Issues with partial repainting when enabling cairo

Issues with partial repainting when enabling cairo

[Yuri D'Elia]

I briefly tested a build from master with lucid+cairo enabled on X11.
How experimental is "experimental" supposed to be? ;)

It seems that some areas of the frame are not repainted correctly.

For instance, scrolling by pages does not refresh the top (or bottom)
lines at all. More properly, it seems to be a fixed area at the
top/bottom - not really related to visual lines.

Maybe related to 'scroll-margin?

There are some bug reports about cairo on the bts, but nothing so
glaring.

Re: Issues with partial repainting when enabling cairo

[Eli Zaretskii]

> From: Yuri D'Elia <wavexx@thregr.org>
> Date: Fri, 27 Jan 2017 15:55:50 +0100
> 
> I briefly tested a build from master with lucid+cairo enabled on X11.
> How experimental is "experimental" supposed to be? ;)
> 
> It seems that some areas of the frame are not repainted correctly.
> 
> For instance, scrolling by pages does not refresh the top (or bottom)
> lines at all. More properly, it seems to be a fixed area at the
> top/bottom - not really related to visual lines.
> 
> Maybe related to 'scroll-margin?
> 
> There are some bug reports about cairo on the bts, but nothing so
> glaring.

Glaring display bugs are the main problem with Cairo, AFAIK.  We need
motivated individuals who'd work on debugging and fixing those
problems.

TIA

Re: Issues with partial repainting when enabling cairo

[Yuri D'Elia]

On Fri, Jan 27 2017, Eli Zaretskii wrote:
>> There are some bug reports about cairo on the bts, but nothing so
>> glaring.
>
> Glaring display bugs are the main problem with Cairo, AFAIK.  We need
> motivated individuals who'd work on debugging and fixing those
> problems.

Apart from being newer, is there any advantage that cairo should bring
to the display engine?

When I was reading the display code (for debugging the line truncation
hang some months ago) it didn't seem quickly approachable to the
uninitiated. In addition, there's quite some historical baggage one
needs to consider.

Similarly, I wanted to fix the fontconf/Xresource discrepancy for the
freetype rendering params, but again I ran out of time during the
weekend :/

Each time I read the internals I can get further, but I've been using
emacs for 10+ years, and I wonder what kind of guru one needs to be to
have the balls^H^H^H^Hbeard to fix cairo ;)

Re: Issues with partial repainting when enabling cairo

[Eli Zaretskii]

> From: Yuri D'Elia <wavexx@thregr.org>
> Date: Fri, 27 Jan 2017 17:39:17 +0100
> 
> On Fri, Jan 27 2017, Eli Zaretskii wrote:
> >> There are some bug reports about cairo on the bts, but nothing so
> >> glaring.
> >
> > Glaring display bugs are the main problem with Cairo, AFAIK.  We need
> > motivated individuals who'd work on debugging and fixing those
> > problems.
> 
> Apart from being newer, is there any advantage that cairo should bring
> to the display engine?

It's supposed to be the future of GUI desktop on Unix, beyond Xlib
interface and the traditional X server-client setup, so I think it's
important for us to move in that direction.

> When I was reading the display code (for debugging the line truncation
> hang some months ago) it didn't seem quickly approachable to the
> uninitiated. In addition, there's quite some historical baggage one
> needs to consider.

You are talking about xdisp.c, which is entirely irrelevant to Cairo.
The Cairo-related code is in xterm.c and xfns.c (and also in
ftcrfont.c (the "font driver") and in image.c, but those are most
probably unrelated to what you see).

> Each time I read the internals I can get further, but I've been using
> emacs for 10+ years, and I wonder what kind of guru one needs to be to
> have the balls^H^H^H^Hbeard to fix cairo ;)

The trick is to investigate specific problems without trying to
understand the entire big picture.  I'm quite sure the basic design of
the Cairo support is sound, the bugs are in minor details.

Let me know if you need some guidance in finding the places where the
scroll-related problems you saw could hide.

TIA

Re: Issues with partial repainting when enabling cairo

[Stefan Monnier]

> Apart from being newer, is there any advantage that cairo should bring
> to the display engine?

IIUC this would be to most natural path towards adding Wayland support.


        Stefan