Chrome 133 issues

[krausest]

My hopes that the run for chrome 133 would just finish without problems vanished this morning when I looked at the logs.
Lots of errors are reported for a special check. I'm expecting a single mousedown event in the chrome trace. Especially for benchmark 07 (create many rows) there appear to be multiple mousedown events reported for many frameworks.

If I open the trace in the chrome profiler there's just one mousedown event:

but there are 11 in the JSON trace:

21795542062 {"args":{"data":{"type":"mousedown"}},"cat":"devtools.timeline","dur":2,"name":"EventDispatch","ph":"X","pid":47288,"tdur":1,"tid":259,"ts":21795542062,"tts":52629}
21795647346 {"args":{"data":{"type":"mousedown"}},"cat":"devtools.timeline","dur":1,"name":"EventDispatch","ph":"X","pid":47288,"tdur":0,"tid":259,"ts":21795647346,"tts":149062}
21795663663 {"args":{"data":{"type":"mousedown"}},"cat":"devtools.timeline","dur":1,"name":"EventDispatch","ph":"X","pid":47288,"tdur":0,"tid":259,"ts":21795663663,"tts":160771}
21795735585 {"args":{"data":{"type":"mousedown"}},"cat":"devtools.timeline","name":"EventDispatch","ph":"I","pid":47288,"s":"t","tid":259,"ts":21795735585,"tts":230110}
21795755207 {"args":{"data":{"type":"mousedown"}},"cat":"devtools.timeline","dur":1,"name":"EventDispatch","ph":"X","pid":47288,"tdur":0,"tid":259,"ts":21795755207,"tts":241391}
21795822211 {"args":{"data":{"type":"mousedown"}},"cat":"devtools.timeline","dur":1,"name":"EventDispatch","ph":"X","pid":47288,"tdur":0,"tid":259,"ts":21795822211,"tts":301339}
21795838718 {"args":{"data":{"type":"mousedown"}},"cat":"devtools.timeline","name":"EventDispatch","ph":"I","pid":47288,"s":"t","tid":259,"ts":21795838718,"tts":312851}
21795905519 {"args":{"data":{"type":"mousedown"}},"cat":"devtools.timeline","name":"EventDispatch","ph":"I","pid":47288,"s":"t","tid":259,"ts":21795905519,"tts":372630}
21795921956 {"args":{"data":{"type":"mousedown"}},"cat":"devtools.timeline","dur":1,"name":"EventDispatch","ph":"X","pid":47288,"tdur":0,"tid":259,"ts":21795921956,"tts":382919}
21795997596 {"args":{"data":{"type":"mousedown"}},"cat":"devtools.timeline","dur":1,"name":"EventDispatch","ph":"X","pid":47288,"tdur":0,"tid":259,"ts":21795997596,"tts":451928}
21796115929 {"args":{"data":{"type":"mousedown"}},"cat":"devtools.timeline","dur":1,"name":"EventDispatch","ph":"X","pid":47288,"tdur":0,"tid":259,"ts":21796115929,"tts":513519}

There are even 8 X phase events ("complete events") if we decide to ignore the I phase events (events without a duration).
There's also the same number of click events:

