Redmine

I ran some tests with pipeline enabled.

• you have to manage manifests and indexes explicitly to tell sprockets which ones it should handle together, I don't fint it very "magic" (like in "Rails magic"), I whish there would have been a smarter way to handle dependencies using both manifests and usage of (javascript|stylesheets|images|etc.)_path in code. Moreover, if you fail to declare these dependencies nicely, you may not see missing precompiled assets before runtime at production stage
• depending on the activated settings, the precompilation (combination and compression) feature before deployment at production stage could mean that the Redmine project team will then also have maintain code, tests and documentation specific to production deployment
• precompilation rake task seens a bit unstable; I had some recurring 'stack level too deep' error on Windows

please, don't start a discussion on whether the asset-pipeline is a good idea or not. We had that since the day it was introduced. Everybody was skeptical first and everybody who has been using it has been convinced after using it for two or three days. Give it a spin, get convinced and then let's add it to the list of things that need to be done.

Stefan Frank wrote:

please, don't start a discussion on whether the asset-pipeline is a good idea or not.

Why not?

From a development pov, I'm already convinced by the language usage and the performance improvement possibilities.

From a maintainance/deployment pov, I feel like this might add some overkill.

I guess we eventually get it commmitted on trunk, the question is to know when; as I said, I had this precompilation issue on Windows and I'm not sure that the pipeline code is mature enough in RoR.

Please, what are the arguments that push you to switch to its use?

Why not?

basically, because there already has been a lengthy discussion about the asset-pipeline on the rails-list: if it is a good idea, if it belongs into core or should be optional etc.- the result is, it is there, it is now an integral part of rails, it is inside the core and it does not look like it's going to go away anytime soon. I've also been reluctant to use it, but it has grown on me and now I don't wanna miss it.

I agree, that it does add some testing overhead: In the end my guess would be, that it will also make lives easier for plugin-developers (e.g a custom task for copying assets like rake redmine:plugins:assets is no longer needed) - that is if a plugin is already a mountable Engine(is it?!): I just started out with plugin-development and I'm still struggling with the concepts here - if a plugin would "just" be an engine, I'd feel much more at home. There would not be a need a for a separate ruby script/rails generate redmine_plugin <plugin_name> - instead it would be possible to fall back on the standard rails-mechanisms for creating mountable engines. I really do not oversee all the implications, but my guess the asset-pipeline would only be a smaller (but vital) part of that bigger project...