I've heard back from Mali upstream and this particular bug has been fixed in r20 of their blob, and I'm pushing Rockchip to get that driver released. I'm still of the opinion that adding the error check is a good idea though. Any thoughts?

In D18460#398164, @zzag wrote:

I don't think that's the right way to fix the problem.

Do you have OpenGL ES 3.2?

In D18460#398165, @zzag wrote:

Also, it looks like we don't create debug context.

I submitted a talk about deploying Plasma on Embedded devices.

Hey
I'd like to attend and help run the booth and anything else that might need doing, I've worked on getting Plasma working on the Pinebook and other misc. ARM devices.

Update return types

Update as per comments

Make sure we return the right value

LGTM

BTW this patch works as tested on both the Rock64 and Pinebook, but kwin fails in another place now. Will continue investigating.

Simple Login does not evaluate it. It should, or we should use sddm on the phone too.

Can this also go into the next 5.10 release?

You can checkout the XML on http://www.nationalgeographic.com/photography/photo-of-the-day/ by just viewing the source :)

Unfortunately I couldn't figure out a way to relax the XML parser, so if anyone has tips on how to achieve that, I'd highly appreciate them :)

Remove useless debug

Remove accidentally commited file

Update as per David's comments

Mention the name of the widget too

@scarlettclark Sounds about right to me. The tricky part imo is probably how one restarts the job on the publisher.

Poke? Could someone point me in the right direction here? :)

Nope, needs to block all jobs on Jenkins somehow so that Jenkins doesn't try and publish while aptly is restarting.

Oh cool, perhaps you can simply restart aptly once in a while to trigger this?

So I was thinking of somehow clipping the buffer that we present to the backend to only represent the damaged region. I'm not quite sure yet how to clip the QImage to the damaged QRegion we have, but does that sound sensible?

Oh, after reading the code a bit more, would it be more sensible to simply skip presenting the buffer if it hasn't changed? Something like so ( I couldn't quite figure out how else to compare two DrmDumbBuffers ) :

Hey
I've been looking at this and wanted to make sure if I understood this correctly before writing a patch.

Ideally SDDM should grow support for a wayland based greeter instead of a X11 greeter. This would involve looking at what XorgDisplayServer does and why the Greeter needs it.

/etc/profile.d gets evaluated by bash but not simplelogin. David needs to look into this to get it working.

Why not simply mount the device to the relevant folder instead of having another level of indirection ?

You also changed it in the jenkins master cookbook, which might or might not be relevant for other jenkins masters

This looks inconsistent, why is one a link and the other a folder?

Why not just define these requirements in the Gemfile?

I would argue that breeze-gtk should already have had this change in order to not break on systems where GTK is >= 3.20 and that introducing a new build-dep is the only way of fixing thee issue.

Simplify setting up the variable

Don't cache string, set a CMAKE_MODULE_PATH too

In D4596#85993, @lbeltrame wrote:

FTR, the reason this was not done at CMake time was to avoid depending on a development package (and associated dependencies like a compiler) to install something that's architecture-dependent. My first. local patch I made at the time used pkg_config, but it was discarded for this reason.

Not that I have hard feelings about this, should this be changed, but I just wanted to outline the rationale for it at the time.

Make finding GTK required, move find_package calls together

This should ideally also be backported for Plasma 5.9

Last I checked we explicitly agreed no git on slaves. So I'm not favor of this till I get more info about why we're changing this requirement.

Update

Update as per comments

Ah cool, missed that.

Perhaps mobster needs to be fixed before this change lands ?