19:28 enunes: sarbes: cool, did you get that from GLES 1 or from guessing at the hardware description?
19:30 enunes: anarsoul: do you have thoughts on setting the mixed_framebuffer_sizes cap to enable GL_ARB_framebuffer_object (in addition to EXT_framebuffer_object)? I feel like this was already considered before?
19:30 anarsoul: enunes: it's not supported by the hardware
19:31 enunes: perhaps we were too strict if there is some corner case, but lima already implements all the "multiple disjoint extensions into a single ARB extension" listed in https://registry.khronos.org/OpenGL/extensions/ARB/ARB_framebuffer_object.txt
19:31 anarsoul: i.e. color and z buffers cannot have different dimensions
19:31 enunes: it prevents hundreds of piglit tests from running based on saying that glBlitFramebuffer is not supported
19:31 enunes: lima is the only driver which does not set it, vc4 does with the reason to enable those piglit tests
19:32 enunes: and the GL 2.1 support is kind of a lie anyway, so... maybe we don't have to be that strict for that particular case
19:32 anarsoul: it is supported in vc4 though
19:33 anarsoul: we'd have to do 2x draws (if color widthxheight doesn't match z widthxheight) if we want to support it
19:35 anarsoul: i.e. utgard cannot do that:
19:35 anarsoul: - Where appropriate, relax some of the constraints expressed by previous
19:35 anarsoul: FBO-related extensions. In particular the requirement of matching
19:35 anarsoul: attachment dimensions and component sizes has been relaxed, to allow
19:35 anarsoul: implementations the freedom to support more flexible usages where
19:35 anarsoul: possible.
19:35 enunes: https://pacific.enunes.eu/user/mesa/-/commit/2b6711cc5fbb06620f46d3a42823bf03b58e9e34
19:36 enunes: oops, thats my local gitlab
19:36 enunes: https://gitlab.freedesktop.org/mesa/mesa/-/commit/2b6711cc5fbb06620f46d3a42823bf03b58e9e34
19:37 enunes: I meant it more like this, so that we can allow all those piglit tests to run
19:38 anarsoul: well, sure
19:39 enunes: I worked on the srgb stuff and when trying to test it, most tests don't actually run because of this
19:42 enunes: on my run thats around 400 tests
19:43 enunes: not sure how many actually test the multiple sizes part though
19:45 sarbes: It had to be implemented in hardware, or so I thought. And the "hole" in the blend func was pretty telling.
19:46 sarbes: Logicops, that is.
19:53 enunes: sarbes: yeah, nice. so just a hint don't know if you had considered it, but this is also in GLES 1 which is also implemented by the blob driver, GLES 1 is a bit of a forgotten source of info for these older features
19:54 sarbes: Are there more GLES 1 features missing? I'm not very knowledgeable in this area.
19:57 enunes: one time I started this https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7825 , apparently it's been 4 years... probably wont realistically finish that myself any time soon, so if anyone is ever interested I guess it would be free to get :)
19:58 enunes: this one might be more relevant though being useful for more real applications https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16136 , same situation