00:00fdobridge: <marysaka> I was skipping everything that was after a GPFIFO entry that have the sync flag
00:01fdobridge: <marysaka> Pushed some fixes
00:02fdobridge: <marysaka> for the random sigbus I guess another thread might be unmapping the region while I'm trying to read it...
00:03fdobridge: <marysaka> meaning that me setting up the page as read only isn't enough and might actually still trigger the actual write on the kernel side...
00:03fdobridge: <karolherbst🐧🦀> do it like a pro, pause all threads 🙃
00:04fdobridge: <marysaka> I will still be racing at that moment tho :vReiAgony:
00:05fdobridge: <marysaka> that or events are broken (quite possible I forgot to remap the fds for thoses)
00:06fdobridge: <karolherbst🐧🦀> mhhh.. right
00:07fdobridge: <marysaka> if it's not events then the other solution is just to map a dummy page there, and present the right page when we want the write to pass I guess
00:11fdobridge: <karolherbst🐧🦀> yeah.. could also just let all threads run into the dummy page and then serialize handling them
00:19fdobridge: <gfxstrand> Okay, I have a lot more to sort through now. 😅
00:24fdobridge: <gfxstrand> Pass!
00:24fdobridge: <gfxstrand> Okay, Now I'm making progress!
00:25fdobridge: <gfxstrand> I don't know why I need to set those bits that way but now that I have the tools to see what NVIDIA is doing I can maybe finally figure out how to configure viewport clipping properly
00:25fdobridge: <gfxstrand> apparently, I needed to set some stuff to CLAMP?!?
00:28fdobridge: <gfxstrand> So weird...
00:42fdobridge: <gfxstrand> Okay, I get it now
00:42fdobridge: <gfxstrand> I think
00:42fdobridge: <gfxstrand> I guess I'll do a full CTS run while I pack for XDC
00:57fdobridge: <gfxstrand> https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25723
01:13fdobridge: <gfxstrand> Fedora no longer has 32-bit image so I can't use my backup plan of building 32-bit in a VM. 😭
01:28fdobridge: <airlied> you could build it from fedora packaging at a guess
01:38fdobridge: <gfxstrand> Yeah, probably? There's a tool for that, isn't there? Probably enough for setting up a chroot. 🤔
02:02fdobridge: <gfxstrand> Looks like mock should do the trick
03:23fdobridge: <gfxstrand> It's a bit clunky but it'll work, I think.
03:23fdobridge: <gfxstrand> It really wants to build srpms, not stuff from git
03:33fdobridge: <gfxstrand> Okay, have a working i686 build!
03:33fdobridge: <gfxstrand> Once x86_64 runs, I'll run i686
03:46fdobridge: <airlied> Yes mock, looks like you found it, yeah all based on rpms, though you can get a mock shell
04:25fdobridge: <gfxstrand> Yeah, just took me a bit to find all the things
04:25fdobridge: <gfxstrand> It's not bad once you figure it out.
04:26fdobridge: <gfxstrand> ```
04:26fdobridge: <gfxstrand> Test run totals:
04:26fdobridge: <gfxstrand> Passed: 401810/3988868 (10.1%)
04:26fdobridge: <gfxstrand> Failed: 0/3988868 (0.0%)
04:26fdobridge: <gfxstrand> Not supported: 3587054/3988868 (89.9%)
04:26fdobridge: <gfxstrand> Warnings: 4/3988868 (0.0%)
04:26fdobridge: <gfxstrand> Waived: 0/3988868 (0.0%)
04:26fdobridge: <gfxstrand> ```
04:28fdobridge: <gfxstrand> 32-bit is running
04:28fdobridge: <gfxstrand> It may OOM because we're leaking shaders. 😳
04:28fdobridge: <gfxstrand> I can fix that. I just haven't bothered too yet and it's annoying while we have NAK and codegen both wired up
04:32fdobridge: <gfxstrand> Hrm... OOMing... Maybe we're leaking a map somewhere? 🤔
04:34fdobridge: <gfxstrand> Yeah, I have some leaks to fix
05:25fdobridge: <gfxstrand> Wasn't cleaning up the disk cache. Oops.... That probably does lots of mmap(). 🙃
15:09fdobridge: <gfxstrand> Fixed unbind of images (which was apparently only tested by WSI tests. WTF?!?)
