[MPlayer-cvslog] r27849 - trunk/libvo/x11_common.c

[Uoti Urpala]

On Fri, 2008-10-31 at 12:08 +0100, Reimar Döffinger wrote:
> On Thu, Oct 30, 2008 at 09:22:33PM +0200, Uoti Urpala wrote:
> > The data is only required during expose events which are
> > the exceptional case, so backing store should not give any benefit
> > during basic playback while the requirement to always keep the data
> > ready in case an expose event happens can add overhead.
> 
> E.g. for XVideo expose events are the only events that matter.

I'm not quite sure what you're trying to say here. Expose and resize
events are the only events that do anything nontrivial performance-wise,
but no event type matters for basic playback performance. Expose events
should only happen when a part of the window that was previously hidden
becomes visible.

> But I do not really want to discuss that again,

I do not remember any claim of possible "performance improvement" by
enabling backing store in earlier discussion.

>  I just want a commit
> message that give at least some idea of the point of it, which hopefully
> the committer can answer.

I agree that the commit message did not explain the rationale clearly,
but I'm not sure what you mean by the "which hopefully the committer can
answer" part. Do you think it's still unclear whether a justified
rationale exists? IMO the fact that there are documented cases where
backing store is harmful and no indication of it being useful does
justify the commit.