Re: [ccextractor-users] [GSoC] Tests on current 0.70a build - conclusions

[Ruslan K. <kuc...@gm...>]

>
> Explains it. If not, of course EIA-608 should.

It seems that something wrong with WLS.mpg. If we reset cursor position
when we receive roll-up again then the output is wrong. If we don't reset,
everything is OK. But we need to reset. I don't know where else to check
this file.

Let's settle with clear when our PMT disappears from PAT

There were already the condition for that (at line #535 ts_tables.c in
master)

>   and (in the PMT code) if our
> elementary stream disappears, even if there's changing affecting other
> streams.

Function parse_PMT() overwrites caption buffer variables.
So, to fix this bug we should remove clearing of this vars when new pat is
received.

It's worth checking out but my prediction is that parallel processing
> will make things worse, because the bottleneck is the disk

Oh, indeed, that would not make it faster for sure.



On Wed, May 28, 2014 at 10:17 AM, Carlos Fernandez <cf...@gm...> wrote:

> On Wed, May 28, 2014 at 12:04 PM, Ruslan Kuchumov <kuc...@gm...>
> wrote:
> >
> > Yes. It was that wired conditions I didn't pay attention to. I mean, if
> > context->mode was already in roll-up and we received roll-up again, then
> we
> > should reset the cursor. These conditions were messed up. I did't find
> > anything about resetting column in Robson's book.
>
> I believe this document:
> http://edocket.access.gpo.gov/cfr_2007/octqtr/pdf/47cfr15.119.pdf
>
> Explains it. If not, of course EIA-608 should.
>
> > By the way, I fetched repository and I didn't find the changes about
> > clearing after new PAT, but I pushed them. So images.mpg still doesn't
> work.
>
> Ah! Yes. Now I remember. If we're going to alter the PAT/PMT rules for
> clearing/not clearing let's do it correctly. If not, it will work for
> some files and fail for others and we're just going to change which
> files are in the correct/fail sets depending on the strategy.
>
> Let's settle with clear when our PMT disappears from PAT (even if the
> PAT changes for different reasons, such as a new channel that we don't
> care about is added or removed), and (in the PMT code) if our
> elementary stream disappears, even if there's changing affecting other
> streams.
>
> > Willem, your test suite is quite useful :) I have an idea. Since each
> test
> > execution doesn't depend on the others, it's possible to do them in
> parallel
> > to improve time.
>
> It's worth checking out but my prediction is that parallel processing
> will make things worse, because the bottleneck is the disk, not the
> CPU - so processing several files at the same time will just make the
> HD heads move more :-)
>
> Anyway let's see, this is one of the times I'd rather be wrong.
>