13:18 imirkin: sigod: significant firmware issues only exist for some GTX660 boards
13:18 imirkin: sigod: driver issues exist for all GPUs, of course
13:19 karolherbst: mhh
13:19 imirkin: it's unlikely for the GTX660 issues to ever be fixed as the only person capable of performing those fixes has never been able to lay his hands on a problem board
13:19 karolherbst: that one spirv test still fails
13:19 imirkin: (and i'm not 100% sure he would spend the time to fix them were he to obtain one now)
13:20 karolherbst: mhh.. KHR-GL45.gl_spirv.spirv_validation_capabilities_test
13:20 karolherbst: that's a CTS bug though for sure
13:20 imirkin: probably yeah
13:20 karolherbst: "glGetError() returned GL_INVALID_OPERATION"
13:21 imirkin: probably doesn't check for some ext?
13:21 imirkin: are you advertising some spirv exts?
13:21 karolherbst: the other spirv tests are passing
13:21 karolherbst: or well
13:21 karolherbst: notsupported
13:21 karolherbst: :)
13:21 imirkin: right, probably missing some silly check
13:21 imirkin: they're pretty open to taking patches
13:21 imirkin: although as a khr member, you have to go through some internal thing afaik
13:25 karolherbst: a "spirvUtils::checkGlSpirvSupported(m_context);" is missing :)
13:25 imirkin: yeah, that's common
13:25 imirkin: most of the fixes to get things working with GL33 were to such checks
13:26 karolherbst: nice, now that test also reports Not supported :)
13:26 karolherbst: I should check the gtf as well at least once
13:29 imirkin: was that the only fail in the CTS series?
13:29 karolherbst: yeah
13:30 imirkin: cool. must be a new test, or not in mustpass or something
13:30 imirkin: all the GTF tests should pass too
13:30 karolherbst: it's there, but I also run 4.6.1 :p
13:30 karolherbst: I am on the opengl-es-cts- tag
13:30 imirkin: oh, i don't think that's right
13:30 karolherbst: I think it is though
13:30 imirkin: i think you want the opengl-cts-4.6.0 branch
13:31 imirkin: at least that's what i was testing
13:31 karolherbst: I think the GLES and GL stuff got merged
13:31 karolherbst: the release notes explciitly state GL fixes in the gles tags
13:31 karolherbst: also, origin/opengl-cts-4.6.0 is from feb 2018
13:31 karolherbst: ;)
13:32 imirkin: https://github.com/KhronosGroup/VK-GL-CTS/commits/opengl-cts-4.6.0
13:32 imirkin: commits from Apr 13, 2020
13:32 karolherbst: heh...
13:32 imirkin: i think you're looking at the original date of the commit that was cherry-picked
13:32 karolherbst: yeah.. I guess so
13:32 karolherbst: interesting
13:32 imirkin: i need to send a PR to integrate my glx patch too ...
13:33 karolherbst: gtf: Failed: 1/1742 (0.1%) :/
13:33 karolherbst: GTF-GL45.gtf30.GL3Tests.sgis_texture_lod.sgis_texture_lod_basic_getter
13:34 imirkin: uhm
13:34 imirkin:goes to check
13:35 karolherbst: ERROR/sgis_texture_lod_basic_getter: ERROR:sgis_texture_lod_basic_getter:GL_TEXTURE_MAX_LOD, 2147483647 != 0
13:36 imirkin: uh crap. something changed.
13:36 karolherbst: fails for you as well?
13:36 imirkin: i can't get my old glcts build to run at all
13:36 karolherbst: :/
13:36 imirkin: X operation 152:34 failed: GLXBadFBConfig
13:36 karolherbst: ehh..
13:37 karolherbst: imirkin: enable 4.5
13:37 karolherbst: :p
13:37 karolherbst: don't ask how much I debugging this shit already
13:37 imirkin: OH!
13:37 imirkin: yes, thanks
13:37 imirkin: Passed: 1/1 (100.0%)
13:37 imirkin: on mesa 20.0.5
13:37 karolherbst: heh
13:37 karolherbst: okay
13:37 karolherbst: let me check
13:38 imirkin: fails on master
13:38 imirkin: super.
13:38 karolherbst: cool
13:39 karolherbst: I don't udnerstand why the CTS always returns 0 :/
13:39 karolherbst: would be soooo helpful if it would return !0 on error
13:40 karolherbst: at least bisecting mesa isn't as painful as bisecting the kernel :)
13:41 imirkin: hope it wasn't my changes that broke it
13:41 imirkin: i'm finally going to try to finish the gs passthrough thing, before the 20.1 branchpoint
13:42 imirkin: we'll see how it goes
13:42 karolherbst: cool
13:42 karolherbst: also try to take a look at the shader cache stuff once you find some time for that :p
13:43 karolherbst: I am actually concerned about the nv50 changes there as.. well.. nv50 is different
13:46 imirkin: gr, yes
13:46 imirkin: i keep promising to do that.
