no Moksha restart issues in the few computers were we installed it as a pilot, with the idea to deploy it as default when Bodhi 4.0 were released (which we actually did in Escuelas Linux 5.0).

 

First test

Using the default .e/ included in Escuelas Linux on the M3800. Dragging around  Terminology and PcmanFM. In less than a minute, the expected behavior: Moksha restarted.

 

Second test

Closing session. Press CTRL-ALT-F1 to open a Terminal, and from that place, deleted .e/ directory. Exit terminal and ALT-F7 to return to Login Screen, access the system. The default Moksha Desktop appears.

Open PCManFm and dragged it around the screen and... to my disappointment, Mosksha restarts... I was expecting that in the default Moksha configuration this would not happen :-(

 

So, the same behavior was observed with our custom .e/ config, and with the freshly created .e/ Moksha default config.

 

Some more info just in case it could be useful: the notebook has installed the nvidia-367 package, but the tests were done using its Intel GPU. I currently use the Intel GPU not only for better battery duration, but because I had a bad experience using the Nvidia GPU using the previous Escuelas Linux 4.5 (based in Bodhi 3.2). In a formal conference I had active the Nvidia one, and found to my dismay that GPU does not work well with an HDMI to VGA adapter: the screen in the laptop showed a slide, while in the projection didn't. I had to turn back and forth on each slide to refresh the external display, and when opening apps, with the Nvidia GPU I had to drag around the app window in order to be make it visible on the projector. Aside from the embarrassment, now I stick with the Intel GPU, which does not have that behavior...

 

lspci

00:02.0 VGA compatible controller: Intel Corporation 4th Gen Core Processor Integrated Graphics Controller (rev 06)

02:00.0 3D controller: NVIDIA Corporation GK107GLM [Quadro K1100M] (rev ff)


 



#8 Jeff

Jeff

    Lead Developer

  • Developer
  • 12524 posts
  • LocationBloomington, IL

Posted 13 December 2016 - 07:14 AM

Is this only happening on one hardware configuration? Does it only happen with particular themes? I have four different computers I use several times a week with Bodhi 4 and I don't see segfaults when moving windows around the screen.

 

In fact - just spent 2~ minutes dragging windows around on my main desktop that has dual screens with no segfaults.

Have you tried the EFL 1.18.3 packages in the testing branch yet? I am currently using that on all my systems, but I also don't recall seeing crashes prior to this update.



#9 Astroboy

Astroboy

    Member

  • Members
  • 339 posts
  • LocationZacatecas, Mexico

Posted 13 December 2016 - 01:21 PM

It does not happen at all in some computers. It happens sporadically in some other computers (say, once in a week) without any apparent reason, such as when using a web browser or typing in LIbreOffice, which means that it does not happen exclusively when dragging windows around.

 

In the test computer does not mean that it always happens when dragging windows around, sometimes I can move windows without any issue. For this test I had to provoke the Moksha restart, that is, moving a window until the segfault appears (much as like bothering a dog to make it bark).

 

It is not a deal breaker situation, for the users running some app in full screen it looks to them like a brief visual glitch in which the title bar dissappear and reappear without any consequence, no lost work. But of course would be better to not have that glitch.

 

Only used the EFL packages as they came included in Bodhi 4.0 ISO.



#10 Jeff

Jeff

    Lead Developer

  • Developer
  • 12524 posts
  • LocationBloomington, IL

Posted 13 December 2016 - 03:59 PM

I see occasional segfaults here as well, but nothing I can reproduce on command so essentially impossible to track down.

Thinking on things that are different from Bodhi 3 - are you using the clipboard module in your profiles? If so maybe try unloading this and see if the segfaults go away. Another possible culprit are our changes to the screenshot module (though those are fairly minimal).



#11 The waiter

The waiter

    Module Master

  • Developer
  • 1659 posts
  • LocationBanska Bystrica, Slovakia

Posted 13 December 2016 - 04:49 PM

I think unloading those modules was already tested in another thread and did not help. Anyway I am for testing it again.

#12 The waiter

The waiter

    Module Master

  • Developer
  • 1659 posts
  • LocationBanska Bystrica, Slovakia

Posted 13 December 2016 - 07:02 PM

I have read some threads about Ubuntu 16.04 freezing and restarts and some advices pointed to changing kernel related to GPUs. BL 4 uses kernel 4.4 as default. BL 3.2.1 the 4.2 version. Astroboy are you helpful installing the older version in order to make some progress in revealing this issue?



#13 The waiter

The waiter

    Module Master

  • Developer
  • 1659 posts
  • LocationBanska Bystrica, Slovakia

Posted 13 December 2016 - 07:52 PM

Hmm. one more idea which came on my mid is using Xephyr. It is a debugging tool which could help us...

