When the project limit is reached the user is not allowed to create new ones.
Instead of error code 404 the status code 403 (Forbidden) is returned with error
message via API.
Extracted a method for 400 error (Bad request) and adjusted code accordingly. The name of
the missing attribute is used to show which one was missing from the request. It is used to
give an appropriate message in the json response.
When using project snippets via API the functions now provide status codes for
different situations other then only returning 404 error. If required parameters are missing,
e.g. `title` when creating a project snippet a 400 (Bad request) error is returned. The snippet
delete function now is idempotent and returns a 200 (Ok) regardless if the snippet with the
given id is available or not. Changing return codes of these functions has the advantage that
the 404 error is used only for resources, which are not available.
Tests added to check these status codes when handling project snippets.
Different status codes in the API lib are returned on hook creation, update or deletion.
If a required parameter is not given (e.g. `url` in `/projects/:id/hooks/:hook_id`) status
code 400 (Bad request) is returned. On hook deletion a 200 status code is returned, regardless if
the hook is present or not. This makes the DELETE function an idempotent operation. Appropriate tests
are added to check these status codes.
Mac OS uses launchd instead of /etc/init.d to start daemons and tasks to be started by launchd MUST NOT daemon itself. So "nohup" here won't work for Mac OS.
Can we add a "launchd" task to the rake file so that we can start sidekiq as "bundle exec rake sidekiq:launchd" ?
When adding a project hook a url must be specified or a 400 error code is returned
* Specs added to check status code on handling project hooks
* refactored code, extracted a method
When a user is added to a project that is already a member of, a status code 201 is now returned to
signal an idempotent operation. If something fails then instead of returning error code 404 different
more specific error codes are returned. Status code 400 (Bad request) is returned when a required
attribute, e.g. `access_level` is not given or 422 if there is a semantic error, e.g. should
the `access_level` have an unsupported value.
Specs are added to check these status codes.
If a note is created with a POST request via API (`/projects/:id/notes`) status
code 400 is returned instead of 404. The resource itself exists but the request
is incomplete. Specs added to check different status codes when accessing, creating
and updating notes.
If a milestone is created via API but no title given then status code 400 (Bad request)
is returned instead of 404. A small helper method handles the errors collection of a
milestone.
Creating a comment to an existing merge request via API without providing a note
returns a status code 400 now, suggesting a bad request. The reason for this
is the resource itself (MR) exists but the required property is not set.
Using the API library to create or update a merge request at the moment a 404 error is returned.
This is fine when the merge request in question does not exist, but does not provide good
information that for example a required attribute is missing.
A status code of 400 (Bad request) is returned when creating or updating a merge request
when either `source_branch` or `target_branch` is missing. A status code of 409 is returned
when `source_branch` and `target_branch` are the same. Tests are added for these cases.
Author date is not updated, if the commits is rebased.
So the network graph having many rebased commit turns round and round,
that it is very difficult to undarstand history.
The attempt to revoke project access for a user that was not member of the
project results in a 500 Internal Server error where it actually should
result in a 200 OK since after the operation, the user is not member of
the project. This turns the operation into an idempotent call that can
be repeated with no ill effects.
Updated the spec and changed the code accordingly. However, the result differs
slightly, as we can't return the users project access level if the user was not
member. I'm not aware if anybody relies on the result of this call.
Fixes#2832
In case we rescue from a fatal error, we want the error and the backtrace to
the error logged, so we can debug later on. This change injects the configured
logger from the rails app to the grape API and logs error as well as backtrace
in a rails-like fashion.
A new queue of "project_web_hook" is used to process web hooks asynchronously, allowing each to succeed/fail (and be retried) independently.
(Basically, project web hooks now process the same as system hooks.)
