The line `allowfilter = true` is inserted both in new Git repositories
and existing ones (via an UpdateStep). This enables clones with
`--filter` parameters.
Co-authored-by: René Pfeuffer<rene.pfeuffer@cloudogu.com>
With this change, most modifications of git repositories (like inserting, deleting and updating files and merging branches) do no longer work inside clones held in temporary working directories but are done directly inside the bare git repository data. This resolves in a massive performance boost for the editor plugin and pull requests, especially in larger repositories.
Co-authored-by: René Pfeuffer<rene.pfeuffer@cloudogu.com>
Committed-by: René Pfeuffer<rene.pfeuffer@cloudogu.com>
Document titles represent the pages, for example in lists
with bookmarks. They are important for navigation and
orientation in websites. If the offer or the content of the
page is not labeled, orientation is impaired.
This changes the behavior for setting document titles.
The functionality has been removed from the Page and
Title components and is now represented by `useDocumentTitle`
hook to better describe the content of inividual pages.
Co-authored-by: Anna Vetcininova<anna.vetcininova@cloudogu.com>
- Description of logout button is read once
- SCM-Manager logo is better decribed for assistive technologies
- The state of the mobile menu is better accessible
To still support the editor plugin, a new field in the
browse command results is needed to indicate, whether
such a result or rather the requested revision can be
modified by a new commit.
This is the case, when
- for Subversion repositories either the new 'head' or the
latest revision has been requested, or
- for Git and HG when a branch (or the default by specifying
no concrete revision) has been used.
Over the course of a very long time, the interpolation strings for the versions in the scm-git-plugin and scm-svn-plugin had not been processed. One assumption was that this was related to a discontinued Maven process.
The behavior was now realized by two tasks process after the gradle build lifecycle step. It remains to be discussed whether it's worth the time and effort to integrate the change into the smp plugin.
The caching for local storage seems way over-engineered. With this the context for the local storage is removed and the local storage is accessed directly.
Pushed-by: k8s-git-ops<admin@cloudogu.com>
Pushed-by: Rene Pfeuffer<rene.pfeuffer@cloudogu.com>
Co-authored-by: René Pfeuffer<rene.pfeuffer@cloudogu.com>
Originally, the state of the DiffDropDown was not coupled to the LoadingDiff and instead a toggle method was forwarded. This was causing problems due to their detachment, and so the state of the DiffDropDown was tied to the LoadingDiff one. The original boolean values have been replaced with strings, so that nuances between 'hide all whitespaces' and 'show all whitespaces' in future feature additions can be recognized without affecting the API a second time.
Co-authored-by: Till-André Diegeler<till-andre.diegeler@cloudogu.com>
Co-authored-by: Thomas Zerr<thomas.zerr@cloudogu.com>
Pushed-by: Thomas Zerr<thomas.zerr@cloudogu.com>
Pushed-by: Till-André Diegeler<till-andre.diegeler@cloudogu.com>
We encountered NullPointerExceptions during hg repository imports. This happened, when neither authentication nor proxy definions had been given. In this case, no hgrc will be created and therefore we get the exception in the clean-up step. Now, the clean-up is only triggered, if the hgrc had been created beforehand.
To prevent further error in the clean-up, we no catch other potential exceptions and log them but do not let them break the import.
Pushed-by: Rene Pfeuffer<rene.pfeuffer@cloudogu.com>
Co-authored-by: René Pfeuffer<rene.pfeuffer@cloudogu.com>
With this change,
- logs are only written to console when the server is started in foreground (prevents writing to scm-server.out file)
- the location of the log file is determined correctly, when it is configured with a relative path
- less logs are written to console on startup
- enabling file and console appenders via config.yml
The BaseReceivePackFactory re-used the GitReceiveHook. The problem is, that the hook is not thread safe. Due to the re-usage, the repository could have been changed during processing post receive hooks. With this, the factory will always create a new receive pack.
Pushed-by: Rene Pfeuffer<rene.pfeuffer@cloudogu.com>
Co-authored-by: René Pfeuffer<rene.pfeuffer@cloudogu.com>
Committed-by: René Pfeuffer<rene.pfeuffer@cloudogu.com>
