16:59 <Wizzup> enunes: I believe our latest image should have a fix for the virtual keyboard problem you were seeing -- the one that freemangordon could reproduce (he now reports it is fixed in this image). https://maedevu.maemo.org/images-devel/pinephone/20220119/

17:32 <enunes> Wizzup: ok, I'll give it a try then

21:18 <enunes> Wizzup: freemangordon: question, does this latest image attempt to set EGL_BUFFER_PRESERVED or not?

21:26 <freemangordon> pfew, finally identified myself to the nickserv :)

21:27 <freemangordon> enunes: no, however we tried to set that to no use

21:27 <enunes> yeah, so heres the thing

21:27 <enunes> https://www.khronos.org/registry/EGL/sdk/docs/man/html/eglChooseConfig.xhtml EGL_SWAP_BEHAVIOR_PRESERVED_BIT

21:27 <enunes> mesa does not support that at all currently, so setting it wont have any effect, in fact it will return an egl error

21:28 <freemangordon> mhm

21:29 <enunes> I came up with a hack that "adds support" for it, and then I can see results with a simple app drawing a couple of triangles using it

21:29 <enunes> that could make your application work, since it seems to rely on preserved buffers

21:29 <freemangordon> yes, it does

21:30 <freemangordon> is this a mesa patch?

21:30 <enunes> however, as we mentioned before this is not a lima bug at all, its just a mesa infrastructure patch

21:30 <freemangordon> I see

21:30 <freemangordon> Wizzup: ^^^

21:31 <enunes> I would suggest if you can test with it and it fixes your app, please dont use the patch anyway and fix the application

21:31 <freemangordon> fix the application like?

21:31 <enunes> https://paste.centos.org/view/raw/4bde4e15

21:32 <enunes> fix the application to not require preserved buffers as mesa does not support it

21:32 <enunes> I'm not going to pursue merging something like this myself

21:32 <enunes> I'd suggest to open a more general feature request to mesa then

21:34 <freemangordon> enunes: do I get it right that EGL_SWAP_BEHAVIOR_PRESERVED_BIT is something that's mandated by the specs?

21:36 <enunes> no, it is something that can be either supported or not, and if you plan to use EGL_BUFFER_PRESERVED then you should add EGL_SWAP_BEHAVIOR_PRESERVED_BIT to eglChooseConfig so that you get a config that will support that

21:36 <freemangordon> I see

21:37 <enunes> mesa basically does not set that at all today for dri3 configs, but it seems that some implementation is there, so just forcing it to be set, it kind of works

21:37 <enunes> I'm not all that much interested in validating that it does indeed work well enough to actually enable it in mesa

21:38 <enunes> it does seem to work well enough for my triangle app anyway

21:41 <freemangordon> ok, we'll try the patch, thanks

21:41 <freemangordon> I guess it makes sense to see if we can implement "buffer age" in clutter

21:41 <Wizzup> enunes: thanks for the help, we'll try the patch and look at fixing clutter, but also file a more general mesa bug/request I think :)

21:51 <enunes> Wizzup: freemangordon: for reference https://github.com/enunes/mesa-test-programs/tree/buffer-preserved https://github.com/enunes/mesa-test-programs/commit/47fa0ce4ce430a566e37d28017bb9302f6b5dff9

21:55 <enunes> and remember that if you care about efficiency and e.g. battery life etc at all for the pinephone, please dont use this feature at all :)

21:55 <freemangordon> thanks a lot !

21:56 <freemangordon> enunes: you say it is better to always redraw?

21:57 <enunes> probably best case buffer_age, partial_update, since that avoids back copying the entire frame every time

21:59 <freemangordon> I see. well, I don't think it is that bad (with EGL_BUFFER_PRESERVED) but not sure we can do some proper mesurements

22:02 <enunes> https://community.arm.com/arm-community-blogs/b/graphics-gaming-and-vr-blog/posts/mali-performance-3-is-egl_5f00_buffer_5f00_preserved-a-good-thing is a good overview

22:09 <freemangordon> given that hildon-desktop does heavy transparent compositing, I think I know what our case is :)

22:10 <enunes> I just think it would not be so special. this has been disabled for 10? years and probably nobody complained, I dont see any test cases for it either, its just not a popular thing to do

22:11 <freemangordon> well, I don;t think there were any users on mobile linux either

22:12 <freemangordon> but lets not go into that

22:15 <freemangordon> anyway, appreciate your help. it is late here, going to have some sleep. night!