and request them each session. Pass these tokens to the project import data.
This prevents the need to encrypt these tokens and clear them in case they
expire or get revoked.
For example, if you deleted and re-created OAuth2 keys for Bitbucket, you would get
an Error 500 with no way to recover:
```
Started GET "/import/bitbucket/status" for x.x.x.x at 2015-08-07 05:24:10 +0000
Processing by Import::BitbucketController#status as HTML
Completed 500 Internal Server Error in 607ms (ActiveRecord: 2.3ms)
NameError (uninitialized constant Import::BitbucketController::Unauthorized):
app/controllers/import/bitbucket_controller.rb:77:in `rescue in go_to_bitbucket_for_permissions'
app/controllers/import/bitbucket_controller.rb:74:in `go_to_bitbucket_for_permissions'
app/controllers/import/bitbucket_controller.rb:86:in `bitbucket_unauthorized'
```
Closes#1871
Do not delete the SQL dump too early
The change in baa157926d broke backup
restore fucnctionality. This would not lead to data loss, but it
prevented the restore script from working. This bug exists only in
7.14.0 release candidate versions, not in 7.13.
Reported in https://github.com/gitlabhq/gitlabhq/issues/9571 .
See merge request !1182
The change in baa157926d broke backup
restore fucnctionality. This would not lead to data loss, but it
prevented the restore script from working. This bug exists only in
7.14.0 release candidate versions, not in 7.13.
Reported in https://github.com/gitlabhq/gitlabhq/issues/9571 .
Include more information in hook error messages
When pushing commits to the remote, if a pre-receive hook gets triggered that prevents the push, it may not be clear which commit is causing the issue. Having the values of what caused the comparison to fail in the error message will help debugging, whether it be the user or the administrator.
See merge request !14
Make the `provider` argument for the LDAP group links API endpoint required
Related to #329
If no `provider` is passed to the API endpoint, the group link is not created properly. Making it required fixes the issue temporarily. It should default to the first provider.
Next step, as mentioned in #329, is to do that.
/cc @job
See merge request !475