[mythtv-users] Playback problem -- random short pauses

Kenni Lund kenni at kelu.dk
Mon Mar 28 22:18:32 UTC 2011

2011/3/28 Kenneth Emerson <kenneth.emerson at gmail.com>:
> I can only guess that what I need now is some way to get a constant stream
> (data collection) of I/O and CPU activity that has a small enough
> granularity that it might catch some anomaly when I see the pause in the
> playback.  Any suggestions gratefully accepted.

Another thing to try, which would help a lot, is to figure out what
commit introduced the issue in MythTV. If it was introduced at some
point in 0.24-fixes, then the faulty commit can be identified quite
easily with the help of "git bisect". The following actions would need
to be performed:

1. Checkout MythTV at the first git commit in 0.24-fixes (~november
2010), compile it and see if the problem exists in this version. If it
DOESN'T exist in this version, we now know that the issue was
introduced in 0.24-fixes and we can continue with git bisect to
identify the commit.

2. Tell git that we want to perform a git bisect, something like:
git bisect start
git bisect bad $latest_0.24-fixes_git_revision_which_has_the_issue
git bisect good $the_initial_0.24-fixes_git_revision_which_DOESNT_have_the_issue

3. Git now performs a checkout of a revision in between the two
revisions. Compile it and see if the problem exists or not.

4. Depending on the test results, give git feedback on the tested revision:
git bisect good
- or -
git bisect bad

Git will checkout a new version to test; test it, give feedback, etc.
In the end, git will spit out the commit which introduced the issue -
which should be the code that needs fixing.

But take care when judging whether a given revision is good or bad -
one wrong answer will identify the wrong commit in the end, so make
sure it's the same issue you're tracking during the testing (other
things, like damaged frames, can also cause "short pauses" in
playback, so make sure to keep an eye on the logs).

Best regards

More information about the mythtv-users mailing list