[mythtv-users] Problems with FireWire recording

[Jason]

>> Occasionally, my firewire connection to the DCH-3200 goes nuts and
>> needs to be reset.  I'm OK with that,
>
> You shouldn't be.

 You're right, but until software isn't written by people and bugs don't
 happen, I'll accept living with problems.  :)

> This was due to a virtual memory bug in the kernel. The 1394 libs
> would report an error until reset even though the connection to
> the device was really okay. A reset didn't assure that it wouldn't
> happen again minutes or seconds later.
>
> This was fixed in 2.6.23.6 or later. There may also be a workaround
> in new versions of the 1394 libs but if you use 2.6.24.x, it works
> fine.

 I didn't realize that this had been fixed in the kernel.  I'll have to
 investigate whether this fix has been backported into the latest CentOS
 kernel.  You don't happen to have a reference to the VM bug in the kernel 
do
 you (or is a 1394 driver error)?  That would make it easier to track 
down...


> There were tons of scripts and procedures to try to deal with
> the problem. Here's one that would work as well as any other:
> Get up and go to the next room, scratch your ass, come back and
> sit down, use firewire_tester with -R then try again and hope
> it doesn't fail.

 I had been using such scripts with the 0.20.2.  I didn't use the scripts
 because I'd seen a lot of list traffic that 0.21 would make it all better
 and not require such things.  Obviously, that was wrong.  It's sometimes
 hard to keep track of which things we have to do for a certain version 
don't
 need to be maintained for a future release.  Doc on changes is 
unfortunately
 (and expectedly) slower than the changes themselves.

 It would still be nice if Myth could figure out that it's communicating 
with
 something that's not there and give up.  It seems like a bug (poor behavior
 at least) if it sees an error in the previous 20,000 tries at something and
 then tries again expecting a new behavior.  I accept that the next result
 could change, but after a certain point, as a viewer, I'm not going to care
 that it got the last 35 minutes of an hour show, as I'm going to tell it to
 try to re-record it anyway (I think most users are going to do this). 
There
should be a way for it to try for a minute or so, then move on.

> A much more effective procedure is to upgrade your kernel.

 Agreed.

 Thanks, Bruce!

 Jason