mozilla :: #gfx

18 May 2017
00:35pulsebotCheck-in: - sotaro - Bug 1354198 - Fix WebRenderBridgeParent::RecvShutdown() r=nical
00:41jgilbertLenzak, are you looking at ANGLE's skip-validation?
02:01daoshengmustandups: profiling Zen Garden demo and confirm the real bottleneck would not be validatebufferfetching()
02:37Lenzakjgilbert: Yes, but I only found an attribure we can pass to create an ANGLE context, and it can only skip some minor validation.
02:37LenzakThe performance gain is not significant.
02:37LenzakI'll continue looking at it since I have found the cause of bug 1354997.
02:37firebot NEW, OOM @ libglesv2 while compiling WebGL shader
02:56pulsebotCheck-in: - Morris Tseng - Bug 1336986 - Enable layers.advanced.bullet-layers if webrender is enabled. r=kats
02:56pulsebotCheck-in: - Morris Tseng - Bug 1336986 - Mark fails-if for some tests. r=mchang
02:56pulsebotCheck-in: - Morris Tseng - Bug 1336986 - Return false in ContainsOnlyColoredGlyphs if no glyphs are drawing. r=jrmuizel
03:11Lenzakstandups: Conirmed bug 1354997 is an ANGLE bug, having reported it to crbug.
03:11firebot NEW, OOM @ libglesv2 while compiling WebGL shader
05:55Jerry_IRCCloudstandups: 5/17 fix the crash for external image:
09:43pulsebotCheck-in: - Ethan Lin - Bug 1363998 - Part1. Correct the coordinate space of webrender mask layer. r=kats
09:43pulsebotCheck-in: - Ethan Lin - Bug 1363998 - Part2. Add pref "layers.advanced.background-color" to all.js and turn it on for certain reftests. r=kats
10:50pulsebotCheck-in: - Andrew Osmond - Bug 1363347 - Allow multiple external image IDs to be bound to the same image host if owned by the same WRBridge. r=sotaro
12:40BasDoes anyone know what Angle magic gets me a texImage2D from a DXGI SharedHandle?
12:57katsethlin_: reftest failures from your patches on windows
12:57katslooks like we need a larger fuzz on windows
13:17pulsebotCheck-in: - 230 changesets (bugs 1358420, 1298803, 1356871, 1344924, 1364185 and 122 other bugs)
13:28katsBas: ping
13:29Baskats: pong
13:30katsBas: can you join #buildduty ? there appears to be a windows bluescreening issue, possibly related to your patch
13:35mrobinsonmstange: I see you worked on CSS filters somewhat recently. Do you happen to know where mFrameSpaceInCSSPxToFilterSpaceTransform from nsFilterInstance is actually initialized?
13:37mrobinsonmstange: Ah, I see now that it's from FrameSpaceInCSSPxToUserSpaceOffset(...). Sorry!
13:38Jerry_IRCCloudkats: ping. I need the wr updating again..... it fix the crash when we use the external image.
13:38Jerry_IRCCloudit's just 7 commit ahead of the current graphic-branch
13:39katsJerry_IRCCloud: ok, i can do it today. try runs have been green
13:39Jerry_IRCCloudkats: I will do that later. or is that bug already exist?
13:39Jerry_IRCCloudkats: oh. thanks.
13:39katsJerry_IRCCloud: there's always a bug already
13:40katsyou're even cc'd on it
13:41Jerry_IRCCloudkats: thanks for the wr updating.
13:48katsJerry_IRCCloud: it's a trivial update this time, just one patch. i flagged you for review
13:49Jerry_IRCCloudkats: ok. I'm looking for that.
13:51Jerry_IRCCloudkats: is it no cargo.toml updating for these changing?
13:52katsJerry_IRCCloud: there were no version bumps
13:52Jerry_IRCCloudkats: lol
13:54Jerry_IRCCloudkats: r+
13:55katsJerry_IRCCloud: thanks. i'll wait for the try push just to be safe and then land it
14:21jrmuizelkvark: are you working today?
14:22kvarkjrmuizel: look to your right
14:29jrmuizelmstange: ping
14:51pulsebotCheck-in: - Kartikaya Gupta - Bug 1364993 - Update webrender to 102603520d52f335f152ab74b6bcfdae061b6bc8. r=Jerry
15:00pulsebotCheck-in: - JerryShih - Bug 1362049 - Add R8G8 type in SurfaceFormat. r=bas.schouten
15:00pulsebotCheck-in: - JerryShih - Bug 1362049 - Remove the unused MacIOSurfaceTextureSourceOGL. r=mattwoodrow
15:00pulsebotCheck-in: - JerryShih - Bug 1362049 - Update the texture target and texture format in MacIOSurface. v5. r=mattwoodrow
15:00pulsebotCheck-in: - JerryShih - Bug 1362049 - Update the ext-image interface in WR binding. v2. r=nical
15:00pulsebotCheck-in: - JerryShih - Bug 1362049 - Add AddWRImage() to call the proper WR commands for all TextureHost types. v3. r=nical
15:00pulsebotCheck-in: - JerryShih - Bug 1362049 - The BufferTextureHost::AddWRImage() and MacIOSurfaceTextureHostOGL::AddWRImage() implementations. v3. r=nical
15:00pulsebotCheck-in: - JerryShih - Bug 1362049 - Use channel_index to get the correct channel data info from RenderTextureHost. r=nical
15:00pulsebotCheck-in: - JerryShih - Bug 1362049 - Handle multiple-channel format for RenderMacIOSurfaceTextureHostOGL. v4. r=mattwoodrow
15:00pulsebotCheck-in: - JerryShih - Bug 1362049 - Use the MacIOSurface texture directly for video playback. r=nical
15:22katsGankro: are you planning to update that webrendest patch?