click 21795542438 {"args":{"data":{"type":"click"}},"cat":"devtools.timeline","dur":67084,"name":"EventDispatch","ph":"X","pid":47288,"tdur":65225,"tid":259,"ts":21795542438,"tts":52990}
click 21795647723 {"args":{"data":{"type":"click"}},"cat":"devtools.timeline","dur":11431,"name":"EventDispatch","ph":"X","pid":47288,"tdur":10176,"tid":259,"ts":21795647723,"tts":149422}
click 21795663758 {"args":{"data":{"type":"click"}},"cat":"devtools.timeline","dur":39992,"name":"EventDispatch","ph":"X","pid":47288,"tdur":38378,"tid":259,"ts":21795663758,"tts":160848}
click 21795735929 {"args":{"data":{"type":"click"}},"cat":"devtools.timeline","dur":11076,"name":"EventDispatch","ph":"X","pid":47288,"tdur":9896,"tid":259,"ts":21795735929,"tts":230439}
click 21795755309 {"args":{"data":{"type":"click"}},"cat":"devtools.timeline","dur":37641,"name":"EventDispatch","ph":"X","pid":47288,"tdur":36384,"tid":259,"ts":21795755309,"tts":241473}
click 21795822538 {"args":{"data":{"type":"click"}},"cat":"devtools.timeline","dur":11136,"name":"EventDispatch","ph":"X","pid":47288,"tdur":9937,"tid":259,"ts":21795822538,"tts":301647}
click 21795838811 {"args":{"data":{"type":"click"}},"cat":"devtools.timeline","dur":37752,"name":"EventDispatch","ph":"X","pid":47288,"tdur":36446,"tid":259,"ts":21795838811,"tts":312927}
click 21795905854 {"args":{"data":{"type":"click"}},"cat":"devtools.timeline","dur":10151,"name":"EventDispatch","ph":"X","pid":47288,"tdur":8976,"tid":259,"ts":21795905854,"tts":372948}
click 21795922053 {"args":{"data":{"type":"click"}},"cat":"devtools.timeline","dur":33364,"name":"EventDispatch","ph":"X","pid":47288,"tdur":31851,"tid":259,"ts":21795922053,"tts":382995}
click 21795997944 {"args":{"data":{"type":"click"}},"cat":"devtools.timeline","dur":10020,"name":"EventDispatch","ph":"X","pid":47288,"tdur":8833,"tid":259,"ts":21795997944,"tts":452258}
click 21796116032 {"args":{"data":{"type":"click"}},"cat":"devtools.timeline","dur":410294,"name":"EventDispatch","ph":"X","pid":47288,"tdur":410012,"tid":259,"ts":21796116032,"tts":513605}

only the last one corresponds to the duration reported by chrome:

[krausest]

Here are the timestamps in msecs (start, stop, duration):

[(0.0, 65.225, 65.225),
 (105.285, 115.461, 10.176),
 (121.32, 159.698, 38.378),
 (193.491, 203.387, 9.896),
 (212.871, 249.255, 36.384),
 (280.1, 290.037, 9.937),
 (296.373, 332.819, 36.446),
 (363.416, 372.392, 8.976),
 (379.615, 411.466, 31.851),
 (455.506, 464.339, 8.833),
 (573.594, 983.606, 410.012)]

All intervals are disjoint. Are they from different runs?
Something else is strange: Seems like there are missing trace files when that occurs (only 3 of 15 traces for alpine keyed are available. The third file with the many click events is larger than the other files (but only 1,8 times larger).

Happens for puppeteer and playwright and for alpine for far always in the third benchmark iteration.

[krausest]

If I change the runner such that for every benchmark iteration a new browser instance is created it works as expected.

But this will (almost certainly) increase the duration of the benchmark: I'll have to check if it's practical. 😢

[leeoniya]

I'll have to check if it's practical.

#1787

https://en.wikipedia.org/wiki/Jevons_paradox strikes again! 😢

[krausest]

So true that it hurts.

[krausest]

First run for keyed looks really bad. But I noticed that I didn't turn WIFI off (despite being on my checklist). Hopefully it was just that. Otherwise we're in trouble:

variance for all but create 1|10k and replace rows is just too high.

[krausest]

Total duration for just the keyed benchmarks was 19 hours 45 minutes 😢 .

[krausest]

Another run - but I couldn't get variance down. After the run the Mac was pretty slow (Windows Server > 100%). Not sure if that happened during the run and caused results bad or if all of those chrome processes caused the window server to go berzerk.

Summary: No chrome 133 this time!

[krausest]

The issue happens with chrome 133.0.6943.126 for Testing. Good news: Chrome canary didn't have this issue - hope is on the rise...

133.0.6943.141 for Testing also has the issue, but my normal desktop browser with 133.0.6943.142 doesn't. Strange, but better than having the issue in the desktop version too.
134.0.6998.35 (beta) for Testing is broken too.

I'll switch to the non-testing version.