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

Issue 834477 link

Starred by 1 user

Issue metadata

Status: Duplicate
Merged: issue v8:7668
Owner:
Closed: Apr 2018
Cc:
Components:
EstimatedDays: ----
NextAction: ----
OS: Linux , Android , Windows
Pri: 1
Type: Bug



Sign in to add a comment

Memory metrics regression in v8

Project Member Reported by erikc...@chromium.org, Apr 18 2018

Issue description

We're observing a massive memory metrics regression for v8 [4*10^9 on Android], 3*10^19 on desktop. 

Since v8 can't be using that much memory, there must be a regression in the accounting. 

Last known good: 551570
First known bad: 551591

autorolls: 

https://chromium.googlesource.com/v8/v8/+log/059b1287..bde5be65
https://chromium.googlesource.com/v8/v8/+log/bde5be65..f70a4135

We should try not to release a canary with this - it will make a muck of all of our metrics.
 
Cc: mlippautz@chromium.org u...@chromium.org
Components: -Blink>JavaScript Blink>JavaScript>GC
Owner: petermarshall@chromium.org
Status: Assigned (was: Untriaged)
It's likely 
  https://chromium.googlesource.com/v8/v8/+/6c68efac1482f293c4810babdcc874f8b453f900

The CL apparently forced an underflow somehwere. I don't know the specifics though (I am currently traveling). The CL has been reverted here: https://chromium-review.googlesource.com/c/v8/v8/+/1016600

Assigning to memory sheriff to double check.
Mergedinto: v8:7668
Status: Duplicate (was: Assigned)
Yes, that is the same problem. It looks like the canary has already been released, without the revert included because it was too late - 68.0.3400.0 contains the bug. Merging with the main tracking bug.

Sign in to add a comment