[aur-general] Cycled dependencies in AUR

[Tim Meusel]

Hi,
I'm the TU that recommended this.

On 24.08.21 17:12, Mikhail f. Shiryaev via aur-general wrote:
> Hello.
> 
> Today I tried to build ruby-rubocop [1] package and it's in a strange
> state. A lot of packages for the package are cross-dependent for the check,
> for example ruby-ruby-progressbar [2] and ruby-fuubar [3]. The proposal of
> how to solve it from Mario [4] is to skip the check() phase, see comment
> [5].
> 
> The same is valid for package ruby-rubocop-performance [6] that is
> dependency for rubocop check and requires it to be built
> 
> IMHO the state when the packages aren't able to be built by the simple AUR
> helper command like `yay -S ruby-rubocop` should be considered as room for
> improvement.
> 

I'm not sure about this. The AUR was always a tricky environment. There 
is a fine line between proper packages with documented dependencies and 
tests or an easy way to install packages. The "easy" way is always to 
just install the gem artifact from rubygems.org within a PKGBUILD and 
ignore all dependencies/tests. This will just break on every ruby 
upgrade and you need to ensure somehow you've all dependencies. 
Sometimes it's the best to build stuff locally and maintain the related 
packages in a private repo or to patch the PKGBUILDs and ignore the deps.

This was also done in a preparation to move the packages into community. 
We already have other strange depenency circles, like ruby-faraday-* has 
them.

> If I'm wrong, please say so. But I'd like to have AUR packages in an easily
> installable state and not passing additional arguments like `--nocheck` for
> makepackage commands.
> 
> [1] - https://aur.archlinux.org/packages/ruby-rubocop
> [2] - https://aur.archlinux.org/packages/ruby-ruby-progressbar/
> [3] - https://aur.archlinux.org/packages/ruby-fuubar/
> [4] - https://aur.archlinux.org/account/supermario
> [5] - https://aur.archlinux.org/packages/ruby-rubocop-rspec/#comment-823420
> [6] - https://aur.archlinux.org/packages/ruby-rubocop-performance/
> 
> Best regards,
> Mikhail f. Shiryaev
> 
> p.s. Mario, sorry for sending you two messages.
> 
>>