Issue metadata
Sign in to add a comment
|
Regression: Page become unresponsive during timeline recording for 'adrifelt.github.io'.
Reported by
jshan...@etouch.net,
Jun 27 2016
|
||||||||||||||||||||||||
Issue descriptionChrome Version: 53.0.2780.0 (Official Build) 5453dc6abc2d7c797c7f9f16df8aed8e6358a217-refs/heads/master@{#402053}-32/64 bit OS: Windows(7,8,10), Mac (10.10.5)(10.11.4), Linux(Ubuntu 14.04 LTS) URL: https://adrifelt.github.io/demos/bubble-demo.html Steps: 1. Launch Chrome, navigate to above URL and open Dev-tool. 2. Go to Timeline section, check all check-box and press 'CTRL+R' key 3. Press ESC key multiple time till 'Loading timeline' message is seen. 4. Now close Dev-tool, switch tab and observe. Actual: Blank page is seen and page become unresponsive after clicking on reload button. Expected: Blank page should not be seen and page should not become unresponsive after clicking on reload button. This is a regression issue broken in M-52, below is bisect info. Good build: 52.0.2708.0 Bad build: 52.0.2709.0 Narrow bisect: https://chromium.googlesource.com/chromium/src/+log/f8563372f7ba26f49ea5b4f9fcff0ef4a5a5ac68..cb13dd10365604c6d8a64e6d123ae388e41a1bfa?pretty=fuller&n=10000 Suspecting: r387347 ? Please help to re-assign is your change is not the cause for this issue.
,
Jun 27 2016
It's not my CL, which has no functional effect. There's only two CLs in that regression range that I'd guess are potential candidates: https://codereview.chromium.org/1888073002: cc : Delete unused properties in effect node https://codereview.chromium.org/1812273003: Eliminate copies of encoded image data I've CCed the authors of those CLs and randomly passed off ownership.
,
Jun 27 2016
https://codereview.chromium.org/1888073002 deleted dead code. So it can't be the cause.
,
Jun 28 2016
This looks to be the same problem as issue 607227 (although I think this one has a more reliable repro case). Marking as a duplicate.
,
Jun 29 2016
The following revision refers to this bug: https://chromium.googlesource.com/chromium/src.git/+/7ba0aa90179d42ac4806386a88d50ed3778ae30b commit 7ba0aa90179d42ac4806386a88d50ed3778ae30b Author: scroggo <scroggo@google.com> Date: Wed Jun 29 22:36:08 2016 Fix dead-lock issue attempting to decode During serialization, if DecodingImageGenerator::onRefEncodedData() returns nullptr, this may require the serialization code to decode so it can have something to then encode and serialize. This can cause a deadlock if the DecodingImageGenerator is using discardable memory, which cannot be used from certain threads (where serialization may be occurring). Currently, onRefEncodedData() returns nullptr if !m_allDataReceived. This was to fix crbug.com/568016 , but that has been further fixed by crrev.com/1970773002, which avoids calling onRefEncodedData() to pass it to the GPU, since even if m_allDataReceived is true, we do not yet support any GPU texture formats. This change returns the encoded data even if the data has not all been received, except if the GrContext argument is not NULL for the future when the GPU starts calling it again. This will result in serializing the existing data and avoid the deadlock. BUG= 607227 , 623475 Review-Url: https://codereview.chromium.org/2108163003 Cr-Commit-Position: refs/heads/master@{#402963} [modify] https://crrev.com/7ba0aa90179d42ac4806386a88d50ed3778ae30b/third_party/WebKit/Source/platform/graphics/DecodingImageGenerator.cpp |
|||||||||||||||||||||||||
►
Sign in to add a comment |
|||||||||||||||||||||||||
Comment 1 by tkonch...@chromium.org
, Jun 27 2016