[vlc-devel] [RFC PATCH] vout: add the window provider module

[Rémi Denis-Courmont]

Le vendredi 5 avril 2019, 08:45:00 EEST Steve Lhomme a écrit :
> Same here until I implemented the standalone vout_window_t for Windows.
> I know vout_window_t existed but never understood what it does. I never
> realised there was one in Qt (the one we mostly use). And never
> understood why some code was marked deprecated and still not used. Now I
> know that because the vout window stuff was implemented in the display
> modules, hence the total confusion.

When I took over the windowing code, it was full of bugs, void of any 
documentation, its developper had gone MIA, and there was no budget to 
organize workshops.

Stop chasing unicorns, please.

I am already spending more of my free time on code reviews than is good for my 
own sake, and we cannot count on Laurent coming back any time soon. Reading 
documentation, code, specs and making inferences is unavoidable at your or 
Thomas' level.

> The documentation on the vout window is scarce in the code, let alone
> the callback ones.
> 
> And I'm still blurry on the display calls, that seem to have the same
> vout window calls which also have some report calls.

I believe that the last remnants of those were removed by Thomas already.

> An example of confusion call: vout_window_ReportClose()
> "Reportsarequesttoclosethewindow." Does it mean it's done ? Should be
> done ? It calls a closed() callback. The receiver assumes it's closed ?
> The core implementation does nothing with it (just an error log ?!).

It should probably stop the input manager, but nobody ever wrote the missing 
glue code, so it does nothing.

As for the subtlety of closing vs closed, what the meaning of display size is, 
etc, it boils down to setting a sensible convention for a given window type. 
Nobody did that for Win32 as far as I am aware.

-- 
Rémi Denis-Courmont