alephone (Marathon) segmentation fault

sleve_mcdichael

It doesn't happen until I quit, so it doesn't seem to affect playability any. Saving still works, as far as I can tell (even after a reboot).

Segfault occurs both from runcommand (in log) and from console:

/opt/retropie/ports/alephone/bin/alephone "/path/to/folder/Marathon/"

Does this happen for everyone else? Or if not, how to troubleshoot? I'm on Pi4, RetroPie 4.8, alephone updated from binary (Build Date: 2022-03-20T18:52:18-07:00).

Edit: it happens for all three Marathon games.

mitu

I wonder if it's not a game data version issue. The module installs the 2015 data files, while the program version is still updated to the current git.
Can you replace the 2015 versions with 20220115 in these lines and re-install the game ?

sleve_mcdichael

@mitu said in alephone (Marathon) segmentation fault:

Can you replace the 2015 versions with 20220115 in these lines and re-install the game ?

Still getting segfault with 20220115 data (I deleted my roms/ports/alephone directory first so it would download fresh, and everything in configs/ports/alephone also):

Parameters: 
Executing: /opt/retropie/ports/alephone/bin/alephone "/home/pi/RetroPie/roms/ports/alephone/Marathon/"
/opt/retropie/supplementary/runcommand/runcommand.sh: line 1304: 24184 Segmentation fault      /opt/retropie/ports/alephone/bin/alephone "/home/pi/RetroPie/roms/ports/alephone/Marathon/"

Manual check:

pi@retropie:~ $ cd temp/marathon/
pi@retropie:~/temp/marathon $ ls -l
total 25860
drwxr-xr-x 5 pi pi     4096 Oct 30 13:25 Marathon
-rw-r--r-- 1 pi pi 26473230 Jan 15 16:44 Marathon-20220115-Data.zip
pi@retropie:~/temp/marathon $ /opt/retropie/ports/alephone/bin/alephone ./Marathon/
Aleph One Linux 2022-01-15 1.5
https://alephone.lhowon.org/

Original code by Bungie Software <http://www.bungie.com/>
Additional work by Loren Petrich, Chris Pruett, Rhys Hill et al.
TCP/IP networking by Woody Zenfell
SDL port by Christian Bauer <Christian.Bauer@uni-mainz.de>

This is free software with ABSOLUTELY NO WARRANTY.
You are welcome to redistribute it under certain conditions.
For details, see the file COPYING.

Built with network play enabled.

Built with Lua scripting enabled.
GL_VENDOR: Broadcom
GL_RENDERER: V3D 4.2
GL_VERSION: 2.1 Mesa 19.3.2
Segmentation fault
pi@retropie:~/temp/marathon $

mitu

Thanks for the test. If the game works with the new data, we should add it to the script.
For the crash, the traditional way to debug it is with gdb, but that might not be easy to grok. I'll take a look when some free time comes up and maybe I can figure it out where's the issue.

sleve_mcdichael

@mitu said in alephone (Marathon) segmentation fault:

If the game works with the new data, we should add it to the script.

It seems to. Didn't get very far in but it runs & saves & loads the save. Didn't try 2 or 3 yet.

Is the help instruction about switching to software renderer still relevant? It seemed to run a lot smoother using OpenGL, for me. (Does screen size matter? I'm on 720p.)

mitu

@sleve_mcdichael Probably the instructions were written when there was no OpenGL driver available on the RaspberryPi, before Pi4 came along with OpenGL (i.e. non-GLES) support out-of-the-box.

mitu

I've taken a look at this, it's a SDL2/Mesa issue (see here) which gets triggered with the KMSDRM video driver.

TL;DR - using SDL_Quit in an atexit handler will lead to a segfault in this particular scenario. Various other programs have implemented alternative ways to cleanup SDL's state without triggering this bug (see here, here or the patch sent by @psyke83 for gzdoom here).

We'll see if the upstream projects wants to prevent this crash, which otherwise is harmless.