13:46 imirkin: are the latest patches on list?
13:46 imirkin: or somewhere else?
13:46 karolherbst: there is an MR at least where I left some comments, but I think the latest patches on the MR are the same: https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4264
13:48 imirkin: k
13:49 imirkin: want to make some progress on the passthrough gs thing though
13:49 imirkin: i feel like i might already be close, but i also expect there will be random unexpected issues
13:49 imirkin: in theory it's just "copy outputs, add some additional error checking, done"
13:49 karolherbst: well, it's up to you anyway :p
13:49 imirkin: well, i do try to be good about doing reviews. but i don't always succeed.
13:49 imirkin: it's a problem at work too :)
13:50 karolherbst: imirkin: I left a comment for you though, which is quite high level
13:50 imirkin: k
13:50 imirkin: i'll have a gander
13:50 imirkin: or is it a goose...
13:53 RSpliet: nah, it's a propah gandah!
13:56 imirkin: =]
13:56 imirkin: glad _someone_ got it
14:00 karolherbst: imirkin: 53c36dfcfe3eb3749a53267f054870280afb0d71 broke it
14:00 imirkin: lol
14:00 karolherbst: yeah.... double checking
14:00 karolherbst: but yeah
14:00 karolherbst: yep
14:00 karolherbst: i965 also affected
14:00 karolherbst: no surprise though
14:01 imirkin: makes sense, yeah
14:01 imirkin: i think the whole int/long thing is what kills it
14:01 karolherbst: probably
14:02 imirkin: but i'd have to analyze more. i'll grudgingly allow you that privilege :)
14:02 imirkin: could also be the test that's bogus
14:02 karolherbst: yeah.. dunno
14:07 karolherbst: imirkin: 1.09951163e+12 is the value
14:07 imirkin: which feels > maxint
14:07 karolherbst: which... is nowhere close to 0 :)
14:07 karolherbst: yeah
14:07 imirkin: could be test breakage
14:07 imirkin: or i dunno
14:07 imirkin: again, analysis required :)
14:08 karolherbst: yeah.. but lround(1.09951163e+12) == 0?
14:08 karolherbst: *lroundf
14:08 karolherbst: heh...
14:08 imirkin: i haven't the faintest clue
14:08 karolherbst: ohhh
14:08 karolherbst: :O
14:09 karolherbst: what the heck
14:09 imirkin: i expect it's maxint
14:09 karolherbst: that's evil
14:09 karolherbst: p (long int)lroundf(obj->Sampler.MaxLod)
14:09 karolherbst: 1099511627776
14:09 imirkin: but it's an int
14:09 karolherbst: 0x10000000000
14:09 imirkin: right?
14:09 imirkin: ;)
14:09 karolherbst: how evil is this...
14:09 karolherbst: :D
14:09 imirkin: so we need an extra int clamp
14:09 imirkin: or look for a iclampf
14:10 imirkin: or revert
14:10 karolherbst: well.. the result type is GLint
14:11 karolherbst: mhhh
14:11 imirkin: GLint or GLuint?
14:11 karolherbst: GLint
14:12 imirkin: hm
14:13 karolherbst: yeah well.. it's also a monster MR
14:13 karolherbst: I let dcbaker deal with it...
14:14 mixfix411: who's dcbaker
14:15 mixfix411: oh i see hes a coder
14:16 imirkin: the author of the change in question
14:17 karolherbst: not even the MR can be cleanly reverted :/
14:18 karolherbst: maybe I didn't mess up the revert.. let's see
14:18 karolherbst: ahh, I did :)
14:31 karolherbst: Passed: 1742/1742 (100.0%)
14:32 mixfix411: good ole mesa build :)
14:32 karolherbst: soo... let's see what the real CTS does
14:37 karolherbst: imirkin: ..... remember those out of mem errors users had with glamor? Wanna bet it's the same bug I fixed today?
14:37 imirkin: could be!
14:43 karolherbst: although it sounded like it was more likely on older GPUs or GPUs with less VRAM, but maybe with a weaker CPU the race is more likely...
14:43 karolherbst: I was never able to hit that bug myself
15:02 karolherbst: imirkin: do you still have the logs from an old full cts run?
15:02 imirkin: everything passed, until it didn't...
15:03 imirkin: in the same spot each time
15:03 imirkin: i don't think i have those logs anymore though
15:03 karolherbst: yeah, just wondering what test it was and if you still have the qpa files from that
15:03 karolherbst: mhh :/
15:03 imirkin: qpa files don't really matter
15:03 imirkin: the failure was a hang
15:03 karolherbst: ohh.. interesting
15:04 imirkin: or maybe not. but it was an OOB PTE error
15:04 imirkin: or something
15:44 sigod: fyi my card it a gigabyte gtx 660
15:45 sigod: is*
15:45 sigod: thanks for the info