scripts: add script to count pushes to repositories Laura wanted this data. The script is hacky but gets the job done.

Bug 1139661 - mozreview: Fix link to BMO in error report. r=smacleod

testing: log in newly-created Bugzilla users (bug 1164688); r=smacleod Bugzilla recently made it so users with no activity in the past 60 days can't have review flags aimed at them. See bug 751862. We change the user creation code to log in the user by default to simulate Bugzilla activity. This is easier than modifying every test to touch newly-created users. This required a bit of refactoring to make the login code reusable.

ansible: add variable for djblets and review board package versions

MozReview: Upgrade to Review Board 2.0.15 (Bug 1160176) Review Board site upgrades print any manual steps after "Upgrade Complete!". During deploy print site upgrade stdout when "Upgrade Complete!" is not the last line of site upgrade.

MozReview: Disable RBTools API caching to avoid errors (Bug 1160176) RBTools 0.6.1 has known bugs in its API cache implementation. Some newer Review Board versions (e.g. 2.0.15) cause our tests to fail with errors from the RBTools API cache. Temporarily disable the API cache until we can upgrade to a newer (currently unreleased) version of RBTools and better investigate the cache problems.

testing: add a decorator to restart MozReview between test methods This was requested by mdoglio. While we were here, MozReview cluster instances now track their state and prevent simultaneous starts. This should help prevent a footgun where people could start multiple instances of a cluster, completely confusing state.

testing: add --use-last-tests to run-tests to make tests run faster Generating Docker images on every test run adds considerable overhead, especially when developing and debugging tests. In many scenarios, the images won't have changed and generating new images just adds overhead. Add an option to recycle images from last time. This option is a footgun. A lot of files are copied to containers/images. Improper use could result in changes not taking effect in the test environment. Use with caution. On my machine, this reduces test start overhead from 20-25s to ~4s.

Reverting accidental checkin.

Bug 1055021 - mozreview: Prevent ship-its in the UI from parent review requests. r?smacleod The required hook isn't yet available in Review Board, so this just uses CSS rules to hide the ship-it options.

Bug 1055021 - mozreview: Create attachments for all children. r?smacleod An attachment is now created on Bugzilla for each child. No attachment is created for the parent. Review logic has changed a bit as well. For children, a ship-it results in an r+ on the corresponding attachment from the reviewer. A review with no ship-it clears any '?' review flag with the requestee being the reviewer, and any '+' with the setter being the reviewer. For parents, reviews with ship-its are not allowed, and the corresponding UI functions have been removed. A review with no ship-it results in a comment unassociated with any attachment, and no flag changes.

Bug 1055021 - mozreview: Include reviewer BMO ids in summary API. r?smacleod

mqext: make 'reviewers' and 'components' accept revisions (bug 1160440) r=gps

testing: support Ubuntu Ansible containers We want to use Ansible inside Ubuntu containers. Create a base image for it.

autoland: autoland should strip draft revisions after pushing to try (bug 1159822) r=gps If we don't strip draft revisions, we'll end up with many, many heads that will eventually make mercurial unhappy.

testing: properly unicode handling when writing xunit report (bug 1141720) This is essentially 7b0a20cd8c95 from upstream, which is a patch I contributed to Mercurial after this was reported at Mozilla.

add explicit support for Mercurial 3.4 (bug 1162304); r=smacleod Mercurial 3.4 was released a few days ago on May 1. We need to advertise Mercurial 3.4 support in extensions or Mercurial may think the extension is incompatible. Do it. A test run verifies that extensions continue to work with Mercurial 3.4.

drop explicit support for Mercurial 3.0 (bug 1162304); r=smacleod Mercurial 3.0 is now over a year old. People have had plenty of time to upgrade to 3.1 or newer. We're about to introduce official support for 3.4. So if nothing else, let's drop 3.0 support so we're testing on one less configuration.

MozReview: Stop generating parent diffs (Bug 1159860); r?gps MozReview requires parent diffs be generated and uploaded when the repository Review Board is configured with does not contain the commit a diff is based on. Review Board will always be pointed at the review repository itself meaning it will contain the base commit for every diff (squashed and commit) - parent diffs are not needed. Stopping the use of parent diffs has the following effects: - MozReview to use less space in the database (*Much* less in requests with many commits as every commit parent diff contains the commits before it). - Diff application and generation is faster - 1/2 to 2/3 as many applications are required and diff sizes are smaller. - MozReview requires more requests to the hg server to fetch base files. Using parent diffs the same base files are used for every request, without them each commit request has it's own base commit and files modified in several commits will not be cached (Fetched files are cached by sha/filename). Also, provide a base_commit_id with each diff. Default diffs generated by hg include information about revisions the diff is generated between but git style diffs do not. Starting in version 2.0.16, Review Board will properly parse git style diffs and take base_commit_id into account. Switching to git style diffs will require this base_commit_id be provided. Alternatively "hg export" style headers could be included in diffs to tell Review Board which commit to use as a base - this is unnecessarily complex as the same Review Board version which supports git diffs fixes base_commit_id support.

MozReview: Allow specifying base_commit_id and make parent_diff optional (Bug 1159860). r?gps In future releases of Review Board base_commit_id will be respected by the mercurial SCMTool allowing us to properly set a diffs base. We now allow specifying the base_commit_id as part of the commits structure passed to post_reviews(). Parent diffs are also made optional. Originally they were mandatory since Review Board was going to point at the mainline repository, rather than the review repository - meaning we could not apply just the desired diff as the base was unknown to Review Board. In reality Review Board is pointing at the review repo itself, which will be mandatory going forward.