 

open terminal emulator and type:

 

Xephyr -screen 1024x768 -br :1

 

 and in another terminal or tab:

 

DISPLAY=:1 enlightenment_start

 

Another moksha window will appear. Try to do all actions which lead to Moksha restart and roll back in message window to catch what happened before restart. Go back in time to 00:00. Post the screenshot here or better all content. Thanx

 

Unfortunately I have no restarts here to try it on my PC. I had them, but they dissapeared after some days. Because of some dist-upgrades? Who knows...



#14 The waiter

The waiter

    Module Master

  • Developer
  • 1659 posts
  • LocationBanska Bystrica, Slovakia

Posted 13 December 2016 - 09:47 PM

Last post today. 

Enlightnement generates crash report which is placed here:

 

~HOME/.e-crashdump.txt

 

Please let me know if this file does exist and what is inside (thanx okra for info).



#15 Astroboy

Astroboy

    Member

  • Members
  • 339 posts
  • LocationZacatecas, Mexico

Posted 14 December 2016 - 04:37 PM

These are the results, since the start of Xephyr until the Moksha restart (which was fairly quick, around 5 seconds after this session started. No clipboard module loaded.

Spoiler


#16 Astroboy

Astroboy

    Member

  • Members
  • 339 posts
  • LocationZacatecas, Mexico

Posted 14 December 2016 - 05:00 PM

I also installed kernel 4.2.8 (the last iteration of the 4.2 series). Same issue, the Moksha restart appears in this kernel version.



#17 Jeff

Jeff

    Lead Developer

  • Developer
  • 12524 posts
  • LocationBloomington, IL

Posted 14 December 2016 - 05:17 PM

Need a bit more information if you can help. 
 

Open up a terminal and run:
Xephyr :1
 
 
Then open a second terminal and run:
gdb enlightenment
 
At the gdb prompt type this command:
b src/lib/eo/eo_ptr_indirection.c: 47
 
then
 
run -display :1
 
Now trigger the crash. When the crash trigger Moksha should stop in Xephyr.
 
In the gdb terminal run:
 
bt 
 
And paste me the backtrace.


#18 Astroboy

Astroboy

    Member

  • Members
  • 339 posts
  • LocationZacatecas, Mexico

Posted 14 December 2016 - 07:26 PM

Here are the results:
Spoiler


#19 Jeff

Jeff

    Lead Developer

  • Developer
  • 12524 posts
  • LocationBloomington, IL

Posted 14 December 2016 - 09:03 PM

You missed running "bt" after the crash it looks like.



#20 Astroboy

Astroboy

    Member

  • Members
  • 339 posts
  • LocationZacatecas, Mexico

Posted 14 December 2016 - 10:02 PM

Oooops!

(gdb) bt
#0  _int_malloc (av=av@entry=0x7ffff414fb20 <main_arena>,
    bytes=bytes@entry=544) at malloc.c:3414
#1  0x00007ffff3e1021a in __libc_calloc (n=<optimized out>,
    elem_size=<optimized out>) at malloc.c:3234
#2  0x00007ffff6402af0 in ?? () from /usr/lib/x86_64-linux-gnu/libevas.so.1
#3  0x00007ffff6397d1a in ?? () from /usr/lib/x86_64-linux-gnu/libevas.so.1
#4  0x00007ffff6398e20 in evas_cache_image_data ()
   from /usr/lib/x86_64-linux-gnu/libevas.so.1
#5  0x00007fffe9813f84 in ?? ()
   from /usr/lib/x86_64-linux-gnu/evas/modules/engines/software_x11/v-1.18/module.so
#6  0x00007ffff6443c49 in ?? () from /usr/lib/x86_64-linux-gnu/libevas.so.1
#7  0x00007ffff63778fa in ?? () from /usr/lib/x86_64-linux-gnu/libevas.so.1
#8  0x00007ffff6379c87 in ?? () from /usr/lib/x86_64-linux-gnu/libevas.so.1
#9  0x00007ffff62f01ad in evas_canvas_render_async ()
   from /usr/lib/x86_64-linux-gnu/libevas.so.1
#10 0x00007fffe9a29357 in ?? ()
   from /usr/lib/x86_64-linux-gnu/ecore_evas/engines/x/v-1.18/module.so
#11 0x00007ffff5e43d86 in ?? ()
   from /usr/lib/x86_64-linux-gnu/libecore_evas.so.1
#12 0x00007ffff606941f in ?? () from /usr/lib/x86_64-linux-gnu/libecore.so.1
#13 0x00007ffff175563b in ?? () from /usr/lib/x86_64-linux-gnu/libeo.so.1
#14 0x00007ffff1754313 in eo_event_callback_call ()
---Type <return> to continue, or q <return> to quit---
   from /usr/lib/x86_64-linux-gnu/libeo.so.1
#15 0x00007ffff606bd9b in ?? () from /usr/lib/x86_64-linux-gnu/libecore.so.1
#16 0x00007ffff606c387 in ecore_main_loop_begin ()
   from /usr/lib/x86_64-linux-gnu/libecore.so.1
#17 0x0000000000434759 in main ()

 






0 user(s) are reading this topic

0 members, 0 guests, 0 anonymous users