<Sachiel>
old enough to drink, at least outside the US
ddavenport has joined #dri-devel
<HdkR>
Old enough for cigarettes at least. That network is going to need one after what it goes through :P
sdutt has quit [Remote host closed the connection]
thellstrom1 has joined #dri-devel
thellstrom has quit [Remote host closed the connection]
pnowack has quit [Quit: pnowack]
pnowack has joined #dri-devel
Guest2154 is now known as gruetzkopf
pnowack has quit [Quit: pnowack]
pnowack has joined #dri-devel
pnowack has quit []
pnowack has joined #dri-devel
pnowack has quit [Read error: Connection reset by peer]
pnowack has joined #dri-devel
thellstrom has joined #dri-devel
thellstrom has quit [Quit: thellstrom]
thellstrom1 has quit [Ping timeout: 480 seconds]
thellstrom has joined #dri-devel
pnowack has quit [Quit: pnowack]
pnowack has joined #dri-devel
pnowack has quit []
pnowack has joined #dri-devel
pnowack has quit [Remote host closed the connection]
pnowack has joined #dri-devel
nirbheek has joined #dri-devel
nirbheek has left #dri-devel [User left]
rasterman has joined #dri-devel
ddavenport has quit [Remote host closed the connection]
pochu has joined #dri-devel
rgallaispou has joined #dri-devel
camus has quit [Remote host closed the connection]
camus has joined #dri-devel
lynxeye has joined #dri-devel
pnowack has quit [Quit: pnowack]
pnowack has joined #dri-devel
MrCooper has quit [Quit: Leaving]
MrCooper has joined #dri-devel
naveenk2 has joined #dri-devel
ella-0 has joined #dri-devel
ella-0 has quit []
ella-0 has joined #dri-devel
ella-0 has quit []
pcercuei has joined #dri-devel
vivijim has joined #dri-devel
robert_mader has joined #dri-devel
Lucretia has quit []
Lucretia has joined #dri-devel
imirkin has quit [Ping timeout: 480 seconds]
cedric has joined #dri-devel
Ekkie has joined #dri-devel
Ekkie has quit [Remote host closed the connection]
bluebugs has quit [Ping timeout: 480 seconds]
imirkin has joined #dri-devel
kokotchY has joined #dri-devel
kokotchY has quit [autokilled: spambot. Dont mail support@oftc.net with questions (2021-06-15 09:46:22)]
bjorn3_gh has joined #dri-devel
<bjorn3_gh>
An open source game I sometimes play recently switched from OpenGL to Vulkan+SPIRV. It now crashes reproducably with a null pointer dereference inside mesa during the shader compilation. I did some debugging and it seems that a switch has no cases: https://gist.github.com/bjorn3/b9ad74fa1165e00e7bc3b2af078f2bd0
<neobrain[m]>
bjorn3_gh: probably best to report an issue on gitlab about it, ideally with a fossilize capture (https://github.com/ValveSoftware/Fossilize) in this case. Though if the game just recently switched, chances are it's a game bug
<bjorn3_gh>
It recently switched, but the wgpu-rs library it uses should validate all spirv before attempting to pass it to the graphics driver.
<neobrain[m]>
ah, fair
<bjorn3_gh>
I will try to get a capture with fossilize.
<bjorn3_gh>
I am using debian, so my mesa version is relatively old. I don't know if it is already fixed in newer versions. Is that a problem?
<neobrain[m]>
if there's a convenient way for you to update libvulkan_intel it would be great if you could test that first, though otherwise it's simple enough for a dev to confirm latest Mesa has the same issue if the fossilize capture is provided
rasterman has quit []
rasterman has joined #dri-devel
naveenk2 has quit []
dllud_ has joined #dri-devel
dllud has quit [Read error: Connection reset by peer]
dllud has joined #dri-devel
dllud_ has quit [Remote host closed the connection]
NiksDev has quit [Ping timeout: 480 seconds]
naveenk2 has joined #dri-devel
madroach has joined #dri-devel
madroach has quit [Remote host closed the connection]
<naveenk2>
Hi Imirkin: Regarding fps measurement of GPUs using glxgears app, I see a GUI hang when run with vblank_mode=0 glxgears with DRI_PRIME=1
<naveenk2>
However it works fine with DRI_PRIME=0 i.e. with igpu. So issue is seen only when using dgpu.
