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

Issue 890983 link

Starred by 1 user

Issue metadata

Status: Fixed
Owner:
Closed: Oct 1
Cc:
Components:
EstimatedDays: ----
NextAction: ----
OS: Android
Pri: 1
Type: Bug



Sign in to add a comment

Client blacklisting is not reported to DRP PLM correctly

Project Member Reported by ryansturm@chromium.org, Oct 1

Issue description

The bool of whether the lite page was client blacklisted was not appropriately being copied to the UI thread, and was always reported as not client blacklisted.


 
Project Member

Comment 2 by bugdroid1@chromium.org, Oct 1

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/af76daadb696529131a73690da28a01fc0b85d56

commit af76daadb696529131a73690da28a01fc0b85d56
Author: Ryan Sturm <ryansturm@chromium.org>
Date: Mon Oct 01 22:30:03 2018

Fixing DRP PLM pingback previews type reporting for blacklist

When a preview is client blacklisted, this should be reported to the
Data reduction proxy pageload metrics pingback. However, the boolean for
this is not properly copied before content moves the navigation data
cross thread. This is resulting in never reporting the client blacklist
blocking the optimization.

Another bool (offline_preview_used) is also not copied, but it is only
used on the IO thread currently, so there is not functional bug.

Bug:  890983 
Change-Id: I6fafaa91d875b590ee5c5c725bc6f487e63395c4
Reviewed-on: https://chromium-review.googlesource.com/1255889
Reviewed-by: Doug Arnett <dougarnett@chromium.org>
Commit-Queue: Ryan Sturm <ryansturm@chromium.org>
Cr-Commit-Position: refs/heads/master@{#595598}
[modify] https://crrev.com/af76daadb696529131a73690da28a01fc0b85d56/components/previews/core/previews_user_data.cc
[modify] https://crrev.com/af76daadb696529131a73690da28a01fc0b85d56/components/previews/core/previews_user_data.h
[modify] https://crrev.com/af76daadb696529131a73690da28a01fc0b85d56/components/previews/core/previews_user_data_unittest.cc

Status: Fixed (was: Started)
Ryan, do you have a sense for how common this issue was?

I.e. do you expect it to significantly shift the results we get in our PLM data?

I'm asking because we've noticed that the signal we get from opt-outs in PLM data is pretty weak. Could this bug significantly shift the results, or is is only triggered in some rare cases?
I talked with Tom, and this signal (The type being reported as CLIENT_BLACKLIST_PREVENTED_PREVIEW) has 0 hits in records, which makes sense based on the code not copying the bool.
Note that this "type" was only supposed to be available for 69, and 70.
Oh my bad, this is a different field from the one I was thinking of. This doesn't affect the analysis I am doing.

Thanks!

Sign in to add a comment