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

Issue 740843 link

Starred by 1 user

Issue metadata

Status: Duplicate
Merged: issue 73512
Owner: ----
Closed: Jul 2017
Cc:
Components:
EstimatedDays: ----
NextAction: ----
OS: Linux , Android
Pri: 2
Type: Bug



Sign in to add a comment

memory-infra of chrome://tracing for a remote Android device doesn't work

Project Member Reported by hajimehoshi@chromium.org, Jul 11 2017

Issue description

Chrome Version: ToT
OS: Host: Ubuntu Linux x64, Guest: Android Nexus 4

What steps will reproduce the problem?
(1) Connect the device (Android)
(2) Open a tab with a URL like https://google.com/
(3) Open chrome://inspect?tracing and open chrome://tracing for the remote device
(4) Try to record with checking memory-infra

What is the expected result?
Obtained memory infra dumps

What happens instead?
Nothing

Screen shot: https://drive.google.com/file/d/0BwW8PrCcts4WaS1ERUx6b2N6Ums/view?usp=sharing

Please use labels and text to provide additional information.
* After pressing 'Record' button in the dialog, the progress was not changed from 0%. It looks like most of the other data were not recorded.
* The attached JSON file is the dump result when trying to record blink/cc/netlog/toplevel/v8 and memory-infra.

For graphics-related bugs, please copy/paste the contents of the about:gpu
page at the end of this report.

 
trace_android.json
244 KB View Download
Cc: hjd@chromium.org
hjd@ can you take a look to see if you can repro?
> (2) Open a tab with a URL like https://google.com/

This is in the guest device.

> (3) Open chrome://inspect?tracing and open chrome://tracing for the remote device
> (4) Try to record with checking memory-infra

These are in the host device.

Comment 3 by hjd@chromium.org, Jul 11 2017

I picked a ChromePublic.apk off the top of the waterfall[1].

Then I tried repoing with both 59.0.3071.115 (stable) and 61.0.3141.7 (dev) as the host.

The problem reproduced on 59.0.3071.115 but not on 61.0.3141.7.
Perhaps we changed how memory-infra is triggered?

hajimehoshi@: Which version was the host Chrome? If it was not very recent could you try with a very recent build for the host, thanks!

primiano@: Maybe something to do with Erik's changes to the trace config? https://bugs.chromium.org/p/chromium/issues/detail?id=735124

[1]: https://build.chromium.org/p/chromium.linux/builders/Android%20Builder/builds/85126
 

Comment 4 by alph@chromium.org, Jul 11 2017

Components: -Platform>DevTools>Tracing Speed>Tracing
Thanks,

> hajimehoshi@: Which version was the host Chrome? If it was not very recent could you try with a very recent build for the host, thanks!

b1601a7ae547960927ebc20de9776dc0c2e779e3 (Sun Jul 9 21:44:42 2017 -0700)
Version 61.0.3154.0 (Developer Build) (64-bit)

Hmm, this should be fixed?
Sorry but the Chrome (at comment #5) is the guest, and the host is Version '59.0.3071.115 (Official Build) (64-bit)'. Does the host Chrome version matter in this issue? Thanks.
Mergedinto: 73512
Status: Duplicate (was: Untriaged)
> The problem reproduced on 59.0.3071.115 but not on 61.0.3141.7.
That is the range when Erik fixed that:

$ git log 59.0.3071.115..61.0.3141.7 --author=erikchen --pretty=oneline -- \*mem\*
c72902d0a30ee1fc71e7890d14c9ca79c3d387b3 Change memory dumps to not use periodic dumps by default.
...
8b0b2a16a1e141102620e7d173271197cd6b7f3d Revert of Change memory dumps to not use periodic dumps by default. (patchset #3 id:40001 of https://codereview.chromium.org/2912483002/ )
...
0529413d84c06f3720bf4a56b28bb2c7358cefd1 Change memory dumps to not use periodic dumps by default.


> Does the host Chrome version matter in this issue? Thanks.
Yes, I think because of this: https://codereview.chromium.org/2948023002
Which rolled into 247215d39e070e9e9d9ff4dc58a6f37682a5d125
which is between 59 and 61.

Mystery solved.
Thanks hjd@ for checking this.

Sure, thank you very much!

Sign in to add a comment