15:22Gankrokats: oh sorry, I didn't see the comments
15:24katsGankro: do you get bugmail?
15:25Gankrokats: yeah sorry it got lost in a bunch of other noise, I need to add some filters for mozilla stuff
15:26GankroJust need to eat breakfast and wake up...
15:26gw280jrmuizel: I missed this
15:26gw280is "Frontier Edition" just AMD's name for "Founder's Edition"
15:27gw280also that's a very wide bus
15:27gw280haha "Launch Price: (A profitable number)"
15:34mrobinsonjrmuizel: This is a fairly early WIP.
15:34firebotBug 1365972 UNCONFIRMED, Convert nsDisplayFilter to a WR display item
15:34jrmuizelmrobinson: I'll take a look
15:34mrobinsonjrmuizel: Thanks!
15:37jrmuizelmrobinson: from my very quick look it seems sane
15:38jrmuizelmrobinson: rhunt is probably a good person to review it more throughly
15:43rhuntmrobinson: Yeah, I can probably take a look at that
15:43bdahljrmuizel: i was looking into using gfxConfig, but it looks like that's initialized too late, since i need IsHeadless pretty early on in startup. I'm thinking there are two options: 1) use a static and only check PR_GetEnv once in IsHeadless and store it or 2) add a gfxPlatform::EnableHeadless to set it during startup, but then i need to explicitly call that
15:43bdahlin every new process
15:44jrmuizelbdahl: I maybe slightly prefer the 2nd option
15:44jrmuizelbdahl: do you have a preference?
15:46jrmuizelbdahl: though I can see how being able to inherit the environment is nice
15:46jrmuizelbdahl: so maybe it's fine to just do that
15:48bdahljrmuizel: no preference, 1 is less work, but 2 seems cleaner though i also need to pass on to whether the process is headless to all the child processes so that may complicate it more
15:48bdahlalright, maybe i'm leaning towards 1
15:49jrmuizelgo for it
16:26pulsebotCheck-in: - JerryShih - Bug 1364922 - Add a new op OpAddExternalVideoImage for multiple-channel video image. r=nical
16:26pulsebotCheck-in: - JerryShih - Bug 1364922 - Update AddWRImage() interface for multiple ImageKeys. v2. r=nical
16:26pulsebotCheck-in: - JerryShih - Bug 1364922 - Fix the error calculation of ExternalImageId. r=nical
16:26pulsebotCheck-in: - JerryShih - Bug 1364922 - Create individual imageKey for each channel in the textureClient. v2. r=nical
16:32Jerry_IRCCloudstandups: 5/18 land Bug 1364922 and Bug 1362049
16:32firebot FIXED, hshih [WR] handle the multiple-channel image format
16:32firebot FIXED, hshih [WR] handle the multiple-channel MacIOSurface for video playback
16:53daoshengmustandup: debugging base64 image encode for Bug 1343730 and get a correct result.
16:53firebot NEW, [webvr] VRDisplayPuppet submitFrame support
16:53daoshengmustandups: debugging base64 image encode for Bug 1343730 and get a correct result.
18:01ashughesdvander: I'm working on a TelEx for Compositor Process (Software) - is there a way to detect via JS if a user is unaccelerated?
18:02ashughesI want to make sure we're only giving the experiment to those users
18:29katsGankro: thanks. want me to land the patch?
18:29Gankrokats: sure
18:29Gankrokats: also omg at functions coercing to bool
18:29pulsebotCheck-in: - Alexis Beingessner - Bug 1365418 - Expose other layers.advanced configs, and add webrendest option to toggle them all on. r=kats
18:30katsGankro: yeah, c++ is sad sometimes
18:32dvanderashughes: does the JS have chrome privileges?
18:32dvanderashughes: actually it shouldn't matter
18:32dvanderif you flipthe software pref on it won't affect accelerated users
18:34ashughesdvander: I want to make sure the experiment only goes out to unaccelerated users so the data I get back is clean.
18:41dvanderashughes: what does it mean that it wouldn't be clean?
18:42ashughesdvander: it's easier for me to do an A/B comparison if I know I'm only get data from users who'd be affected by the change
18:43dvanderashughes: can you just filter out people who have acceleration, or does that limit the number of responses you'd get back?
18:43ashughesif a bunch of HWA users are getting the experiment it might make the results look more positive than they actually are
18:44ashughesI don't get that info in the crash_aggregates table so I wouldn't be able to filter that out
18:45dvanderashughes: how early in startup does the experiment try to activate?
