Dropping X11 VCL backend

[Luboš Luňák]

On Tuesday 29 of December 2020, Ilmari Lauhakangas wrote:
> We could start by gathering the concrete things that we use X11 backend
> for.
>
> One of them is UI tests, where we advise to use SAL_USE_VCLPLUGIN=gen
> when we want to observe the UI while running tests or while logging UI
> events in the domain specific language:
> https://wiki.documentfoundation.org/Development/UITests
>
> Hopefully someone can explain why we still need to do this.

 I think this is simply because 'gen' is the lowest common denominator and 
thus considered the safest. You can build LO without Gtk or KF5 backends, 
but 'gen' is always there.

On Tuesday 29 of December 2020, julien2412 wrote:
> There's also kf5 rendering (corresponding to kde, I don't know why
> it's not called "kde5") which doesn't use accessibility but need a lot of

 The organization is KDE, the code is KDE Frameworks. Calling the VCL 
backend 'kde5' would be like calling the Windows one 'microsoft10'.

On Tuesday 29 of December 2020, julien2412 wrote:
> I understand but IMHO Skia support in LO is not enough mature (first commit
> on master branch of LO is in 2019-11-27 so just 1 year whereas it's an
> essential component) and above all, it needs at least one additional
> maintainer (for me the main pb here) (see my response to Chris).

 Maturity is not measured only by time. I can easily spin this by saying that 
Skia is the best performing drawing code and the most maintained (or 2nd most 
maintained, maybe, after gtk3), and most backends AFAIK have zero 
maintainers. So you're ignoring the other side of the equation, such as that 
this whole thread is about the possibility of removing that other side of the 
equation. And I'd be concerned about other things, such as the fact that with 
Skia we're much more at the mercy of upstream than with other options.

(Obviously, I'm biased.)

-- 
 Luboš Luňák
 l.lunak at collabora.com