nednguyen@: is there anybody else that can pick up and drive this?

Roberto: are you able to pick this up this quarter?

Bisect team is quite overloaded in Q4. But I think someone from WebRTC team should be able to pick this up by:

1) Searching for 'angle' in bisect code to get examples: https://cs.chromium.org/search/?q=angle+file:bisect&sq=package:chromium&type=cs

2) robertocn is adding a new method for end-to-end testing the bisect, when it is documented you can test fully before enabling.

Would that be reasonable? Note that we have bisectors on chromium.perf but NOT chromium.perf.fyi waterfalls.

I noticed https://chromium-review.googlesource.com/417182 was submitted referencing  bug 670657 . Is that fixing this bug or are there more work?

I spoke to Roberto and he said the support is there now on staging, and just needed a regression to confirm it on. You can test yourselves by finding bisecting via the dashboard and making sure to hit the "launch on staging bots" checkbox before submitting the bisect.

I looked for a recent regression and found https://chromeperf.appspot.com/group_report?bug_id=670657 but for that one, a bisect job (https://chromeperf.appspot.com/buildbucket_job_status/8993446045747445264) already seemed to have been fired on the staging bots (according to the results).

Since that job failed, I tried firing another one here:
https://chromeperf.appspot.com/buildbucket_job_status/8991857145763570224

Did the previous one not work because the auto-bisect support was not yet fully deployed, or were we just unlucky in this particular regression?
It doesn't seem to be possible to see who fired it in the available metadata (which would be a useful feature).

re: #c9

Looks like it just couldn't reproduce the regression. There are often a # of reasons for that, slightly different way of running the test, bot environment, etc. In this case looking a bit closer, it looks like there was some weird mismatch of when the ref build ran, but over roughly the same range (starting at chromium:433560) they both jumped up, so most likely  crbug.com/670657  isn't a real regression and can be closed.

https://chromeperf.appspot.com/report?sid=905cf9e0cf2bcd2ffa6b3869f0d13755faf8790a431baa418d0f8e7720b07b8e

Both value and ref for reference.

We've been waiting many weeks for a regression, but not a single one has happened - is there another way to verify it works? Is there another way to verify it works, or could we just move it from staging to prod now?

No unfortunately there isn't any way to easily test these kind of changes at the moment beyond actually running it against a regression. Is there a reason you want to move this to prod right away?

No, we were just wondering if there was a way to move forward with it. I guess we'll have to be patient then.

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. If you change it back, also remove the "Hotlist-Recharge-Cold" label.

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

Should have been fixed long ago in recipe, and we've switched to Pinpoint which supports bisecting into 3rd party automatically.

Hmm. If I look at https://chromeperf.appspot.com/group_report?keys=agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDg3sHtkAgM and click on the alert, the Bisect button is greyed out. Why is that? Is it simple configuration to turn it on? I suspect the problem is that those tests are MANUAL_ and can't run on any bot.

What about this one: https://chromeperf.appspot.com/report?sid=6ff9918bf09a3859eefd6a4f4e91c40cc29927f45973e4ed8f2b75d6ba125262&rev=21983? It's from webrtc_perf_tests, which is a test binary we use ourselves. What would it take to enable bisecting for that? Note, this binary runs on the webrtc.org code, so this isn't a question about bisecting into WebRTC rolls into chromium.

Was bisect on these other masters ever supported in the past? The first example is running on ChromiumWebRTCFYI, second on WebRTCPerf. Bisect button appears on chromium.perf tests, such as https://chromeperf.appspot.com/report?sid=2cd98650720577a3aa4698e9c8ed8d723157e9c564cce5f36390071c3188f21d

We should file a new issue to track getting bisect working on those, if that's a goal.

Re #18: probably not, no. Sure, I'll file a new issue.