Chromium Commit Queue

Contents

  1. 1 What (is it)?
  2. 2 Design document
  3. 3 How (does it work)?
    1. 3.1 Current process for the user
  4. 4 Why (is it broken)?
  5. 5 Options
    1. 5.1 Is the CQ broken?
    2. 5.2 The CQ seems hung
    3. 5.3 Please Help! I just want to ask on irc !
    4. 5.4 My patch failed to apply
    5. 5.5 What about LKGR?
    6. 5.6 Where is the dashboard?
    7. 5.7 What's my position on the queue?
    8. 5.8 Sending CL through CQ in dry-run mode
    9. 5.9 Picking custom trybots
    10. 5.10 Try job results aren't showing up consistently on Rietveld
    11. 5.11 Binary files?
    12. 5.12 My CL has a bazillion files, will it blend?
    13. 5.13 Moving, renaming or copying files
    14. 5.14 Are CQ users required to be around when the patch is landed?
    15. 5.15 What determines the set of tests and targets the try bots run?
See also: Chromium OS Commit Queue.

What (is it)?

It's a service (aka a bot) that commits rietveld changes for you, instead of your directly committing the change. It is enabled for the following projects:
...but not GYP, for instance. This list is not complete, though, and may grow over time. Rietveld will show a 'Commit' checkbox if the CQ supports to land the CL.

Design document

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 rietveld where it gets reviewed and LGTM'ed.
  2. One of:
    1. Check (select) the 'Commit' checkbox on a Rietveld issue, on the particular patchset that has been approved. Checking the checkbox is the only action required and there will be no immediate feedback in the form of messages, etc, to notify you that something has happened.
      • Only the issue owner, someone at @chromium.org or @google.com can check the box.
      • Yes, non-Chromium committers are allowed to use the commit queue but cannot LGTM a change.
    2. At the command line, type git cl set_commit
    3. Have a reviewer use 'Quick LGTM & CQ'.
  3. Wait an hour. The current list of patches to be queued can be found at Commit Queue Patches, while CQ progress can be tracked at Commit Queue Progress. The commit queue will wait automatically for the tree to reopen.
  4. Wait for an email from commit-bot@chromium.org with success or failure.

Why (is it broken)?

Please follow these general guidelines:
  1. If compilation or individual tests are failing on one or more CQ bots and you suspect that your CL is not responsible, please contact your friendly neighborhood sheriff or file a sheriff bug. If the code in question has appropriate OWNERS, consider contacting or CCing them.
  2. If other parts of CQ bot execution (e.g. bot_update) are failing, if you have reason to believe the CQ daemon itself is broken, or if you can't really tell what's wrong or why something is failing, please file a trooper bug according to the process documented here.
In both cases, when filing bugs, please include links to the build and/or CL in question.

If you have a feature request, feel free to file a bug, use label Infra-CommitQueue. Be sure to search for current feature requests first.

Options

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 Rietveld's dry run feature.

TBR=<username> 
See policy of when it's acceptable to use TBR ("To be reviewed"). If a change has a TBR line with a valid reviewer, the CQ will skip checks for LGTMs.

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 green the tree, 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.

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, contact commit-bot@chromium.org.

If your CL hasn't been touched after a few minutes of checking the CQ bit, CHECK THE PROJECT FIELD ON YOUR ISSUE. If there is no project field, it will be ignored by the Commit Queue.

The CQ seems hung

Is the tree open?
It commits (if not configured differently for a project) 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

What about LKGR?

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

Where is the dashboard?

It's at https://chromium-cq-status.appspot.com/. You can follow the link posted to your CL to see the testing progress.

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:

Sending CL through CQ in dry-run mode

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

Picking custom trybots

See the CQ_INCLUDE_TRYBOTS option, above.

Try job results aren't showing up consistently on Rietveld

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

Binary files?

Sorry, unsupported. Some known issues include bug 579778, bug 395626, and bug 375734.

For reference, older bugs related to this include bug 339068 and bug 23608.

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.

Moving, renaming or copying files

Was implemented in bug 125984 and bug 125983. If the diff on Rietveld doesn't look right, use the --similarity (defaults to 50%) and disable/enable file copy with --find-copies/--no-find-copies. In case of confusing;

git cl help
man git diff

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 back out your crap 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 (chromium_trybot.json for most trybots). Also see this document for details on the analyze step.