New issue
Advanced search Search tips
Note: Color blocks (like or ) mean that a user may not be available. Tooltip shows the reason.

Issue 644742 link

Starred by 4 users

Issue metadata

Status: Fixed
Owner:
Closed: Sep 2016
Cc:
EstimatedDays: ----
NextAction: ----
OS: Linux , Windows , Mac
Pri: 1
Type: Bug-Regression

Blocking:
issue 373987



Sign in to add a comment

Chrome v53.x breaks raycasting in our Unity WebGL project

Reported by bryanat1...@gmail.com, Sep 7 2016

Issue description

Chrome Version       : 53.0.2785.89 m
URLs (if applicable) : 
sample project:
http://a.gardenacres.com.s3.amazonaws.com/webgl/ChromeBug.2/index.html

our production game:
https://apps.facebook.com/gardenquest

Other browsers tested:
  Chrome 52.0.2743.116 m: OK
  Chrome 55.0.2851.0 canary (64-bit): OK

    Firefox 48.0.2: OK

What steps will reproduce the problem?
(1) For the sample project, click the "Get Ray" button a few times and you will eventually see the ray direction data become invalid.
(2) For our production game, since the ray data becomes invalid, all raycasting is fully broken and we use it for much of our mouse/pointer tracking in the 3D space

What is the expected result?
(1) The sample project should simply output a valid Ray direction value (a Vector3 in Unity terminology)

What happens instead?
(1) The ray directional data is NaN

Please provide any additional information below. Attach a screenshot if
possible.

Obviously this issue is a fatal issue for our production game for any players who have updated to the latest stable version of Google Chrome :(.

Thanks.
 
I have no idea if it's related, but there's a certificate error in this repro.

config.uca.cloud.unity3d.com presented a certificate that did not validate, due to RemoteCertificateChainErrors.
0 - A required certificate is not within its validity period when verifying against the current system clock or the timestamp in the signed file.
EXPIRED: 9/6/2016 7:00:00 AM

Comment 2 by ajha@chromium.org, Sep 9 2016

Cc: gov...@chromium.org ajha@chromium.org
Labels: -Type-Bug -Pri-3 ReleaseBlock-Stable M-53 hasbisect OS-Linux OS-Mac OS-Windows Pri-1 Type-Bug-Regression
Owner: hablich@chromium.org
Status: Assigned (was: Unconfirmed)
Able to reproduce the issue on the latest stable(53.0.2785.101) but the same works fine on latest beta(54.0.2840.16) and the latest canary(55.0.2855.0) on Windows-10, Mac OS 10.11.6 and Linux Ubuntu 14.04. Tested with the sample project URL attached above.

Reverse bisected to identify the CL that fixed this(Bisected twice on both Linux and Windows and got the good and bad combination for the below changelog).

Last bad build:   54.0.2787.0
First good build: 54.0.2788.0

Changelog: https://chromium.googlesource.com/chromium/src/+log/a5fa9cf39496b97e0db21e27b5949bc2efca0d60..02f7f5836dab4f3d269618f9b699f7555221d9b0

V-8 changelog: https://codereview.chromium.org/2117933002 

Marking this as Stable blocker for M-53 to get the Fix merged to M-53 if there is any stable refresh.

Michael@: Could you please take a look and help in routing this to appropriate owner. Bisect tool is pointing to the above V8 related change.

Thank you!
Cc: seththompson@chromium.org
Having a look ...

Comment 5 by habl...@google.com, Sep 9 2016

Cc: bbudge@chromium.org
Per comment #6, this is already merged to M53. If nothing is pending for M53, please remove "Merge-Approved-53" label and apply "merge-merged-5.3" label.
Project Member

Comment 8 by bugdroid1@chromium.org, Sep 9 2016

Labels: merge-merged-5.3
The following revision refers to this bug:
  https://chromium.googlesource.com/v8/v8.git/+/8fbcbb4c7e20a66226c6128c122b37957533ef7e

commit 8fbcbb4c7e20a66226c6128c122b37957533ef7e
Author: hablich <hablich@chromium.org>
Date: Fri Sep 09 17:45:22 2016

Merged: [Turbofan] Simplify operand canonicalization on archs with simple FP aliasing. - Changes ...

Revision: 4b76dc85979094ac26499f2e527e422d4f0f6e58

BUG= v8:4124 , chromium:644742 
LOG=N
NOTRY=true
NOPRESUBMIT=true
NOTREECHECKS=true
R=bbudge@chromium.org

Review-Url: https://codereview.chromium.org/2316303007
Cr-Commit-Position: refs/branch-heads/5.3@{#67}
Cr-Branched-From: 820a23aade5e74a92d794e05a0c2b3597f0da4b5-refs/heads/5.3.332@{#2}
Cr-Branched-From: 37538cb2c1b4d75c41af386cb4fedbe5566f5608-refs/heads/master@{#37308}

[modify] https://crrev.com/8fbcbb4c7e20a66226c6128c122b37957533ef7e/src/compiler/instruction.cc
[modify] https://crrev.com/8fbcbb4c7e20a66226c6128c122b37957533ef7e/src/compiler/instruction.h
[modify] https://crrev.com/8fbcbb4c7e20a66226c6128c122b37957533ef7e/src/compiler/move-optimizer.cc
[modify] https://crrev.com/8fbcbb4c7e20a66226c6128c122b37957533ef7e/src/compiler/register-allocator.cc

Labels: -Merge-Approved-53
Removing "Merge-Approved-53" label as it is already merged to M53 at #8.

Comment 10 by bbudge@google.com, Sep 9 2016

I need to merge two more CLs for this issue.
OK, please re-request a merge to M53 once CLs are ready. Michael will have to approve them. 
Please note that merges have to be happened by 3:00 PM PT, Monday, 09/12 in order to make to next week Stable cut. 

Comment 12 by bbudge@google.com, Sep 9 2016

After attempting those merges, they look to be messy, so I'm holding off.

I believe the CL just landed fixes this bug which is the most serious. The other changes would fix a problem on 32 bit ARM which is different from this bug (and not yet reported so unclear how much impact it would have.)

If those changes have to be merged, I will probably have to work no the branch and would be a riskier CL.
OK, sounds good. Thank you.
Status: Fixed (was: Assigned)

Comment 15 by kbr@chromium.org, Sep 9 2016

Cc: kbr@chromium.org

Comment 16 by kbr@chromium.org, Sep 9 2016

Blocking: 373987

Comment 17 by ajha@chromium.org, Sep 12 2016

Labels: TE-Verified-53.0.2785.111 TE-Verified-M53
Verified the merge on the latest M-53(53.0.2785.111) of Windows-10, Mac OS 10.11.6 and Linux Ubuntu 14.04 on the sample project URL from C#0 and this is working as intended.

Thank you!

Sign in to add a comment