Follow-up of #5953.
Refactor the concerns with two goals:
- Getting closer from the way ActiveStorage adds its own hooks.
Usually ActiveStorage does this using an `Attachment#after_create`
hook, which then delegates to the blob to enqueue the job.
- Enqueuing each job only once. By hooking on `Attachment#after_create`,
we guarantee each job will be added only once.
We then let the jobs themselves check if they are relevant or not, and
retry or discard themselves if necessary.
We also need to update the tests a bit, because Rails'
`perform_enqueued_jobs(&block)` test helper doesn't honor the `retry_on`
clause of jobs. Instead it forwards the exception to the caller – which
makes the test fail.
Instead we use the inline version of `perform_enqueued_jobs()`, without
a block, which properly ignores errors catched by retry_on.
Test helpers are separated between two files: spec_helper and
rails_helper. This separation is meant to allow tests that do not
require Rails (like testing standalone libs) to boot faster.
The spec_helper file is always loaded, through `--require spec_helper`
in the `.rspec` config file. When needed, the rails_helper file is
expected to be required manually.
This is fine, but:
- Many test files have a redundant `require 'spec_helper'` line;
- Many test files should require `rails_helper`, but don't.
Not requiring `rails_helper` will cause the Rails-concerned section of
the test environment not to be configured–which may cause subtle bugs
(like the test database not being properly initialized).
Moreover, Spring loads all the Rails files on preloading anyway. So the
gains from using only `spec_helper` are thin.
To streamline this process, this commit:
- Configures `.rspec` to require `rails_helper` by default;
- Remove all manual requires to spec_helper or rails_helper.
Reference: https://stackoverflow.com/questions/24145329/how-is-spec-rails-helper-rb-different-from-spec-spec-helper-rb-do-i-need-it
Validate that a dossier on a `for_individual?` procedure always has
an `individual` associated record.
For this, the individual needs to be built before the record is
validated (i.e. even before the `before_create` callback is run).
This should help with #4596: now if a dossier is created without an
`individual`, or if the `invividual` record is later removed, the
validation will fail.
The code paths for deleting a dossier were different, depending on
whether the dossier was deleted by the user, or from the Manager.
This commit unifies the two code paths into one.
This has the effect of:
- An operation log is now recorded when an user deletes its own dossier;
- Gestionnaires are now notified even when the dossier is deleted from
the Manager;
- The `support:delete_user_account` task now requires the email address
of the author.
An Admin may archive a procedure to make it unavailable to
the general public, but before all dossiers are handled. In this case,
the messagerie needs to be available.
Fix#4089
* Use the existing Dossier#messagerie_available? method
* Raise when attempting to build a Commentaire if not messagerie_available?
* Disable the Messagerie form if not messagerie_available?
* Add tests :)
* Tweak the Horaires formatting while we’re here.
The `nearing_end_of_retention` test creates a `just_expired_dossier` “six months ago”, which is 29 Oct 2018. The autumn DST change was on October 28; this is the first time this test runs under these conditions. We workaround the time offset by creating the dossier one hour earlier.
This is technically a workaround, not a fix: the date arithmetics in `nearing_end_of_retention` are probably wrong. It looks like it’s comparing intervals, which seems error-prone, while it should be comparing dates. For now, I’m just making the tests pass.