15:10fdobridge: <gfxstrand> Fixed prime image allocation with max-size on NVK (32-bit rollover).
15:10fdobridge: <gfxstrand> Let's try again!
16:55karolherbst: who is actually coming to XDC? airlied, dakr? We might be enough people that it makes sense to organize some dinner together or something
17:04dakr: karolherbst: Yep, I'll arrive tomorrow at ~5pm.
17:05karolherbst: cool
17:37fdobridge: <gfxstrand> I'm landing Sunday afternoon
17:37fdobridge: <gfxstrand> About to go into airplane mode for my first flight
17:49fdobridge: <karolherbst🐧🦀> I should have booked my flight for Sunday, but oh well...
17:59dakr: karolherbst: My bad, I meant Monday,
17:59dakr: ~5pm.
17:59karolherbst: ahh
17:59karolherbst: I'll arrive at 9pm.. not sure if I make it to the arrival thing, but maybe I will
18:00dakr: The sightseeing walk? Or is there something else I didn't note.
18:01karolherbst: sightseeing is on Tuesday
18:01karolherbst: there is a welcoming party
18:01karolherbst: ehhh more like a gathering
18:01karolherbst: https://indico.freedesktop.org/event/4/sessions/28/#20231016
18:02karolherbst: ehh.. maybe the overview page has more details on that "There will be a welcome event on the evening of the day before the conference starts, October 16th, from 7:30PM to 9:30PM local time. It will take place at the Palexco conference center (see Venue below), and will be a good moment to pick up badges, masks, tests and meet other attendees informally. The event will include some food and drinks courtesy of Igalia."
18:02dakr: Huh, that's a bit subtile, didn't note this one.
18:03dakr: We can plan a dinner on one of the other days though.
18:04karolherbst: yeah.. I'm thinking of wednesday, but I have no idea when or if a RH dinner is planned :D
19:06fdobridge: <airlied> Yeah no for me, A Corun a is nightmare travel due to transfers
19:32fdobridge: <gfxstrand> I mean, it's a nightmare for the rest of us, too. 😝
19:33fdobridge: <gfxstrand> Unless you're on the east side of London and can fly out of Ludwig...
19:33fdobridge: <gfxstrand> Unless you're on the east side of London and can fly out of Gatwig... (edited)
19:33fdobridge: <gfxstrand> Unless you're on the east side of London and can fly out of Gatwick... (edited)
19:36fdobridge: <gfxstrand> Least time we had XDC there, they had a flight to/from Heathrow which is still kinda terrible but at least Heathrow is an airport you can get to.
19:36fdobridge: <karolherbst🐧🦀> yeah, it's terrible for everybody
19:37fdobridge: <karolherbst🐧🦀> for me it was easier to find sane flights to tokyo than to a crouna
20:16fdobridge: <airlied> Maybe suggest xdc Tokyo next year :-p
20:18fdobridge: <Ella> Hmmm train to Tokyo for me... that could be fun
20:24fdobridge: <gfxstrand> Is there any way to increase the nouveau channel kill timeout?
20:50fdobridge: <gfxstrand> Here's hoping that if I run from a disconnected screen session, it won't be fighting with GNOME rendering the text scroll and that will be enough that we'll actually pass this time. 😅
20:51fdobridge: <gfxstrand> Gotta love CTS problems that get solved by "run the GPU at speed". 😂
22:01fdobridge: <karolherbst🐧🦀> yeah soo.. we actually have kernel bugs like that 🙃
22:34fdobridge: <gfxstrand> Ugh... I had so many good runs and now I can't get one. 🙄
22:34fdobridge: <gfxstrand> I'm gonna reboot the machine and try again, I guess.