<naveenk2>
Though simply glxgears app works fine with dgpu without providing vblank_mode=0. Max fps hits to 60 which is aligning with Display fps. I want to understand more on what is limiting dgpu when vblank_mode is set to 0
jperras has joined #dri-devel
jperras has quit [Remote host closed the connection]
robertmader[m] has joined #dri-devel
robert_mader has quit [Quit: Leaving.]
khfeng has quit [Remote host closed the connection]
khfeng has joined #dri-devel
go4godvin has joined #dri-devel
johnvikert6 has joined #dri-devel
johnvikert6 has quit [Remote host closed the connection]
camus1 has joined #dri-devel
camus has quit [Remote host closed the connection]
pjakobsson_ has joined #dri-devel
jernej has joined #dri-devel
jernej has quit []
jernej has joined #dri-devel
naveenk2 has quit [Ping timeout: 480 seconds]
ofourdan has joined #dri-devel
sdutt has joined #dri-devel
shankaru has quit []
sdutt has quit []
sdutt has joined #dri-devel
Sumera[m] has quit []
Sumera[m] has joined #dri-devel
Sumera[m] has quit []
Sumera[m] has joined #dri-devel
s-h-i-n-o-b-i22 has joined #dri-devel
s-h-i-n-o-b-i22 has quit [Remote host closed the connection]
itoral has quit [Remote host closed the connection]
<shfil>
hello there, not sure if it's enough for an issue, but iirc on gitlab you can enable a tab for wiki and point it to external address (like https://docs.mesa3d.org/)
jernej has joined #dri-devel
Danct12 has quit [Quit: Quitting]
Danct12 has joined #dri-devel
Danct12 has quit []
Danct12 has joined #dri-devel
fearnothing has joined #dri-devel
fearnothing has quit [autokilled: Suspected spammer. Mail support@oftc.net with questions (2021-06-15 14:44:58)]
<jekstrand>
kusma: Looking at sphinx C autodoc stuff... It seems breathe is definitely the most mature of the options.
<jekstrand>
But they've not put a lot of effort into ease-of-use. :-(
<kusma>
jekstrand: fair :/
<jekstrand>
I think I can make a sphinx plugin which runs doxygen for us if needed.
<jekstrand>
I've made sphinx plugins before.
<kusma>
Everything Doxygen just makes me sad
<jekstrand>
But I really don't like the weird mix of markdown and RST that using doxygen implies
<jekstrand>
kusma: Yeah, I don't like doxygen at all either
<kusma>
jekstrand: Yeah, I made a few as well... But maybe we should just stuff it in the CI scripts and document it as well?
luzipher__ has joined #dri-devel
<jekstrand>
Maybe? I don't like the way that you have to have fixed paths for stuff
<kusma>
jekstrand: The other option is to just open-code the definitions using C domain roles, but that seems likely to go out of date...
<kusma>
"just"
xaret has joined #dri-devel
xaret has quit [Remote host closed the connection]
<jekstrand>
Ok, sphinx-c-autodoc is pretty terrible
<daniels>
it's so bad that we spent a genuinely surprising amount of time exhausting every other alternative, before concluding that they were all somehow worse
<daniels>
I mean I know developers love posturing about how everything we didn't write is terrible, but my thoughts on C documentation are totally unpublishable, even with the usual allowance for en_AU
<jekstrand>
daniels: Yeah...
<jekstrand>
daniels: I'm happy to copy+paste that
<jekstrand>
Here's the the thing about sphinx-c-autodoc that pushed me over the edge: It doesn't handle new lines in comments properly and so you can't write a comment with multiple paragraphs.
<jekstrand>
If that's busted, what else is?
<daniels>
...
<daniels>
you can also check out the Weston issues/MRs where we went through all the alternatives if you want to read reams and reams and reams of prose
<daniels>
or not if you don't!
<jekstrand>
Nah, I believe you
<jekstrand>
I'll just copy+paste what Weston does
Danct12 has quit [Quit: Quitting]
<jekstrand>
daniels: Does weston work with readthedocs?
