[vlc-devel] [PATCH 3/3] transcode: video: fix insertion of fps filter

[Rémi Denis-Courmont]

Le keskiviikkona 27. tammikuuta 2021, 10.58.32 EET Francois Cartegnie a écrit 
:
> Le 26/01/2021 à 22:03, Rémi Denis-Courmont a écrit :
> >> That's one example why we should have kept full init() clean() cycles
> >> everywhere, and even if the init is zero op.
> > 
> > We cannot keep something we literally never had. And lets not even get
> > started on lack of error handling and the probability to make it happen
> > without introducing even more bugs.
> 
> I remember did remove the union at some point

In other words, you made a ugly hack that partially papered over the issue, 
increased memory consumption, and effectively regressed work I'd done.

And?

> > Unlike video_format_t, there are no realistic ways that we could make
> > es_format_t serial, but your argument is simply invalid, and so far nobody
> > really came up with a good idea.
> 
> There was enough discussions on this ml when ^above happened. filip and
> I were pushing for a full cycle API and cleanup the es_format mess, some
> people didn't want because there was "no need".

Where and when did who say that there is no need to clean up es_format_t?

I maintain that video_format_t should be made serial - meaning that palette 
support should be ditched or moved to es_format_t. That's not the same as 
making es_format_t serial, which is totally impractical.

If you have a concrete workable solution, you propose it. Otherwise, it's just 
pointless ranting and ponting fingers. Removing the union does not fix the 
problem.

-- 
Rémi Denis-Courmont
http://www.remlab.net/