[Openchrome-users] Mythtv color adjustment

[Thomas Hellström]

Lasse Nisted wrote:

> Hello again
>
> I see that the developers are quite active on the mailing list,
> and I know I can't expect them to answer every single question.
> But I was wondering if I'm supplying wrong information, invalid
> bugs or something like that, since no one answers.
>
> Of course I understand if you're busy and such. This mail is
> NOT meant to offend anyone or seem irritated. It's only because
> I want to know if I'm doing something the wrong way.
>
> I'm very impressed by the work put into these drivers :)
> Thanks!

Hi!
This is most probably caused by different interpretation of what the
XvMC standard says and not says.

The standard says that XvMC attributes, like hue, contrast, colorkey
etc. are per context.

Now every XvMC context belongs to an Xv port, and what happens is
probably that an XvMC context is displaying while MythTV tries to change
the Xv port attributes.

The question is what happens when the Xv port attribute is changed.
Should that change immediately be propagated to all its XvMC contexts or
not?

I've taken the approach that Xv attributes are inherited by an XvMC
context when it is created, and subsequent changes to Xv attributes are
NOT propagated to XvMC contexts.
I believe this is what Nvidia (probably unknowingly) implements. (They
only have one XvMC attribute and that is the colorkey), and this way,
should the player choose to mix Xv and XvMC contents, they can have
separate attributes, just as mixed XvMC contexts can have. Let's say the
player wants do display an Xv frame with a certain colorkey. It would be
bad to change all active XvMC contexts at the same time.

So how should your problem be solved?

The XvMC API has functions to change the attributes of the displaying
XvMC context. Apparently MythTV doesn't use these and instead change the
Xv attribute directly.  Xine usually changes both at the same time, and
I think the best thing would be to ask MythTV developers to do the same
thing. It is not a big change, and unless someone comes up with some
very good reasons to change the via XvMC behaviour it will remain as it is.

I hope this shed some light over the issue.

/Thomas