<daniels>
jekstrand: no, we just generate in GitLab CI and serve from Pages
<daniels>
we could've uploaded to RTD as well but didn't quite see why
<jekstrand>
fair
<kusma>
jekstrand: FWIW, we don't use readthedocs for mesa either (any more ish)
<alyssa>
we don't?
<jekstrand>
Ok then
<kusma>
docs.mesa3d.org is just gitlab pages
mattrope has joined #dri-devel
<jekstrand>
daniels: Was there a particular reason for using meson vs. just having a script to invoke sphinx?
<alyssa>
ah
<jekstrand>
I don't necessarily mind having all the files listed in meson.build files, it just seens unnecessary
bjorn3_gh has left #dri-devel [#dri-devel]
Duke`` has joined #dri-devel
<danvet>
jekstrand, daniels looked at sphinx + kerneldoc?
<jekstrand>
danvet: I looked at it yesterday. I'm not sure how it works. Lots of perl
<danvet>
well maybe the reimplemtation jani had lying around, since kerneldoc relies on some horror perl
<danvet>
jekstrand, yeah the implementation is the horror
<danvet>
the way it integrates clears a fairly good bar imo (compared to other C doc systems at least)
<jekstrand>
The advantage to something doxygen-like is that we've written a LOT of doxygen comments in mesa
<jekstrand>
Lots of it needs to be reworked but still
<jekstrand>
The thing that's horrible about doxygen is that it's MD-based not RST-based
<jekstrand>
And, as verbose and ugly as rst can be, the context switch to and from MD is seriously mean.
<thaytan>
Lyude, Linux 5.12 landed in Fedora and I finally have an official kernel with OLED backlight support - thank you! :-D
<thaytan>
I'm not sure when this actually happened - I had given up testing the brightness keys some time ago :)
camus has joined #dri-devel
<danvet>
jekstrand, well that's kinda what I like with the kerneldoc flavour, it removes the need for most of the ugly rst
<jekstrand>
With doxygen, hyperlinking is automagic
<danvet>
and aside from that the rst markup, if used lightly, is imo fairly non-offensive
<jekstrand>
jani: Oh, that's you? I did look at that one. :)
<jekstrand>
If there's someone here I can bother about it, I'd feel a lot more comfortable depending on something and assuming it'll grow with our needs.
mbrost has joined #dri-devel
<danvet>
jekstrand, there's also a sphinx extension which does the kerneldoc hyperlinks in any .rst file text
<jekstrand>
jani: It looks to me like it can only autodoc whole files. Is there a way I can pull in a single function or struct at a time?
<jani>
jekstrand: nope, not yet
<jekstrand>
jani: How much work would it take to wire that up?
<jani>
jekstrand: I've mostly just wanted to keep it dead simple at first
camus1 has quit [Ping timeout: 480 seconds]
<jekstrand>
Yeah, I get that. But the spot I'm sitting in is that I've got a pile of MD that I'd like to sphinxify and get into the mesa tree and I don't want to do a huge amount of python hacking.
tzimmermann has quit [Quit: Leaving]
<jani>
jekstrand: I don't think it's a huge amount of work to make it happen. I've looked into it. probably would want to integrate the whole thing into sphinx autodoc better, and figuring that out is the hardest part I think
<jekstrand>
ah
<jani>
the person who wrote sphinx-c-autodoc contributed something to hawkmoth first, and then wrote his own from scratch. looking at that, I think it's an uncomfortable amount of code for the thing tbh
<jekstrand>
*sigh*
ChristinaAguiler has joined #dri-devel
txenoo has quit [Quit: Leaving]
<jekstrand>
sphinx-c-autodoc works, it just mangles comments unacceptibly
ChristinaAguiler has quit [Remote host closed the connection]
<jani>
can't help with that :)
<daniels>
jekstrand: 'a particular reason for using meson' - yeah we wanted ninja -C build docs to just ... build all the docs ... and not have to worry about another way of specifying the src/output dir, or dependencies on built source, or any of that garbage that a build system does for you once
<daniels>
danvet: I can't remember why sphinx + kerneldoc didn't work; might've been prevalence of existing doxy-formatted comments, or might've been because it relies on kernel specifics (types or build system or whatever, I can't remember), but it didn't work for some reason. so we looked at hawkmoth instead which also didn't work for reasons I've paged out of my brain
<daniels>
I think the main issue with hawkmoth is that you couldn't \group @foo types and functions, but you had to do all the grouping by hand over in doc world
<daniels>
which seemed likely to go out of sync even before it landed
<jani>
jekstrand: part of the chicken-and-egg problem with hawkmoth is that first, I originally wrote it as a pure own time pet project for fun, and second, I don't really know any open source projects that use it for anything bigger. I get the occasional bug report, but I don't know what people want.
