SVN # 15927 - mythtv-setup - channel scanner constantly segfaults while scanning

[dstrang@…]

Since the merge of multirec to trunk (and possibly before), the mythtv-setup scanner crashes in the middle of the scan. This happens on both DVB-S, OTA, and an HDHomeRun connected to Comcast cable.

The scanner in 0.20-fixes works just fine, however constantly downgrading to scan is a bit of a pain.

The attached log and backtrace are from scanning with an HDHomeRun set to unencrypted channels only. The following options were selected:

Frequency Table: Cable Modulation: QAM-256 ATSC Channel Seperator: (5.1) Period Existing Channel Treatment: Minimal Updates

I get the same crashes scanning 8VSB with my Air2PC/AirStar 2 ATSC 1st generation card.

[dstrang@…]

mythtv-setup -v all > output.txt

[dstrang@…]

gdb dump of mythtv-setup channel scanner crash from 0.21-fixes rev 16058

[dstrang@…]

output.txt from mythtv-setup channel scanner crash from 0.21-fixes rev 16058

[dstrang@…]

Some additional information; I tested this against the latest 0.21-fixes branch today, the problem still exists. When scanning 8VSB on my Air2PC/AirStar 2 ATSC card, the scanner appears to work but the GUI locks up at the end. It sometimes takes multiple scans to get all the OTA channels. As for the HDHomeRun, it continues to crash after scanning just 4-5 channels, it never finishes a complete scan.

Here are my compile options for mythtv;

./configure --prefix=/usr --enable-dvb --disable-dbox2 --disable-joystick-menu --enable-hdhomerun --enable-opengl-vsync --enable-opengl-video --enable-mmx --compile-type=debug --enable-valgrind

NOTE: the debug & valgrind were added just to do the backtrace.

[billybobjones2112@…]

I was about to open a very similiar bug. The trunk release consistently cores immeduately when trying to scan DVB-S (QPSK). There is a thread on the DVBN forum where multiple users are complaining about the same issue. I am a little worried that this one seems to have morphed into a an 8VSB specific issue.

note to ijr - please let me know if I should open an additional bug with all logs, etc

[concordfta <dragozet@…>]

I can also confirm this bug. Last tested trunk svn 16133. Using a HDhomerun in 8VSB mode. Scanning segfaults in Mythtv-setup on first proper 8vsb atsc channel found during scan. Playing with hdhr time-outs did not correct problem.

[anonymous]

Just to comment -t his is still going on in SVN as of today. It's really annoying because without the channel scan working it is hard to get going.

It is inconsistant as well because it works maybe 1/15 times. Must be a race condition.

[anonymous]

One other comment.... personally I fid if I use 'space' to select 'Next' instead of the enter key, the crash does not seem to happen hardly ever. So maybe it is due to enter key bindings.

[anonymous]

Using space or enter does not matter here. Usually segfaulting, sometimes even C++ abort. Need to check in more details why this happens, but have to agree that it practically renders whole 0.21-fixes and later mythtv unusable state here :|.

[laga]

Does it work if you bind mythtv-setup to one CPU?

taskset -c 0 /usr/bin/mythtv-setup.real

Of course, this only applies for those with SMP boxes :)

[anonymous]

Yes, this is dual-cpu box, but enabling valgrind with full leak checked made scanner work ok :) Will try forcing to one cpu only.

[anonymous]

Replying to anonymous:

Yes, this is dual-cpu box, but enabling valgrind with full leak checked made scanner work ok :) Will try forcing to one cpu only.

Yep. Taskset to single cpu seems to have fixed the problem.

[danielk]

Use the taskset workaround for 0.21.

This doesn't happen with the scanner in the channel scan branch, but that uses a completely different UI infrastructure.

I'm marking this as won't fix, since the code causing the segfault will be replaced before 0.22.