| Commit message (Collapse) | Author |
|---|
|
This avoids useless rebuilds, just because files like the `README.md`
changed.
|
|
|
|
|
|
|
|
This functionality is now available via the `coImport` feature in the
`mkByName` `nixLib` function.
|
|
|
|
This allows us to also keep markdown and other documents, that aren't
nix, formatted.
|
|
|
|
Other options, for example `git-bug webui --read-only` is just to bugged
to be useful.
|
|
|
|
|
|
The old `libreddit` subdomain still has redirection to avoid this
being a breaking change. But keeping the old subdomain is rather weird
considering their new name.
|
|
This also includes a rename into `redlib` because of upstream changes.
|
|
|
|
|
|
Lix does not support the newer `nix shell` shebang.
|
|
|
|
The current stable version has a bug with regard to parsing the current
reddit json responses.
|
|
|
|
|
|
|
|
|
|
to generate
|
|
|
|
This has been updated to provide a means to send the user to YouTube, if
no invidious instances are available.
|
|
|
|
|
|
The main page does sometimes load, but videos are still not playable.
This new path really checks, whether the instance works.
|
|
Otherwise, the mapping of uid/gid to user name or group name could
change between reboots, which would result in magically change
permissions.
We were already affected by this at some point, so just remove the
possibility of it happening again.
|
|
This option was newly added, as previously only on `fcgiwrap` instance
was run as root. We probably have not been affected by this, as our
`fcgiwrap` instance was already running as `git:nginx`.
Usage of the new options seems better either way, as they provide a
finer grained control over the user _each_ `fcgiwrap`ped service is
running at.
The security advisory: https://discourse.nixos.org/t/51419
|
|
Libolm is marked as insecure and must thus be removed from the system
closure.
|
|
|
|
Filtering regions limits our possible instance selection without
actually providing great value. Let's stop discriminating based on
server location.
|
|
|
|
The 2GB are just not enough.
|
|
|
|
|
|
Take for example a repository name like: `some/organisation/project_a/team_c/repo_b`.
Setting the setting to `-1` means that cgit traverses the path from left
to right, until it has found 1 element (and `section-from-path` (or n
for short)is 0, because n is incremented after each iteration). E.g. :
~ [n=-1]
starting point: `some/organisation/project_a/team_c/repo_b`
~ [n=0]
after the first iteration: `some/organisation/project_a/team_c/repo_b`
Now `some/organisation/project_a/team_c` becomes the section, whilst
`repo_b` becomes the repo name.
|
|
The rust code does not compile, resulting in a really bothersome wait
every time a command is run in the devshell because direnv tries to
build it again.
|
|
|
|
Cgit effectively splits the repo path on '/' and then takes
`section-from-path` segments, which form the section. A negative value
here results in cgit traversing the path from left to right instead of
right to left. Beware that cgit only sets the section, if the path
contains `section-from-path` or more slashes in it (thus rendering this
setting defunct with the previous value of 1000).
There seems to be no way to tell cgit to always use all components up-to
the second to last for the section name, thus requiring all projects
that need a longer than 1 section length to set the `cgit.section`
git config variable via gitolite.
|
|
The previously set variable is only used in the gitolite.conf file for
the `config` specifications on each repo. We can't use that because we
use "wild-repos". Thus we need to add the `user-configs` option to each
repo, allow users to change the git settings specified there with a
simple `ssh git@git.vhack.eu config <repo> --set cgit.owner <name>`.
|
|
|
|
|
|
|
|
|
|
The last line, testing for the about page rendering is not yet working.
I assume that's because of our hand-rolled list-to-string function thingy
in the `cgitrc`. After <https://github.com/NixOS/nixpkgs/pull/317293> is
merged, this should probably improve immensely.
|
|
A lot of the added settings here have been tested. They will get
get tests to ensure they stay working, in later commits.
|
|
|
|
The modules still need to be imported one-by-one in `default.nix`
files because we have yet to rewrite this, but the new test
infrastructure might as well use the new `fileset` functions from `nixpkgs`.
|