18:46ashughesI believe the experiment is a restartless addon so it activates immediately on install
18:46ashughesbut if it's changing a pref that requires a restart then we wouldn't get valid results until that user restarts
18:50ashughesdvander: where does about:support get the compositing info from? if I can detect compositing == Basic would that be sufficient?
18:51dvanderashughes: it gets it from nsIDOMWindowUtils but unfortunately you won't have that available yet. you actually might not have anything acceleration available yet
18:52dvanderashughes: is it possible for the addon to wait for an event to flip the pref, and then wait for a restart later?
18:53ashughesit's something I could look in to
18:53dvanderashughes: so we stick the compositor type in telemetry, like this:
18:54dvanderbut that runs too early most of the time
18:54dvanderso we also listen to this event for the first compositor to be created:
18:55dvanderif you wanted something less accurate, and just use the block list
18:55dvanderyou could qurey nsIGfxInfo.getFeatureStatus for FEATURE_DIRECT3D_11_LAYERS
19:01ashughesdvander: so if I do gfxInfo.getFeatureStatus("FEATURE_DIRECT3D_11_LAYERS") locally I get returned 2
19:01ashughesthat does that mean?
19:01ashughes*what does that mean?
19:01dvanderashughes: looks like 2 is "unknown". I uhhh.. don' tknow what that means :)
19:02dvandermaybe jeff m. does. maybe it is too early to query even that.
19:33mstangebotond: ping
19:34mstangebotond: we can't find the password for the Mac machine
20:31jrmuizeltnikkel: ping
20:31tnikkeljrmuizel: hi
20:31jrmuizeltnikkel: why do you change your irc nick?
20:31jrmuizelwhy did
20:32jrmuizeltnikkel: also do you know the difference between xul:image and html:img?
20:32tnikkeljrmuizel: :tn had a conflict on bugzilla, and when i switched to irccloud i changed nicks so i wouldnt have to deal with remember my nickserv password i guess?
20:32tnikkeljrmuizel: yes
20:34tnikkeljrmuizel: they share almost no code in common
20:35jrmuizeltnikkel: do they have any differences in behaviour?
20:36jrmuizeltnikkel: do we have any reason to keep xul:image around?
20:36jrmuizeltnikkel: is there any reason we should be using it in the ui?
20:36mattwoodrow the cost of rewriting all users? I suspect thats substantial
20:37jrmuizelmattwoodrow: there's maybe only 91 of them
20:37jrmuizelmattwoodrow: and I'm interested in knowing why it wouldn't just be s/image/img/
20:37mattwoodrowjrmuizel: Wed have to deal with addons too if we wanted to get rid of it
20:38jrmuizelmattwoodrow: we're dealing with addons in 57
20:38mattwoodrowjrmuizel: iirc the syntax is different
20:38mattwoodrowindeed we are
20:39jrmuizelare there differences other than syntax?
20:39jrmuizelGankro: will the pref patch land soon?
20:39Gankrojrmuizel: I think kats just merged it
20:39tnikkelxul image has this behaviour where it can use the image from the list-style-image style instead of the src that i know the ui uses, and it also has -moz-image-region that they use too
20:40jrmuizelGankro: glorius
20:40Gankrojrmuizel: you'll know if it's in about:config search
20:40tnikkelui people might have more reasons
20:40jrmuizeltnikkel: ah yes, that sounds sort of familiar
20:40tnikkeljrmuizel: there might be reasons relating to drawing with an undecoded or partially decoded image (ie the ui only wants fully decoded images draw)
20:41jrmuizeltnikkel: that sounds like functionality that we could just add to img
20:41tnikkelthere was a bug where dao had a patch to change favicons to use standard css background images but it couldn't land because of some issue
20:41tnikkeljrmuizel: yes
20:42tnikkeljrmuizel: what brings this up anyway?
20:42jrmuizeltnikkel: we don't implement webrender for xulimage yet
20:43jrmuizeltnikkel: and trying to kill things a little bit, makes me feel better about implementing it
20:44jrmuizelmconley: ping
20:44mconleyjrmuizel: in triage, but what's up?
20:45jrmuizelmconley: will you be around when it's done?
20:45mconleyjrmuizel: briefly, anything I should prep for?
20:45jrmuizelmconley: no just visit me
20:45mconleywill do
20:46mattwoodrowjrmuizel: The xul image display item already supports conversion to ImageLayer doesnt it?
20:47tnikkelmattwoodrow: the code appears to be there, if the -moz-image-region thing isn't used at least
20:47mattwoodrowIt definitely used to work some of the time, the throbber image got layerized at one point
20:48mattwoodrowYeah, if we added subimage/tiling support to ImageLayer, then we could probably make it unconditional
20:48jrmuizelmattwoodrow: yeah, I'm not really worried about the difficulty
20:49jrmuizeljust like to complain
20:49mattwoodrowdont we all!
23:14dvanderBas: does vs2017 community edition work with mozbuild yet?
23:14dvanderor are we able to request professional already?
19 May 2017
No messages
Last message: 91 days and 6 hours ago