10:20 karolherbst: is there a great way to check for the existence of a C++ member function with meson?
10:23 karolherbst: guess I have to use compiles, because has_function doesn't seem to work for C++ member functions
10:24 mceier: maybe has_member works also for C++
10:24 karolherbst: mhhhh...
10:24 karolherbst: worth a try
10:25 karolherbst: error: invalid use of non-static member function :')
10:25 mceier: ok
10:26 karolherbst: one hack with has_function would be to do a `using namespace::class:member_function;` thing, but I think that's illegal
10:33 mceier: how did you try has_member? meson tries to compile `void bar() { {typename} foo; foo.{member}; };` so you should use `function_name(args)` as a member instead of `function_name`
10:33 karolherbst: mhhhhhh
10:34 karolherbst: I guess?
10:34 karolherbst: let's see...
10:35 karolherbst: mceier: that seems to work
10:35 mceier: cool :)
10:35 karolherbst: I wonder what meson devs think of that :D
11:52 karolherbst: I have a SPIR-V question. If you have a three component vector variable in the Input storage class, what's the expected behavior on out of bound reads?
11:52 karolherbst: UB?
11:52 karolherbst: defined as per environment spec?
12:37 zmike: env spec yes
14:22 sima: jfalempe, for the fancy qr code, maybe a text hint which qr reader will work would be good on the panic screen?
14:23 sima: high-res qr code with full dmesg is great, but we need to make sure people can read it :-)
14:23 karolherbst: put in a second QR code to download the qr code reader :P
14:23 sima: qr codes all the way down
14:24 sima: should make the ascii art tux a qr code too with a link to full res one :-P
14:24 karolherbst: anyway, free qr code reader is such a big bikeshed, because try to find one which you can generally advise to use, which isn't a huge piracy violation like most of them :P
14:25 karolherbst: *privacy
14:25 sima: I thought the bigger issue was that default QR codes only encode like 200 bytes or so
14:26 sima: and the high-res ones we need to put a good chunk of dmesg in it are a bit unusual
14:27 emersion: we don't upload the dmesg to a pastebin yet?
14:27 emersion: smh
14:28 sima: if you can do that reliably from nmi context, I'll retire
14:29 karolherbst: if network is down, just use the sound card to transmit data wireless, and if that doesn't work, there is probably also a CPU noise based exploit in the wild somewhere
14:29 karolherbst: just needs an app to recieve the data
14:30 sima: the problem with panic context is you get either lockless (as in try lockless, no retrying allowed) or raw_spin_trylock to synchronize with anything else
14:31 sima: stuff like lockless list walks with the nmi safe rcu variant for updates
14:31 sima: everything else is somewhere between undefined and "probably on fire"
14:32 karolherbst: the proper solution to all those issues is of course, that we just compress the data with xz and then have our "DRM blue screen QR reader" app on the stores (1.4/5 stars best rated comment: "DRM is cancer")
14:33 sima: I see the headline already
14:33 sima: xz embedded payload to exploit the kernel through the drm panic handler
14:34 karolherbst: owning an android device through a malicious drm panic handler QR code would indeed be impressive
14:36 karolherbst: looks like the scanner app I have can read those version 40 QR codes
14:36 karolherbst: soooo
14:36 karolherbst: I think we are fine
14:37 sima: oh I more meant a repeat of xz-utils vs openssh, except this time the kernel
14:37 karolherbst: heh
14:42 sima: agd5f, no -fixes pull this week or just not yet?
14:42 karolherbst: anyway, the one I'm using is this one: https://gitlab.com/Atharok/BarcodeScanner
14:43 sima: mripard, drm-misc-next isn't in linux-next, which is a bit bad at -rc6 :-/
14:43 sima: also I guess explains why alice couldn't find the baseline for the drm-panic series from jfalempe
15:01 karolherbst: do we have anything ready for formatless image store lowering?
15:03 alyssa: karolherbst: don't think so since each hardware has different special cases they can do.. I think
15:04 karolherbst: yeah.... I'm just mostly curious why every other driver I wired up rusticl, doesn't have the issue and v3d is the first one...
15:05 alyssa: because adult hardware just fetches the format from the descriptor and does the hardware
15:05 karolherbst: anyway, v3d seems like it needs the data properly packed, which is pretty annoying
15:05 alyssa: ^does the fmt conversion in hw
15:05 karolherbst: yeah...
15:05 karolherbst: though nvidia didn't for a long time
15:09 karolherbst: though looks like intel has some impressive lowering there
15:16 mripard: sima: what?! we fixed it weeks ago
20:57 sima: mripard, apparently it rebroke
20:57 sima: maybe we should add a check to dim that rejects force-pushes for the linux-next branches?
20:57 sima: or asks, with a hint that unless you're a maintainer you probably misplaced a patch
20:57 sima: maybe this would catch this stuff ...