Redmine 5 #813
Redmine 5 #813
Conversation
|
@alexandermeindl, my very naive attempt to test on Redmine 5 failed miserably for a whole lot of reason. By any chance have you upgraded to Redmine 5 already ? or is it something you might be working on soon-ish ? |
|
Hi @PowerKiKi, no, sorry. I did no work for upgrading to Redmine 5 and currently it is not planned, too. |
|
Hi |
I fixed documentation for the right version: http://redmine-git-hosting.io/get_started/ |
|
Hi @PowerKiKi , did you by any chance make it work with Redmine 5 with your latest commits? @alexandermeindl did you change your mind since June? :) |
…ents Ruby 3.0 does not implicitly destructure a Hash into keyword arguments if it was the last parameter to a method call. Also handle explicitely delegate keyword arguments.
Seems we only want an optional hash not keyword arguments so remove the double sprat and set a default empty hash value.
…Cache>::Timezone Typo Timezone.new instead of Time.zone.now
|
Hi, |
This new gitolite-rugged version brings rugged v1.5.x instead of v1.1.x, which allows gitolite-rugged to specify credentials when pushing to a git (which is expected for it to work correctly). By the way: * pkg-config is necessary to build rugged native C extensions; * starting from v1.5, rugged expects an explicit setting to build with SSH support, which is expected by redmine_git_hosting to work correctly (see: https://github.com/libgit2/rugged/tree/v1.5.0#options)
|
|
|
@dosyfier on this branch, which also contains your changes, I get the following error when accessing http://localhost:3000/settings/plugin/redmine_git_hosting. Is it working for you ? |
No, I didn't get that one (and I was able to go to redmine_git_hosting settings page). I'll try again next weekend. |
|
Something worth noticing though: by upgrading rugged lib to v1.5.0.1, I had to add the following option to bundler before running bundle config build.rugged --with-sshCf. commit b18ebcf. |
The block registered by redmine_git_hosting plugin to Rails.application.config.to_prepare wasn't actually called (at least, in a test context). This appears to be due to the fact that the plugin's init.rb file is already loaded from within a Rails "to_prepare" initialization event block. Cf. Redmine's PluginLoader class and its "load" method defined in lib/redmine/plugin_loader.rb file: a Rails "to_prepare" block iterates over all plugin directory to find and load an "init.rb" file. As a consequence, there's no need to rewrap some code from the plugin's init.rb file into a Rails "to_prepare" event handling block.
Maybe due to upgrading to Ruby 3.x, the code of these modules couldn't
be loaded anymore, due to an infinite loading loop.
For instance, for the Commands module:
* Loading of commands.rb is triggered,
* As its first instruction, the module extends Commands::Base,
* So the loading of commands/base.rb is triggered,
* But then, in that file, the Base module is included in the
Commands module, which has to be loaded first,
* And so, we're back to the beginning...
To work around this, the "extend Xx::Xx" instructions are moved to the
submodules themselves, so that each submodule can be loaded
independently and add its contributions to the main module (Commands or
Config).
After upgrading to Ruby 3.1, following error raises when running
redmine_git_hosting tests depending on loading a fixture based on a YAML
file:
Psych::DisallowedClass:
Tried to load unspecified class: Time
Redmine5 finalization: modules loading
This reverts commit 3db5af6. Because our unit tests assert that
No description provided.