Issue metadata
Sign in to add a comment
|
Epic Zen Garden runs on Firefox Nightly but not on my system chromium
Reported by
kybernet...@gmail.com,
Jan 17 2018
|
||||||||||||||||||||||
Issue descriptionChrome Version : Version 63.0.3239.84 (Official Build) Built on Ubuntu , running on Ubuntu 14.04 (64-bit) OS Version: Ubuntu 14.04 URLs (if applicable) : https://s3.amazonaws.com/mozilla-games/ZenGarden/EpicZenGarden.html Other browsers tested: Add OK or FAIL after other browsers where you have tested this issue: Safari: Firefox (Nightly): OK IE/Edge: What steps will reproduce the problem? 1. Load the url, wait for errors 2. Look on STDERR 9704:9704:0117/141240.281583:ERROR:shader_manager.cc(120)] Shader translator allowed/produced an invalid shader unless the driver is buggy: <shader was dumped here, do you guys really need it? you can get it yourself right?> --info-log-- ERROR: 0:82: '4294967295u' : integer overflow ERROR: 0:82: '+' : wrong operand types no operation '+' exists that takes a left-hand operand of type 'uint' and a right operand of type 'const mediump int' (or there is no acceptable conversion) What is the expected result? A lovely tech demo What happens instead of that? A sad red error box. Please provide any additional information below. Attach a screenshot if possible. 01:00.0 VGA compatible controller: Advanced Micro Devices, Inc. [AMD/ATI] Blackcomb [Radeon HD 6970M/6990M] [ 47.204] (II) Module radeon: vendor="X.Org Foundation" [ 47.204] compiled for 1.18.3, module version = 7.7.0
,
Jan 17 2018
,
Jan 18 2018
Tested this issue on reported version 63.0.3239.84,on latest stable 63.0.3239.132 using Ubuntu 14.04,mesa driver and seeing different errors both on console and chrome://gpu log messages. Log Messages in chrome://gpu: [8053:8053:0118/172703.286367:ERROR:gles2_cmd_decoder.cc(5409)] : Error: 3 for Command kBlitFramebufferCHROMIUM [8053:8053:0118/190053.699707:ERROR:gles2_cmd_decoder.cc(5409)] : Error: 3 for Command kBlitFramebufferCHROMIUM @Reporter: Could you please check the log message,attached console log and let us know if this is the same error you are seeing. Else please let us know where to check the error you mentioned above. This would help in further triaging of the issue. Thanks!
,
Jan 18 2018
,
Jan 18 2018
Nope. Totally different error - I pasted everything from stderr too, except for the dump of the shader. Looks like you got a little further. I get the following text in the browser window, in addition to the stderr pasted - it is mostly same thing. First red box: [Building shaders failed: Fragment shader compilation failed: ERROR: 0:82: '4294967295u' : integer overflow ERROR: 0:82: '+' : wrong operand types no operation '+' exists that takes a left-hand operand of type 'uint' and a right operand of type 'const mediump int' (or there is no acceptable conversion) Current renderer: Google Inc. Google SwiftShader/WebKit WebKit WebGL WebGL 2.0 (OpenGL ES 3.0 Chromium), WebGL GLSL ES 3.00 (OpenGL ES GLSL ES 3.0 Chromium) (software)] Second red box. [Your browser or graphics card does not support the WebGL extension WEBGL_compressed_texture_s3tc. This can impact UE4 graphics performance and quality. Failed to store file wasmModule to IndexedDB, error: DataCloneError: Failed to execute 'put' on 'IDBObjectStore': # could not be cloned.] None of these show up in Firefox.
,
Jan 18 2018
Thank you for providing more feedback. Adding requester "sc00335628@techmahindra.com" to the cc list and removing "Needs-Feedback" label. For more details visit https://www.chromium.org/issue-tracking/autotriage - Your friendly Sheriffbot
,
Jan 18 2018
I decided to attach the shader too - this is the one that gets dumped to STDERR.
,
Jan 18 2018
CCing swiftshader because of: "Current renderer: Google Inc. Google SwiftShader/WebKit WebKit WebGL WebGL 2.0 (OpenGL ES 3.0 Chromium), WebGL GLSL ES 3.00 (OpenGL ES GLSL ES 3.0 Chromium) (software)]"
,
Jan 18 2018
Oh... and... maybe https://bugs.chromium.org/u/3058424708/ should file their errors as a 2nd bug? Maybe they could check to see how Firefox and/or other webgl demos works on their test instance - webgl is such a crapshoot under linux. Also... A friend of mine said Chromium 63 on his ubuntu 16.04 using nvidia blob worked correctly. Also, to verify Chromium was able to handle other webgl, I tried another classic, the aquarium demo. http://webglsamples.org/aquarium/aquarium.html (p.s. - so annoying that I had to type in the http:// 'cause chromium hid it in the url bar when I copied the url after locating in history) Anyway. It ran in Chromium, but ⅟₁₀th the speed of in Firefox. Like, Firefox rendered 60fps for 500 fish and 30fps for 5000, but Chromium did 6fps and 3fps. Wasn't sure if this maybe indicated software rendering or something, so figured I'd toss that in too.
