Try Server and Commit Queue newsletter

[Marc-Antoine Ruel]

TL;DR; This email is FYI.

Today the TS was restarted, bringing goodies

- The prioritization of try jobs from the CQ, which was added last year but broke by the Rietveld integration, was fixed again. The CQ won't hang due to a single backlogged builder even if someone decided to bisect browser_tests on 56 revisions on mac_rel in the middle of the day.

- Thanks for Nicolas and Bev and the team, we got more mac now so the huge backlogging problem should be resolved for mac/mac_rel.

- You can now git cl try -b linux_rel:compile to try compiling "all" on a specific builder. Note that the CQ was already specifying "compile" on win, mac, linux_clang, so the CQ will again ensure all the targets can be built. I had to temporarily disable ios_rel_device because the Chrome on iOS project typoed 'all' as 'All'.

- The CQ sometimes forget your CL. It's a known issue and I'll try to fix it ASAP. You can work around by unchecking the bit for a few minutes* so it completely reinitialize the state of your CL. *The amount of time can be as low as 10 seconds, it depends how much activity is going on.

- The Try Server <-> Rietveld integration is occasionally flaky, help is wanted (http://crbug.com/160709). I have to fix these manually for now, please email commi...@chromium.org for any suspicious behaviour.

Test isolation update

- Test isolation is progressing well; we got browser_tests_swarm green on linux; it's taking a few commits but we're getting near having it green on Windows.

- Only 2 test cases are failing on unit_tests (http://crbug.com/161481) and got a single test case regression since yesterday; then unit_tests will be green on both Windows and linux.

- Tracing on OSX is .. a challenge. Anyone good in D dtrace code around who could optimize this code? I'd appreciate some help with that.

Did you know?

- Thanks to Robbie, you can now tweak git diff rename detection; Look at the last 3 options in git cl help upload for more details. It was a top complain that the diff on Rietveld was invalid because the git diff heuristics was off so you can now tweak it!

- Contrary to popular belief, only git users can upload a patch to Rietveld and the CQ will properly handle file rename, move and copy and +x or -x executable bit change. No need for a svn checkout for any of these, just tweak the --similarity flag.

- You can use the CQ even if the tree is a mess. As a full committer and with a tree as stateful as these days, you can:

$ git commit -a -m "Fixes goat teleporter destination.

Stop sending them to New Zealand, it's overflowing.

TBR=f...@chromium.org

NOTRY=true

BUG="

$ git cl upload --send-mail -c

The CL will be sent with f...@chromium.org as a reviewer and the CQ will commit it on your behalf as soon as the tree opens. TBR bypasses OWNERS check but not PRESUBMIT checks. NOTRY bypasses the Try Server, especially useful for non-code changes.

- You want to set the CQ bit on a CL but that would mean reaching your mouse? No need! Use git cl set_commit.

- You want to read the CL's comments but that would mean reaching your mouse? No need! Use git cl comments.

Note that most commands have integrated help.

Thanks,

M-A

[Marc-Antoine Ruel]

2012/11/17 Marc-Antoine Ruel <mar...@chromium.org>

- You can now git cl try -b linux_rel:compile to try compiling "all" on a specific builder. Note that the CQ was already specifying "compile" on win, mac, linux_clang, so the CQ will again ensure all the targets can be built. I had to temporarily disable ios_rel_device because the Chrome on iOS project typoed 'all' as 'All'.

The problem with ios_rel_device was fixed, it was a misconfiguration on my part, so it's back on the CQ. 

 

Thanks,

M-A

[Thiago Farina]

On Sun, Nov 18, 2012 at 11:00 PM, Marc-Antoine Ruel <mar...@chromium.org> wrote:
> 2012/11/17 Marc-Antoine Ruel <mar...@chromium.org>

> The problem with ios_rel_device was fixed, it was a misconfiguration on my
> part, so it's back on the CQ.
>

Cool! Thank you Marc.

This was causing some noise this sunday ;)

--
Thiago

[Wez]

I'm getting tries failing on ios_rel_device trying to build All.  (See http://build.chromium.org/p/tryserver.chromium/buildstatus?builder=ios_rel_device&number=7855) :(

--
Chromium Developers mailing list: chromi...@chromium.org
View archives, change email options, or unsubscribe:
http://groups.google.com/a/chromium.org/group/chromium-dev

[Marc-Antoine Ruel]

... and fixed for real this time, and properly verified! Sorry for the inconvenience.

M-A

2012/11/18 Wez <w...@chromium.org>

[Lei Zhang]

And thanks for putting some the affected jobs back in the CQ.