Commit Queue (CQ)

What (is it)?

It's a service (aka a bot) that commits Gerrit changes for you, instead of you directly committing the change. We support a wide variety of projects already and can support your project too (see internal docs).

For enabled projects, we display a CQ checkbox/button on Gerrit interface.

How (does it work)?

The commit queue is not really a queue at the moment, since it processes the changes out of order. This may be changed eventually. This means a CL can be committed before another CL that was triggered much earlier. This can happen when a try job is flaky.

Current process for the user

  1. Upload a review to Gerrit where it gets reviewed and LGTM'ed.
  2. One of:
    1. In Gerrit, click “REPLY” and then under “Commit-Queue”, select +2 and then click “SEND”.
    2. At the command line, type git cl set_commit.
  3. After a few seconds, CommitBot should post a message to the review saying something like “CQ is trying the patch”. Tryjobs will show up on the Gerrit UI.

Why (is it broken)?

Please follow these general guidelines:

  1. Please report issues to chrome-troopers.
  2. If you have a feature request, feel free to file a bug, use label Build-CommitQueue. Be sure to search for current feature requests first.

Options

You may include the following options into the desription of your CL.

COMMIT=false

If you are working on experimental code and do not want to risk accidentally submitting it via the CQ, then you can mark it with COMMIT=false. The CQ will immediately abandon the change if it contains this option. To dry run through the CQ please use Gerrit's dry run feature.

TBR=<username>

This stands for “to be reviewed”. If a change has a TBR line with a valid reviewer, the CQ will skip checks for LGTMs. See guidelines of when it's acceptable to use this.

NOPRESUBMIT=true

If you want to skip the presubmit check you can add this line and the commit queue won‘t run the presubmit for your change. This should only be used when there’s a bug in the PRESUBMIT scripts. Please check that there‘s a bug filed against the bad script, and if there isn’t, file one.

NOTRY=true

This should only be used for reverts to make the tree green, since it skips try bots and might therefore break the tree. You shouldn't use this otherwise.

NOTREECHECKS=true

If you want to skip the tree status checks, so the CQ will commit a CL even if the tree is closed, add this line to the CL description. Obviously this is strongly discouraged, since the tree is closed for a reason. However, in rare cases this is acceptable, primarily to fix build breakages (i.e., your CL will help in reopening the tree).

NO_DEPENDENCY_CHECKS=true

The CQ rejects patchsets with open dependencies. An open dependency exists when a CL depends on another CL that is not yet closed. You can skip this check with this keyword.

CQ_INCLUDE_TRYBOTS=<trybots>

This flag allows you to specify some additional bots to run for this CL, in addition to the default bots. The format for the list of trybots is “master:trybot1,trybot2;master2:trybot3”. This feature only works for recipe based bots right now.

FAQ

Is the CQ broken?

Take a look at https://codereview.chromium.org/search?closed=3&commit=2&limit=100&order=modified. If there are issues older than ~4 hours, they could probably be stuck. Note that the Commit Queue could be stuck only for some issues but not all of them. In case of doubt, please contact chrome-troopers.

The CQ seems hung

Is the tree open? It commits 4 CLs every 8 minutes, so a maximum rate of 30 commits per hour.

Please Help! I just want to ask on irc!

Please report issues to chrome-troopers.

My patch failed to apply

See the Try Server FAQ.

What about LKGR?

The Commit Queue has never known, used or cared about LKGR. It always uses HEAD, the tip of tree.

What's my position on the queue?

The CLs are processed out of order, so it‘s not because another is “before” yours that means it’ll be committed before yours. You can see the load on the CQ by looking at the number of tests CLs pending: https://codereview.chromium.org/search?closed=3&commit=2&limit=1000&order=modified

Sending a TBR patch fast

You can't wait for review? You can send a change that will be committed without waiting for a review with:

git fetch origin
git new-branch work_fast
# Quick, write your fix.
echo "A copy is available for 100000$USD upon request." >> LICENSE
git commit -a -m "Fix the license, show new opportunities

TBR=danny@chromium.org"
git cl upload --send-mail -c

This‘ll still check for try jobs; see the next section if you can’t wait for them, either.

The important part is to have TBR=foo@chromium.org in the CL description.

  • --send-mail sends an email right away.
  • -c sets the commit bit right away, short for --use-commit-queue.
  • --cc joe@chromium.org,hppo@chromium.org to cc more people instead of putting everyone as reviewer.

Now, did you know there's git cl help upload?

Sending CL through CQ without committing (dry run)

To dry run through the CQ please use Gerrit's dry run feature.

Picking custom trybots

See the CQ_INCLUDE_TRYBOTS option, above.

Try job results aren't showing up consistently on Gerrit

If you never had a HTTP 500 on GAE, chances are that you will.

Binary files?

Yes, binary files are supported by try jobs as well as CQ now!

My CL has a bazillion files, will it blend?

The CQ was able to commit a CL with 838 files so it is technically possible: https://codereview.chromium.org/12261012/. The likelihood of the CQ failing increases exponentially with the number of files in the CL.

Are CQ users required to be around when the patch is landed?

In general, no, as the CQ can land at any time (including very long time), and any breaking patches can be kicked about by the sheriff. After all, that's the job of the sheriff. You will get an email when the CQ commits, so you can jump on your nearest laptop if necessary.

If you expect your patch to be hard to revert, is touching several files and directories or move files around, you may want to stay around in case there is an incremental build failure or something hard to diagnose.

Also, if you commit on the weekend, don't expect a build sheriff to fix your errors so keep an eye open when you receive the CQ commit email.

What determines the set of tests and targets the try bots run?

This is controlled by a config file cq.cfg (e.g. config for chromium). Also see this document for details on the analyze step.