,
Jan 18 2018
You're definitely on software for some reason. Your about:gpu would likely say why. You can attach the saved contents (webpage/complete) if you are interested in triaging why.
,
Jan 19 2018
Ah-hah. I kinda figured. I should have mentioned the aquarium thing sooner. Well, I did in IRC right away to smcgruger and friends who were helping me, but issue trackers do get more attention. So, I guess all the errors are just due to the limitations of a software renderer. Anyway, about:gpu lists the usual linux stability issues and a tracking bug. Firefox maintains blacklists too, and they often don't match Chrome's - it's kinda odd since the two are doing almost the same thing for webgl these days. Maybe you guys should maintain a joint list of hardware reliability. Might save you both effort and users pain.
,
Jan 19 2018
Also, while I would be astounded that Epic Zen Garden could run at more than a few fps on a software render - the error message is still kinda odd. This might maybe be a legit bug in the software rendering?
,
Jan 19 2018
"ATI/AMD cards with older drivers in Linux are crash-prone: 71381, 76428, 73910, 101225, 136240, 357314" Unfortunately because of the way GPU info collection works, we can't see the date of your drivers since they're blacklisted. Maybe you can try updating your GPU drivers to get off of software. And yes, it could very well be a bug in SwiftShader, hence the Component tag. I'll cc the team directly. Alexis or Nicolas can you PTAL?
,
Jan 19 2018
I pasted my driver version in the initial report text pulled from Xorg log. 7.7.0 released 2016-04-07 which is, well, not that long ago really. And, eh, hardly worth messing w/ the ubuntu 14.04 LTS setup on this work laptop - especially since I can just override the blacklist manually in chromium, and Firefox does not blacklist radeon 7.7.0
,
Jan 19 2018
Hi, we will investigate this issue in SwiftShader. In the meantime, it seems like the minimum driver version for your hardware is 8.98. Older drivers had stability issue, apparently. Reference: https://cs.chromium.org/chromium/src/gpu/config/software_rendering_list.json?l=60
,
Jan 19 2018
Interesting, sugoi... So... The problem might once again be ATI's WTF versioning. That has been a constant struggle for the Mozilla guys too. Searching online, 8.980 was released in 2012 ... for windows - so 7.7.0 is actually far more recent? I'm going to guess that the Chromium blacklisting needs to be more nuanced to identify the linux radeon.
,
Jan 19 2018
Assigning to zmo@ to check the possible driver version number issue (i.e. AMD drivers have a different version numbers on Windows and Linux and different drivers for different hardware might even have different version numbers on the same platform, to be investigated). kyberneticist@, can you post here your 'about:gpu' chromium page when you launch chrome using --ignore-gpu-blacklist? Thanks
,
Jan 19 2018
Huh. Odd. I don't see the radeon driver output that I'd swear I correctly picked out of my xorg log. Seems to match more what glxinfo has to say, that is: Glxinfo: OpenGL renderer string: Gallium 0.4 on AMD BARTS (DRM 2.43.0, LLVM 3.8.0) OpenGL core profile version string: 3.3 (Core Profile) Mesa 11.2.0 OpenGL core profile shading language version string: 3.30 As opposed to the xorg log: [ 47.009] (II) Loading /usr/lib/xorg/modules/drivers/radeon_drv.so [ 47.204] (II) Module radeon: vendor="X.Org Foundation" [ 47.204] compiled for 1.18.3, module version = 7.7.0 [ 47.204] Module class: X.Org Video Driver [ 47.204] ABI class: X.Org Video Driver, version 20.0 Buuut anyway, attached. I assume you guys know what you are looking for. Aaand hope this bug report isn't veering wildly off at this point. Sorry.
,
Jan 19 2018
Oh, and FWIW, I'm using this package: https://ubuntu.pkgs.org/14.04/ubuntu-updates-main-amd64/xserver-xorg-video-radeon-lts-xenial_7.7.0-1~trusty2_amd64.deb.html
,
Jan 20 2018
Linux AMD will be enabled soon with GPU. We blacklist too aggressively on this platform, and I am fixing it.
,
Jan 22 2018
,
Jan 24 2018
So, FWIW, while the fact I was in software rendering at all is obv a dupe, the fact that I triggered that shader error at all seems legit, which was the main reason I bothered reporting it ☺
,
Jan 24 2018
Alright, leaving assigned to Alexis. He said he's fixing the compile error but there's other remaining work for making SwiftShader successfully run the demo. |
|||||||||||||||||||||||
►
Sign in to add a comment |
|||||||||||||||||||||||
Comment 1 by smcgruer@chromium.org
, Jan 17 2018Labels: -OS-Windows