After updating to use the v4 graphql api in the merge tooling, the githubTargetBranch
was set to be the sha of the latest commit rather than the branch name of the target
branch. This caused our tooling to mismatch which branches were actually being targeted
with the effect that if a PR targeted only the patch branch (i.e. labeled `target: patch`
and targeting `11.2.x` in github)it would still expect to merge into both `11.2.x` and
`master`. This is now corrected to once again use the branch name, restoring to the
previous functionality.
PR Close#41523
Add a property, `targetLabelExemptScopes`, to the merge configuration allowing certain
scopes to be exempted from requirements for features and breaking changes only included
in PRs targetting certain labels.
PR Close#41459
Check commits in pull requests to ensure the pr can be merged into the target branch. Confirms
that prs targeting minor do not contain breaking changes, and prs targeting patch or lts do not
contain breaking changes or `feat` commits.
PR Close#41459
As the `test` and `refactor` commit types are not used in release notes and there
are solid use cases for having multiple scopes/scopeless uses of these types, they
are made to be optional instead of required on commits.
PR Close#41486
For commits from git log entries additional fields are available such as the reference
hash and author name, update the utility functions in commit-message to include the
parsed fields. Additionally define, per commit message type, whether to include the
commit in a release notes entry.
PR Close#41458
Create a utility method for the latest git tag, sorted by committerdate, which
matches SemVer, representing the latest version released on the branch.
PR Close#41455
Due to an issue with wombat proxy returning the login state of the generated tokens,
we will need to require a login for all `ng-dev release publish` runs to ensure npm
login has occured.
PR Close#41422
Check a range of commits by retrieving the log files to be parsed with the expected
format for the parser.
This change is in part of a larger set of changes making the process for obtaining
and parsing commits for release note creation and message validation consistent.
This consistency will make it easier to debug as well as ease the design of tooling
which is built on top of these processes.
PR Close#41341
As discovered in #41316, commit body length checks should consider all of the non-header
content as the commit body rather than the conventional-commit-parser's current method
of considering everything after an issue/PR reference to be the footer.
PR Close#41367
Use conventional-commits-parser for parsing commits for validation, this is being done
in anticipation of relying on this parser for release note creation. Unifying how commits
are parsed will provide the most consistency in our tooling.
PR Close#41286
Previously all github releases created by the release tooling tagged
releases as `latest`. Instead releases which are created for the `next`
tag on NPM should be tagged as `prerelease` for github.
PR Close#40999
Clearing the cached bazel outputs before building the artifacts for
publishing prevents an intermittent error found when the version is
cached between publishes.
PR Close#41000
Write to the unique log file, to prevent being overwritten, for `FATAL_ERROR`
failures in the release tooling. This will help to assist in determining where
something goes wrong in the process as well as being able to resume the action.
PR Close#40524
Add support for a `--no-branch-prompt` flag for the `ng-dev pr merge` tool. This
flag enables suppression of the confirmation prompt for which branches the specified
PR will merge into.
PR Close#40528
When multiple target labels are applied to a PR, it should be considered
invalid as our tooling does not support a single PR targetting multiple
trains/versions.
PR Close#40156
Delaying the check if a closed PR was closed by a merge or just closed by 30 seconds
allows for Github to have time to update the PR to be associated to the commit which
closed the PR. Without this delay, a race condition can exist in which we check for
how a PR was closed before this association is made.
PR Close#40181
When building the environment stamp, support two modes: release and snapshot
The release mode will always stamp using the current version of in the root package.json
and in snapshot mode will use a version stamp expressing a version based on the tag and
the number of commits from the tag.
PR Close#40095
With the migration to husky@5, the environment variable used to skip husky changed from
`HUSKY_SKIP_HOOKS` to `HUSKY`. To continue skipping husky during the merge process as
expected, the environment variable used is updated.
PR Close#40025
Previously, when a PR which does not target the master branch in the Github UI was
merged it would not close automatically. This change detects when this case occurs
and closes the PR via the Github API.
For example:
A PR which targets the 11.0.x branch in the Github UI has the `target: patch` label
This PR is only pushed into the 11.0.x branch, which does not trigger Github's
reference based actions to close the PR.
PR Close#39979
Allowing command line arguments to provide the file and source values to
the restore-commit-message command will assist in the the process of
upgrading to husky@5.
PR Close#39739
When fetch is run in normal mode, the `git-fetch-pack` plumbing command
outputs progress to stderr. Since this is unnecessary progress
information for ng-dev usages, it should be suppressed instead.
PR Close#39503
Some usages of the `GitClient` are better served by suppressing the
logging of lines that express what commands are being run. Many usages
of `GitClient` are contained within tools which are best served by
keeping the output clean as mostly read actions are occurring.
PR Close#39474
The node hash bang was incidentally removed in the published ng-dev, it should
be included to allow for the command to be run without having to specify node.
PR Close#39443
Rather than running ng-dev via ts-node, going forward ng-dev is generated and run
locally via node. Additionally, the generated file is tested on each commit to
ensure that the local generated version stays up to date.
PR Close#39089
Joey Perrott
Alan Agius
Kristiyan Kostadinov
Alexey Elin