[helium-dev] xserver-xorg-video-intel optional in Stretch

Head_on_a_Stick · 2017-06-03 22:48:28

johnraff wrote:

Is there some similar test we could do before applying '--backend=glx'?

I've been looking at this again and I think perhaps we should just add --backend glx, it seems to fix the problem for all cards:

https://forums.bunsenlabs.org/viewtopic.php?id=3149

https://forums.bunsenlabs.org/viewtopic.php?id=1472

I thought that there was a specific reason not to apply the GLX backend but I can't actually find it -- does anybody else know better?

hhh · 2017-06-04 02:56:08

^ That looks good to me, though I also can't speak for all, or even some, graphics cards.

johnraff · 2017-06-05 03:34:17

Should we start a new thread to get suggestions on the best compton options to make default, or is '--backend=glx' + the current bl-compositor test enough?

What about bronto's '-cCG --paint-on-overlay'?

brontosaurusrex · 2017-06-05 05:06:29

I have been using that (paint on overlay) on Intel chipset as well, however can't remember the reason for it.

Last edited by brontosaurusrex (2017-06-05 05:07:19)

Head_on_a_Stick · 2017-06-05 06:22:25

johnraff wrote:

is '--backend=glx' + the current bl-compositor test enough?

That seems to work for my Haswell hardware and also for the OP @kafran.

What about bronto's '-cCG --paint-on-overlay'?

That does make things very slightly better for me (I use that along with `--vsync opengl-mswc` in my system) but I seem to remember @nobody being opposed to that particular flag so the effects may be variable.

I would vote to keep it simple and swap "--vsync opengl" for "--backend=glx"; the compton man page does actually warn that our current option may perform poorly on some hardware.

EDIT: there is a trade-off when controlling tearing with vsync in respect of lag during dragging operations, some find the latter more troublesome than the former so it may be best to use the less aggressive setting.

Last edited by Head_on_a_Stick (2017-06-05 07:06:55)

johnraff · 2017-06-05 07:08:23

Head_on_a_Stick wrote:

I would vote to keep it simple

agreed

...and swap "--vsync opengl" for "--backend=glx"

OK so drop the current test

EXECXCOMP='compton -b'
if glxinfo | egrep -iq 'direct rendering: yes'; then
    EXECXCOMP+=' --vsync opengl'
fi

for a plain

EXECXCOMP='compton -b --backend=glx'

?

...it may be best to use the less aggressive setting.

Meaning what?

Head_on_a_Stick · 2017-06-05 07:12:53

johnraff wrote:

so drop the current test

No, keep the test but change EXECXCOMP+=

johnraff wrote:

...it may be best to use the less aggressive setting.
Meaning what?

Just have the --backend glx flag, leave the --paint-on-overlay out (or commented).

johnraff · 2017-06-05 08:39:20

^ah OK. Sounds good for now.

johnraff · 2017-06-10 08:21:34

Sorry, I just changed --vsync opengl to --backend=glx in a test helium-dev package, and in a Virtualbox VM it completely broke the x session.

Even if this turned out to be a VB-specific issue, it's still surely not acceptable? Anyway, for now I'll have to revert the change.

Let's think about this some more.

Head_on_a_Stick · 2017-06-10 14:37:30

johnraff wrote:

in a Virtualbox VM it completely broke the x session

I knew there was a reason we didn't use that, thanks.

johnraff wrote:

Even if this turned out to be a VB-specific issue, it's still surely not acceptable?

Perhaps.

Would this be too hacky:

empty@Diproton:~ $ diff /usr/bin/bl-compositor /home/empty/bin/bl-compositor
62c62,66
<     EXECXCOMP+=' --vsync opengl'
---
>       if grep hypervisor /proc/cpuinfo; then
>               EXECXCOMP+=' --vsync opengl'
>       else
>               EXECXCOMP+=' --backend=glx'
>       fi

DeepDayze · 2017-06-10 14:49:12

johnraff wrote:

Sorry, I just changed --vsync opengl to --backend=glx in a test helium-dev package, and in a Virtualbox VM it completely broke the x session.
Even if this turned out to be a VB-specific issue, it's still surely not acceptable? Anyway, for now I'll have to revert the change.
Let's think about this some more.

If the script can detect whether we are installing in a Virtualbox VM the bypass that setting otherwise apply it if we are installing to bare metal.There has to be a way to detect if installation is into a VM rather than bare metal. Not sure how VMWare or QEMU or other virtualization methods such as Hyper-V will handle this.

johnraff · 2017-06-12 03:08:12

Too fast... no-one said it was a VB issue, just that it was discovered in a VB session!!

See comments by @nobody & @johnraff in the GitHub commit comments (commit since reversed). It's looking to me as if the problem might lie with the combination of compton's -b (daemonize) and --backend glx options. If started without -b then everything's fine.

At worst, we could probably fork compton ourselves with a final & (as it was in the past) instead of using the (neater) -b option.

EDIT: On my regular Jessie system -b and --backend glx work together OK, although firefox scrolling is much better with the old --vsync opengl. Any test reports of problems with -b and --backend glx gratefully received.

Last edited by johnraff (2017-06-12 05:34:15)

johnraff · 2017-07-01 07:55:52

kafran wrote:

...at least an improvement would be "bl-compositor" receiving compton arguments or have this as a configurable option so the user do no need to change the "bl-compositor" program but "~/.conf/openbox/autostart"

I think this is what we might have to do.

Actually, just starting compton without any command line options at all works fine on my system.

#21 2017-06-03 22:48:28

Re: [helium-dev] xserver-xorg-video-intel optional in Stretch

#22 2017-06-04 02:56:08

Re: [helium-dev] xserver-xorg-video-intel optional in Stretch

#23 2017-06-05 03:34:17

Re: [helium-dev] xserver-xorg-video-intel optional in Stretch

#24 2017-06-05 05:06:29

Re: [helium-dev] xserver-xorg-video-intel optional in Stretch

#25 2017-06-05 06:22:25

Re: [helium-dev] xserver-xorg-video-intel optional in Stretch

#26 2017-06-05 07:08:23

Re: [helium-dev] xserver-xorg-video-intel optional in Stretch

#27 2017-06-05 07:12:53

Re: [helium-dev] xserver-xorg-video-intel optional in Stretch

#28 2017-06-05 08:39:20

Re: [helium-dev] xserver-xorg-video-intel optional in Stretch

#29 2017-06-10 08:21:34

Re: [helium-dev] xserver-xorg-video-intel optional in Stretch

#30 2017-06-10 14:37:30

Re: [helium-dev] xserver-xorg-video-intel optional in Stretch

#31 2017-06-10 14:49:12

Re: [helium-dev] xserver-xorg-video-intel optional in Stretch

#32 2017-06-12 03:08:12

Re: [helium-dev] xserver-xorg-video-intel optional in Stretch

#33 2017-07-01 07:55:52

Re: [helium-dev] xserver-xorg-video-intel optional in Stretch

Board footer