[mythtv-users] Upgrading from .25 to .26 worthwhile?

Craig Treleaven ctreleaven at cogeco.ca
Fri Aug 16 12:36:44 UTC 2013


At 12:39 PM +1000 8/16/13, Jean-Yves Avenard wrote:
>On 16 August 2013 12:00, Craig Treleaven <ctreleaven at cogeco.ca> wrote:
>  > Really, J-Y, not everything is a personal attack aimed at you.
>
>never thought it was...
>
>but there's a difference between experiencing issues (unlikely to be
>related to myth) and strongly recommending people not upgrade.
>
>You've stepped into the bandwagon whenever someone ask about 0.26 and
>not upgrading.

What?  Yesterday was my first posting talking about my experience 
running .26.  I said "0.26 seems to be..." and "the backend may 
be...".  How did you read that as strongly recommending something?

>This is non-constructive and unhelpful, and actually prevent work to
>properly progress.... We're up to 0.27 now, even if there *may* be
>issues, that people stick to 0.25 doesn't allow to fix 0.26. Only with
>mass adoption can you acquire the critical mass required to fully
>cover the application

Yes, I've built 0.27-pre a few times.  I'd like to get 0.26 
stabilized before I start testing .27. 

>  > So, should we shut down the bug database and go home since there will never
>>  need to be another change to Myth?
>read above...
>
>
>>  In my case, right now, .26 uses more than .25.  I didn't say that _everyone_
>>  has the same issue.  It doesn't help to shoot the messenger whenever they
>  > report something that isn't a glowing positive.
>
>>  I've spent the last 40+ minutes doing a little more testing.  I started a
>>  single HD recording and mythbackend began using between 22 and 35% of a
>>  core.  As expected, mythmetadatalookup ran and then mythcommflag kicked off.
>
>and there's been no changes between 0.25 and 0.26 in that part of the
>code. Only changes that occurred in the past 2 years in that code is
>related to the UTC changes.

I didn't make up the RingBuf warnings and errors; I'd be glad to post 
the log files.  BTW, there was a change to ThreadedFileWriter 7 
months ago that fixed a problem with slow writes [1] in 0.25 for me. 
I see it was applied to 0.26 as well.  The evidence suggests there is 
another issue.

[1] http://code.mythtv.org/trac/ticket/11061

A quick grep of my myth 0.26 log folder shows there are 13k lines 
including "RingBuf"!! Virtually all of them are from 
mythcommflag--there are 26 of these log files (Aug 10 to yesterday). 
The smallest mythcommflag log is 4K; the largest 1MB.

I'd say it is time to file a bug report.

>  >
>>  After 10 minutes, mythcommflag began searching for logo.  The log shows
>>  "RingBuf ...taking too long" and "RingBuf...Waited x.x seconds for data".
>>  At 19:10, avformatdecoder reported an unknown decoding error and commflag
>>  halted.
>
>and you're not using ext4 and enabled barriers by any chance (barriers
>used to be turned off by default in earlier version of the linux
>kernel, and now are on)
>that alone would explain many of your issues. Which would have
>happened with 0.25 (or any earlier version of myth for that matter)

As I stated, I'm running Mac OS X--no barriers.  BTW, I've been 
running Myth exclusively on OS X since 0.20.

Craig


More information about the mythtv-users mailing list