In a few models we define ActiveRecord enums that are redefined in EE
using the following pattern:
enum :some_enum, {
...
}.merge(EE_ENUM_VALUES)
This particular approach is problematic to deal with, because it
requires that we `prepend` and EE module _before_ defining the enum.
This typically translates to the `prepend` being the first line in the
model in EE, but this can easily lead to merge conflicts when developers
add more `include` and/or `prepend` lines.
As part of https://gitlab.com/gitlab-org/gitlab-ee/issues/8244 and
https://gitlab.com/gitlab-org/gitlab-ee/issues/8241 we are moving
`prepend` to the last line in a file, reducing the chances of running
into merge conflicts. This poses a bit of a problem with the pattern
above, because this pattern does not allow us to move the `prepend`
further down a file.
To resolve this problem, we simply move the Hash value of the enum to a
separate class method. This method is defined in a separate module where
necessary, allowing us to use it like so:
enum :failure_reasons, ::SomeModelEnums.failure_reasons
The method in turn is defined in a very straightforward manner:
module SomeModelEnums
def self.failure_reasons
{
...
}
end
end
This makes it easy for EE to add values without requiring the `prepend`
to be placed before the `enum` is defined.
For more information, see the following issues and merge requests:
* https://gitlab.com/gitlab-org/gitlab-ee/issues/8244
* https://gitlab.com/gitlab-org/gitlab-ee/issues/8241
* https://gitlab.com/gitlab-org/gitlab-ee/merge_requests/8424
Enum in Rails 5 does not map nil values - IOW nil value remains nil,
even if there is a key with nil value in the enum definition.
This commit overrides the underlying Enum methods so nil value is
still mapped. This solution is far from being ideal: it uses dynamic
definition of methods which introduces more magic/confusion
into the codebase.
It would be better to get rid of the nil value in enums.
We still didn't migrate `ci_builds.stage_id`, so we can't use a
belongs_to association. We also have `ci_builds.stage` string attribute,
that we need to phase out in favor of `ci_stages.name`.
Older pipelines can contain jobs without a name, in which case
'nameless' is used right now, so we can properly still handle other
actions on these jobs.
Fixesgitlab-org/gitlab-ce#33403
* master: (524 commits)
Improve user experience around slash commands in instant comments
Fix LFS timeouts when trying to save large files
retryable? is now available for CommitStatus
Resolve "Documentation of `.gitlab-ci.yml` states incorrect example for variables"
Fix test failures
Add slugify project path to CI enviroment variables
Fixed typo: craeted -> created
32118 Make New environment empty state btn lowercase
Expose import_status in Projects API
32832 Remove overflow from comment form for confidential issues and vertically aligns confidential issue icon
Fix test failures
Allow manual bypass of auto_sign_in_with_provider
Fix keys seed
Allow users to be hard-deleted from the API
fixup some classnames and media queries
Enable the Style/PreferredHashMethods cop
Lint our factories creation in addition to their build
Don’t schedule workers from inside transactions
Allow scheduling from after_commit hooks
Forbid Sidekiq scheduling in transactions
...
Conflicts:
app/serializers/pipeline_entity.rb
db/schema.rb
spec/factories/ci/stages.rb
spec/lib/gitlab/import_export/safe_model_attributes.yml
spec/services/ci/create_pipeline_service_spec.rb
spec/spec_helper.rb
Yorick Peterse
Kamil Trzciński
George Tsiolis
Shinya Maeda
Rémy Coutable
gfyoung
Jan Provaznik
Mayra Cabrera
blackst0ne
Tomasz Maczukin
Zeger-Jan van de Weg
Takuya Noguchi
Grzegorz Bizon
Douwe Maan
Filipa Lacerda
James Lopez
Douwe Maan
Regis Boudinot