Using the per-revision bisect providing the bisect results,
Good build:57.0.2985.0 (revision : 444244)
Bad build: 57.0.2986.0 (revision : 444600)

You are probably looking for a change made after 444598 (known good), but no later than 444599 (first known bad).

CHANGELOG URL:

The script might not always return single CL as suspect as some perf builds might get missing due to failure.

 https://chromium.googlesource.com/chromium/src/+log/6936c67f34dfe72954dd522f16f2dfa9c600da84..694194cd5bd8e7ced0c59eac109f0033aa7e1f30

From the CL above, assigning the issue to the concern owner 

@lukasza - Could you please check whether this is caused with respect to your change, if not please help us in assigning it to the right owner.

Review-Url: https://codereview.chromium.org/2639813002

Note : Adding ReleaseBlock- Stable for now, please feel free to remove/edit this.

Thanks!

alph@, I wonder if you could take a look please?

r444599 cannot be the root cause, because it doesn't affect product code at all (it only changes a clang tool that is currently NOT used during the Chromium build).  FWIW, I used per-revision bisecting to arrive at:

c968f85 DevTools: tweak performance landing page style. by alph

This seems plausible, given that the CL above is related to DevTools.  OTOH, I do not that the repro doesn't seem 100% reliable in "bad builds", so it is probably wise to take these bisect results with a grain of salt.

Seems to be a layout bug. The div that changes the width is unrelated to the hovering over other element.

Placing in Blink triage queue.

Not a layout bug, the size of the pane changes on hover triggering a reflow.

alph@: Could you please take a look at this as per C#5.

Gentle ping to get an update on this.

Thanks.

Friendly Ping!!
alph@: Could you please take a look at this as per C#5 as this issue marked as blocker.
Thank you.

Friendly Ping!!
alph@: Could you please take a look at this as per C#5 as this issue marked as blocker.
Thank you.

Just to update, issue is still seen on latest canary 59.0.3040.0 of Windows-7.
alph@: Could you please take a look at this.

Thanks.

This bug existed in previous milestone, not a RB for M58. Feel free to add if needed.

Routing for Blink triaging queue.

Cannot reproduce on Canary on Linux. Could be a layout issue.

Can repro with ToT on Linux.

I don't see the pane change size on hover.

Specifically according to web inspector (how my windows are sized):

<div class="widget vbox timeline-landing-page fill">
is width 358.

The div underneath it is sized as 338 (margin left: 10px, margin right: 10px) which adds to 358.

But after hover the size of the parent div remains the same but the child div is is sized to 323 (margin left: 10px, margin right: 10px)  323+10+10 != 358.

Seems like a layout issue to me.

This is quite hard to debug. Hovering it changes the size of the pane but I can't figure out how or why as I can't get DOM breakpoints to work at all when debugging the inspector. Neither does pretty printing work.

This might very well be a flexbox layout bug but we need someone that knows how dev tools works to describe what happens when the >> icon is hovered and how to get unobfuscated code working to allow debugging.

Either a bisect or (better yet) reduction would also be incredibly helpful but I understand that the latter is hard.

Thank you.

Closing due to lack of feedback. Comment to reopen.

With response to comment #14 :

Narrow bisect info:
https://chromium.googlesource.com/chromium/src/+log/b288ced458b070ee2785301c4600d65806362557..a65f8aef002b29053fcd5ef52d118018c596d1c5?pretty=fuller&n=100

Suspecting : r444568 from Narrow bisect 

@pfeldman  : Could you please help to reassign if your change is not the cause for this change.

Minor issue