[mythtv] Need help with frontend seg fault

[Jerry Rubinow]

> On Sunday 25 July 2004 11:53 pm, Jerry Rubinow wrote:
> > On Sunday 25 July 2004 09:30 pm, Jerry Rubinow wrote:
> > >> I've been having segmentation faults in the frontend while trying to
> > >> play back 1080X1920 and 1280X720 HD content recording using a pcHDTV.
> > >> 704X480 content appears to play back ok. I'm running the stock 0.15
> > >> sources without any deinterlacing or filters.
> > >
> > ><snip>
> > >
> > >> Can anybody help me out? What other information should I supply? What
> > >> else can I try? I'm very close to having a running hi-def MythTV system
> > >> and it's frustrating being able to record shows but not play them back.
> > >> If anyone wants to try playing the recording, it's about 250MB (2
> > >> minutes long).
> > >
> > >Try running it through valgrind?
> >
> > Ah, that looks like a useful thing to do - I was not aware of this utility
> > before.  I've attached the output, does it help any?
>
> Try using a standard file attachment next time, not some non-portable junk.
>
> The output by itself is unlikely to help me any, anyway, it was just a way to
> get you more information so you could debug further.
>
> Isaac

Running with valgrind gave me the following output (at end of this post).  Line 153 of mptegts.c is 
   tss->end_of_section_reached = 1;
which seems like it would be an invalid write, if anything, not an invalid read.  In any case, I don't know enough about this mpeg decoding code to know if this is an indication of a serious error.  Being new to Myth, I don't even for sure whether this could be some kind of configuration error rather than a bug (well, sure it's a bug since it causes a seg fault, but I mean something I could make go away by changing something in the configuration options).  Unless someone has any better suggestions, I would guess that two things I should try first would be:
 
1. Get someone to try playing my data file in 0.15 and see if they get a seg fault, in order to determine whether it's just me.
2. Try upgrading to the cvs code and see if the error goes away.
 
Would anyone be willing to try out my 250MB data file and let me know the results?
 
I'd welcome any other suggestions people have.  
 
-Jerry
 
latter part of valigrind output:
 
==29263== Thread 2:
==29263== Invalid read of size 1
==29263==    at 0x983FCA: write_section_data (mpegts.c:153)
==29263==    by 0x985510: handle_packet (mpegts.c:934)
==29263==    by 0x98574F: handle_packets (mpegts.c:1015)
==29263==    by 0x985B2D: mpegts_read_header (mpegts.c:1133)
==29263==    Address 0x40EC29F8 is 24 bytes inside a block of size 36 free'd
==29263==    at 0x78793D: free (vg_replace_malloc.c:231)
==29263==    by 0x47489E: av_free (mem.c:98)
==29263==    by 0x9841CC: mpegts_close_filter (mpegts.c:221)
==29263==    by 0x9849A8: pat_scan_cb (mpegts.c:543)
==29263==
==29263== Thread 2:
==29263== Invalid read of size 1
==29263==    at 0x983FCA: write_section_data (mpegts.c:153)
==29263==    by 0x985510: handle_packet (mpegts.c:934)
==29263==    by 0x98574F: handle_packets (mpegts.c:1015)
==29263==    by 0x985BBE: mpegts_read_header (mpegts.c:1156)
==29263==    Address 0x40E971E8 is 24 bytes inside a block of size 36 free'd
==29263==    at 0x78793D: free (vg_replace_malloc.c:231)
==29263==    by 0x47489E: av_free (mem.c:98)
==29263==    by 0x9841CC: mpegts_close_filter (mpegts.c:221)
==29263==    by 0x98483C: pat_cb (mpegts.c:490)
Input #0, mpegts, from '/video/programs/1057_20040725010300_20040725010400.nuv':
  Stream #0.0[0x11]: Video: mpeg2video, 1920x1080, 29.97 fps
  Stream #0.1[0x14]: Audio: ac3, 48000 Hz, stereo, 384 kb/s
Input #0, mpegts, from '/video/programs/1057_20040725010300_20040725010400.nuv':
  Stream #0.0[0x11]: Video: mpeg2video, 1920x1080, 29.97 fps
  Stream #0.1[0x14]: Audio: ac3, 48000 Hz, stereo, 384 kb/s
OSS
2004-07-26 19:30:48 Opening OSS audio device '/dev/dsp'.
2004-07-26 19:30:48 Audio fragment size: 4096
2004-07-26 19:30:48 Using XV port 56
Couldn't find OSD theme: abstractosd
2004-07-26 19:30:56 Changing from None to WatchingPreRecorded
Segmentation fault

-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/ms-tnef
Size: 7083 bytes
Desc: not available
Url : http://mythtv.org/pipermail/mythtv-dev/attachments/20040726/53f2b1d2/attachment.bin