Opened 14 years ago

Closed 14 years ago

#1058 closed defect (fixed)

XVideo segfault on input change

Reported by: buzz <buzz@…> Owned by: danielk
Priority: minor Milestone: unknown
Component: mythtv Version: head
Severity: medium Keywords:
Cc: Ticket locked: no

Description

USING SVN version 8636, compiled with "--enable-proc-opt --enable-dvb --compile-type=debug":

attached is the log from running this:

#gdb mythfrontend gdb>run -v all,nodatabase <i hit c to continue gdb, so frontend runs> <in frontend app, press enter to enter LiveTV> <then once the segfault occurs (few seconds later), I type: gdb>bt

if you need the backend log, I have it too.

Attachments (7)

frontend.log.segfault.bt.txt (21.8 KB) - added by buzz <buzz@…> 14 years ago.
gdb_with_proc_opt.txt (7.8 KB) - added by buzz <buzz@…> 14 years ago.
myth_with_proc_opt.log (10.4 KB) - added by buzz <buzz@…> 14 years ago.
gdb_no_proc_opt.txt (8.0 KB) - added by buzz <buzz@…> 14 years ago.
myth_no_proc_opt.log (10.0 KB) - added by buzz <buzz@…> 14 years ago.
1058-dbg-v1.patch (1.6 KB) - added by danielk 14 years ago.
Debug patch for CreateShmImages? problem
myth.log (24.9 KB) - added by buzz <buzz@…> 14 years ago.

Download all attachments as: .zip

Change History (23)

Changed 14 years ago by buzz <buzz@…>

comment:1 Changed 14 years ago by danielk

Milestone: 0.19
Owner: changed from Isaac Richards to danielk
Version: head

comment:2 Changed 14 years ago by danielk

(In [8652]) References #1058, and references #990.

This adds a verbose to VideoOutputXv? when shm buffers are created. This should make any problems on buffer creation debuggable.

comment:3 Changed 14 years ago by danielk

Please retest with the latest svn, using 'mythfrontend -v playback'. And please compile without '--enable-proc-opt'.

I'll need the log of course, and the 'thread apply all bt' backtrace.

Does this happen with all your recorders, or only a specific one? What kind of device is it and what is your video codec and resolution (if applicable)?

comment:4 Changed 14 years ago by buzz <buzz@…>

My device is a DViCO "FusionHDTV DVB-T Plus" card. Chipset is a cx23882-19 This is seen in SVN myth as a "Zarlink MT352 DVB-T'. I only have one recorder.

rebuild of SVN in progress (as requested), will advise.

comment:5 Changed 14 years ago by buzz <buzz@…>

video codec and resolution...? I'm not sure what you mean by that. Do you mean SVGA resolution(1280x1024), or resolution of media stream(variable, how do I get this info)? Codec?... I have no idea what you mean by this, whatever the default is?

comment:6 Changed 14 years ago by danielk

Buzz, just post those logs. I can get the information I need from them.

comment:7 Changed 14 years ago by danielk

Milestone: 0.19
Resolution: invalid
Status: newclosed

Needed info not provided by submitter, closing.

comment:8 Changed 14 years ago by buzz <buzz@…>

Resolution: invalid
Status: closedreopened

sorry I'm not as quick as you want me to be... but I have another problem occuring (tuning is failing) that is more critical than hitting the 4GB boundary, and has prevented me from getting the logs that you require for this one - It's started happening directly when I tried to "retest with latest svn" as you instructed, and I've been trying to revert my svn down again, but I can't fugure out how, as even though svn says I'm on the correct slightly older svn version, I'm getting compile-time errors (even after a dozen different make distclean/configure/make cycles. I haven't opened a ticket for that, though, as it's just a local problem, I'm sure.... just not sure where. I'm re-opening ticket. If I don't get back to you in 7 days, feel free to close it again.

comment:9 Changed 14 years ago by danielk

Resolution: invalid
Status: reopenedclosed
svn co -r REVISION http://svn.mythtv.org/svn/trunk/mythtv mythtv-X

Don't reopen the ticket until you have the needed logs.

comment:10 Changed 14 years ago by buzz <buzz@…>

Resolution: invalid
Status: closedreopened

1) Cnfirming problem still exists with SVN 8733. 2) Problem is channel specific - only happens on entering live TV if one of the problem channel/s is the default (which it will be if it was the last watched when mfe crashed). Ie channel that caused mfe to crash is set as default channel (last watched?) for mfe to watch when starting livetv, and hence causes crash. 3) problem channel/s is High Def, not all high-def channels cause this problem - AFAIK. 3.5) given problem channel/s are high-def, I have initially left enable-proc-opt enabled, but will be providing a second set of logs with it turned off, when my build finishes. 4) definitely occurs when compiled with: ./configure --enable-dvb --enable-proc-opt --compile-type=debug 5) gdb was run thus:

handle SIGPIPE nostop noprint handle SIG33 nostop noprint set logging on set pagination off set args -l myth.log -v playback #run #thread apply all bt full

6) I have attached files gdb.txt and myth.log output when compiled with "--enable-proc-opt" (note, I'll also be providing logs with --enable-proc-opt removed, in a little while.

Changed 14 years ago by buzz <buzz@…>

Attachment: gdb_with_proc_opt.txt added

Changed 14 years ago by buzz <buzz@…>

Attachment: myth_with_proc_opt.log added

comment:11 Changed 14 years ago by buzz <buzz@…>

problem appears identical when compiled without --enable-proc-opt. logs attached.

Changed 14 years ago by buzz <buzz@…>

Attachment: gdb_no_proc_opt.txt added

Changed 14 years ago by buzz <buzz@…>

Attachment: myth_no_proc_opt.log added

comment:12 Changed 14 years ago by buzz <buzz@…>

I can now confirm the following:

A) problem is only high-def channels. B) problem is not all high-def channels (of 3 prime-time high-dev channels we have, 2 suffer the problem) C) work-around to get live-tv functioning temporarily again: schedule a recording of a currently occuring program on a non-problematic channel, wait a few seconds for the tuner to tune to that channel, and then cancel the recording. Tuner is now on an OK channel, and live TV can be started without crashing, Avoid watching problematic channels, including avoid surfing 'over' then.

Changed 14 years ago by danielk

Attachment: 1058-dbg-v1.patch added

Debug patch for CreateShmImages? problem

comment:13 Changed 14 years ago by danielk

Buzz, please send me the log with the attached patch.

It looks like CreateShmImages?() is being called correctly. So maybe there is some problem inside that function. This should print out info about what it is doing and later what clear_xv_buffers() is doing. I don't need the backtrace, unless it somehow changes.

comment:14 Changed 14 years ago by danielk

Summary: mfe segfault entering liveTV - SVN 8636XVideo segfault on input change

comment:15 Changed 14 years ago by buzz <buzz@…>

frontend log with your patch applied is attached. (Sorry I'm not being much help here, but thread debugging is all a bit beyond me):

Changed 14 years ago by buzz <buzz@…>

Attachment: myth.log added

comment:16 Changed 14 years ago by Isaac Richards

Resolution: fixed
Status: reopenedclosed

(In [8741]) Should fix #1058, at least the crashing part of things.

In this case, it looks like a video card and/or driver is being used that can't handle Xv images of 1920x1088. Bail out if we don't get a size large enough when trying to create the image surfaces, since we won't be able to decode anything.

Note: See TracTickets for help on using tickets.