GNOME Bugzilla – Bug 723331
a crash while trying to port from gst 0.10
Last modified: 2014-05-18 10:49:58 UTC
This might seem as if it would be better handled as a support question than a bug, but as awhile back I've asked (on #gstreamer) if my attempt at porting looks sane, I was told it did. So, zdoom has a semi-unofficial openal branch, that uses gstreamer 0.10 (https://github.com/rheit/zdoom/tree/openal). The whole gstreamer code is in src/sound/oalsound.cpp. I've tried porting this to 1.0. I was mostly successful, as it *usually* works quite well, but besides a need for some polish (that is minor code changes due to caps string changes), there are still issues with this patch. First of all, every now and then (probably on each new file added to the queue) gstreamer prints a critical for gst_segment_to_running_time assertion about 'segment->format' (GST_FORMAT_BYTES) != 'format' (GST_FORMAT_TIME). Second, which may or may not be related, in an easily reachable place in hacx.wad from http://drnostromo.com/hacx I'm hitting 'dropping short buffer' condition of helper_find_pick (libs/gst/base/gsttypefindhelper.c) followed shortly by a crash. The same place works fine with 0.10, but as I said, I can't see anything *obviously* wrong with the code. gdb output doesn't really help me - it's a double free assert from glibc triggered by gst_buffer_unref. For the attached patch to work, at minimum an obvious change in src/CMakeFiles.txt in Linux block regarding gstreamer version. In the patch default audio sink of playbin is used instead of openalsink, but that's not the source of the problem - that works just fine in 0.10.
*** Bug 723332 has been marked as a duplicate of this bug. ***
Created attachment 267698 [details] [review] my attempt of 1.0 port
Created attachment 267699 [details] gdb output upon crash
Created attachment 267758 [details] [review] gst-plugins-good patch OK, there's a funny thing - the crash seems to be a mistake on my part (looks like I should have used gst_buffer_wrapped_new_full, not shorthand), but I might have found a bug in gst-plugins-good in flac (and possibly cutter) plugin. It uses "S8LE" (well, NE to be exact), which some of the other plugins reject. Here it cause a caps mismatch between "S8LE" and "S8". I don't know much about audio formats, but does endianness make sense for 8bit format ? As for my zdoom patch, now I'm getting "Unhandled format: 0 bit, 1 channel" instead of a crash, but I may manage to figure that out on my own.
Comment on attachment 267758 [details] [review] gst-plugins-good patch Thanks, is anything else left to be done here? commit ac4df5e2c543acf1834cbaa5534012365b1d4d34 Author: Rafał Mużyło <galtgendo@o2.pl> Date: Tue Feb 4 13:43:56 2014 +0100 gst: Don't use endianness-specific S8 audio format It does not exist. https://bugzilla.gnome.org/show_bug.cgi?id=723331
Closing this bug report as no further information has been provided. Please feel free to reopen this bug if you can provide the information asked for. Thanks!
The last part was independent from gstreamer.