<daniels>
(sorry jani if I'm massively misrepresenting hawkmoth here)
txenoo has joined #dri-devel
<daniels>
jani: maybe you remember why it was it didn't work for us? I only remember that it looked really really awesome on paper and I was pretty disappointed it didn't work out but our conclusion was that if it did work out, we'd bin breathe and switch over in a heartbeat
<jekstrand>
jani: If you want a bit open-source project to use it, I'm happy to use it in Mesa. I really don't want to depend on doxygen. :-)
<jani>
daniels: that's a fair representation. another issue IIRC was that you had the same function names in two parts of the project. old versions of sphinx couldn't handle that, now there's namespaces
<jekstrand>
I'm in the same boat as daniels. If we can get hawkmoth to work for Mesa's needs, I'll kick breathe to the curb in a heartbeat.
<jani>
:)
<jekstrand>
But I want my personal immediate investment to be in writing the docs themselves, not the tool to generate them.
<jani>
I also see small projects use doxygen+breathe+sphinx, and that's huge overkill for most project needs
<jekstrand>
Maybe dcbaker can help out some too?
<daniels>
jani: oh yeah, some types are overloaded between Wayland client/server side
<daniels>
jekstrand: yeah, unfortunately jani is the only one keen enough to work on doc infrastructure
luzipher_ has joined #dri-devel
naveenk2 has quit []
shankaru has joined #dri-devel
<jani>
this being a hobby project, I don't want to make big commitments to implementing stuff on my own. but I'd really like to talk about what the minimal requirements would be to switch over
<jani>
I understand picking and choosing the symbols to document is one, and whole file is too coarse granularity, at least for non-trivial projects
<bl4ckb0ne>
maybe not of a use case, but if the connector id doesnt belong to the supplied drm fd
<bl4ckb0ne>
i don't think that can be classified as an error
<bl4ckb0ne>
the device not matching the drm fd, that's an error
<emersion>
why isn't it an error?
<bl4ckb0ne>
its just not found, there's no error
<emersion>
why?
<emersion>
> The given pname:connectorId must be a resource owned by the provided
<emersion>
pname:drmFd. If not, the error code ename:VK_ERROR_UNKNOWN must be returned
<emersion>
^ spec
<bl4ckb0ne>
ah wait i wrote that
GloriousEggroll has joined #dri-devel
<bl4ckb0ne>
so its VK_ERROR_UNKNOWN and not VK_ERROR_INITIALIZATION_FAILED
<emersion>
imho it's always better to make errors return errors, instead of silently failing
<bl4ckb0ne>
first one should be VK_ERROR<UNKNOWN too
<emersion>
i can't imagine a situation where an invalid connector ID is something a vulkan app would want
<emersion>
ah, yeah, error is UNKNOWN
<bl4ckb0ne>
maybe user error
<emersion>
user error?
<bl4ckb0ne>
like in the case of drm leasing, there's 2 devices advertised and the user gets the connector id from the first and the drm fd from the other
<bl4ckb0ne>
user mistake*
<emersion>
right. we should try our best to detect that and error out
<emersion>
another mistake is if the non-master DRM FD has a connector ID, but it's missing from the leased DRM FD
<emersion>
although the mistake would be spotted in acquireDisplay
<bl4ckb0ne>
ah while we're at it, vkAcquireDrmDisplayEXT has nothing about the drm fd not corresponding to the physical device
<emersion>
right, a devicesEqual check would be nice
<bl4ckb0ne>
i made it in the code but not in the ext
Manderson has joined #dri-devel
Manderson is now known as Guest2272
Guest2272 has quit [Remote host closed the connection]
Highdude has joined #dri-devel
Highdude has quit [Remote host closed the connection]
cedric is now known as bluebugs
luzipher__ has joined #dri-devel
laktech has joined #dri-devel
laktech has quit [Remote host closed the connection]
<jekstrand>
What container does the docs stuff run in?
