2014-07-15, 08:20
I've uploaded another debug build, #0715b, that is basically the same as #0714b but includes PR5034, fixing the segmentation fault when running script.grab.fanart (and possibly others).
Please test this build and if you experience more crashes then paste a backtrace and hopefully someone (most probably @jmarshallnz) will be able to pinpoint where the fix needs to be applied. Without a backtrace this is much more difficult, if not impossible.
Apart from being significantly bigger than a regular build (260MB vs. 117MB) due to all the debug symbols etc., it should behave the same as a regular (non-debug) build, maybe running a little bit slower.
However, when run under gdb (the Linux debugger) it will allow a backtrace to be obtained. You'll only need to run it under gdb if you want a backtrace, otherwise install and run it as you would a non-debug build.
It won't be possible to run this build under gdb on a 256MB board, but running without gdb should work (probably).
Once we're happy that the recent changes are no longer causing a problem I'll switch back to non-debug builds.
To capture a backtrace with gdb:
Please test this build and if you experience more crashes then paste a backtrace and hopefully someone (most probably @jmarshallnz) will be able to pinpoint where the fix needs to be applied. Without a backtrace this is much more difficult, if not impossible.
Apart from being significantly bigger than a regular build (260MB vs. 117MB) due to all the debug symbols etc., it should behave the same as a regular (non-debug) build, maybe running a little bit slower.
However, when run under gdb (the Linux debugger) it will allow a backtrace to be obtained. You'll only need to run it under gdb if you want a backtrace, otherwise install and run it as you would a non-debug build.
It won't be possible to run this build under gdb on a 256MB board, but running without gdb should work (probably).
Once we're happy that the recent changes are no longer causing a problem I'll switch back to non-debug builds.
To capture a backtrace with gdb:
- Make as much RAM available as possible. Ways to increase available RAM include:
- Addding "noram" to /flash/cmdline.txt to stop OpenELEC loading the SYSTEM image into RAM
- Dropping gpu_mem as low as possible, eg. 96
- Running the GUI at 720p and disabling higher quality textures
- Adding 128MB of swap might help, but is very slow and the Pi might appear to freeze when using swap
- Addding "noram" to /flash/cmdline.txt to stop OpenELEC loading the SYSTEM image into RAM
- If necessary, reboot with the new memory configuration set in #1
- Execute the following in ssh:
Code:systemctl stop xbmc
gdb /usr/lib/xbmc/xbmc.bin - In gdb, execute the following commands:
Code:thread apply all bt
run --standalone -fs - Once the crash occurs and control returns to gdb, execute the following command to output the backtrace:
Code:bt