Bisect info: 455719 (good) - 455720 (bad)
https://chromium.googlesource.com/chromium/src/+log/772559ed..5e921f82?pretty=fuller
The only CL is r455720 "Update V8 to version 5.9.34.2 (Switch on I+TF)"
Landed in 59.0.3037.0

Since this been there since M59 not tagging any blocker labels tagging with M63.

Untriaging it so that it gets addressed.

Thanks, compiler & ignition folks, please investigate and retriage.

This issue has been Available for over a year. If it's no longer important or seems unlikely to be fixed, please consider closing it out. If it is important, please re-triage the issue.

Sorry for the inconvenience if the bug really should have been left as Available.

For more details visit https://www.chromium.org/issue-tracking/autotriage - Your friendly Sheriffbot

Looks like this works now without jank?

It's still broken in 69 and Canary.
The bisect still points to the same CL.
Expected: the balloon slowly moves at its initial position.
Observed: the balloon position is reset every second to the bottom of the viewport.

What is the purpose of

for (var k = 0; k < 10000000; k++)
                    var blahblahblah = 100 * 312;

It does not seem to do anything except looping?

Hm, tight loop like that could mean an OSR issue. I don't have time to take this on at the moment, but that would be my first guess.

I've tried commenting out or deleting these two lines and I still see the bug.
Used a locally edited version served via Requestly extension instead of https://www.blend4web.com/tmp/blend4web_ce/src/data.js
This reminds me of those bug reports where seemingly irrelevant changes to the code produce/hide some bug.
Might be helpful to do a bisect prior to I+TF was shipped but I don't know how to force-enable it in older builds.

Try running older builds with "--ignition --turbo". It would also be helpful to figure out which function(s?) have incorrect data, by e.g. printing inputs and/or return values with console.log and diffing them. I realise that's somewhat of a tough request though.

How do other browsers behave?

re 13 part a:

Bisect info: 410370 (good) - 410383 (bad)
https://chromium.googlesource.com/chromium/src/+log/9d009539..ba4eb17e?pretty=fuller
Suspecting r410383 "Update V8 to version 5.4.373"
Landed in 54.0.2824.0

V8 log: https://chromium.googlesource.com/v8/v8/+log/d4fa8ea7..d67f12f7
There are only two TF commits but it might be helpful if someone does a per-revision bisect of V8.

Notes:
* Used the original page code.
* Command line: --js-flags="--ignition --turbo"
* In these old builds the balloon appears after a ~15sec pause.

re 13 part b: that's probably up to the reporter.

Stab in the dark would be https://codereview.chromium.org/2220973002 and some sort of weird incorrect NaN behaviour.

Does the issue repro with "--no-opt"? Might have to use "--no-crankshaft" for older builds.

--js-flags=--no-opt fixes the bug. Tested in r410383, r589108.

Cool, looks like a TurboFan issue then, assigning to jarin@ for further investigation.

+ahaas: your CL came up in a bisect of this issue. 

The NextAction date has arrived: 2018-09-20