<dcbaker>
jekstrand: I read the backlog, what kind of help with docs stuff are you looking for?
<jekstrand>
dcbaker: Currently, I'm attempting to use doxygen+breathe to pull docs into sphinx
<jekstrand>
dcbaker: jani has a thing called hawkmoth which does it directly by invoking clang to parse C files and snag the docs.
<jekstrand>
dcbaker: I really hate doxygen and it seems everyone else does too
<dcbaker>
yeah, doxygen is pretty awful...
<jekstrand>
dcbaker: But I don't want to spend my time writing docs tools. I'd rather (and I think most others would agree) I spend it dumping my brain into actual documentation.
<dcbaker>
I've actually been meaning to write a sphinx module for Meson but have never come up with a justification for putting the time in... lol
<jekstrand>
dcbaker: I don't care about meson integration
<jekstrand>
dcbaker: I care about pulling docs from C
<jekstrand>
Which hawkmoth does, mostly. But it pulls whole files at a time. I want to do more of a sphinx/kerneldoc thing and grab one function at a time so we can actually organize things.
<dcbaker>
so the problem is breathe requires doxygen, hawkmoth works file-at-a-time and sphinx-c-autodoc was?
<daniels>
jekstrand: search .gitlab-ci.yml for ^\.docs-base:
<anholt>
jekstrand: Top of the test-docs-mr job says "Using Docker executor with image quay.io/freedesktop.org/ci-templates:ci-fairy-sha256-[...]", so looks like something built by the ci-templates project.
<daniels>
quite
<daniels>
which is alpine + python, then the job definition I mentioned above uses apk to install Graphviz and pip to install Sphinx
<jekstrand>
Let's see if I can apk install doxygen
<jekstrand>
Hey, would you look at that. It seems to work. :D
<jekstrand>
Ok, I've got doxygen+breathe working. Good enough for now.
ngcortes has joined #dri-devel
khfeng has quit [Ping timeout: 480 seconds]
jwashin1 has joined #dri-devel
jwashin1 has quit [Remote host closed the connection]
<daniels>
jekstrand: now you're generating docs out of CI, public/ is what's served by https://docs.mesa3d.org/
<daniels>
jekstrand: the regexes in the pages: block control when the docs are regenerated
lynxeye has quit [Quit: Leaving.]
alanc-away has quit []
alanc has joined #dri-devel
camus1 has joined #dri-devel
camus has quit [Ping timeout: 480 seconds]
idr has joined #dri-devel
gouchi has joined #dri-devel
bcarvalho has quit [Remote host closed the connection]
bcarvalho has joined #dri-devel
noortrypah has joined #dri-devel
noortrypah has quit [Remote host closed the connection]
blkqi has joined #dri-devel
blkqi has quit [Remote host closed the connection]
mlankhorst has quit [Ping timeout: 480 seconds]
aswar002 has joined #dri-devel
lemonzest has quit [Quit: Quitting]
naveenk2 has joined #dri-devel
RobertC has quit [Ping timeout: 480 seconds]
pekkari has quit []
<Lyude>
thaytan: glad to hear it :3
craftyguy has quit [Quit: craftyguy]
craftyguy has joined #dri-devel
<siqueira>
ivyl danvet I want to apply one IGT patch upstream, but it has been a long time since the last time I did it. Just for checking, are these the correct procedures: (1) make sure that the patch passes everything, (2) add patches with Reviewed-by, (3) on top of the master branch, apply the target patch, (4) just push the change. Is that correct? Did I
<siqueira>
miss something?
<danvet>
sgtm
<danvet>
and for questions Adrinael and ivyl are maintainers, not so much me
<Adrinael>
That is correct
PocketKnife has joined #dri-devel
PocketKnife has quit [autokilled: Suspected spammer. Mail support@oftc.net with questions (2021-06-15 18:12:28)]