I’m working in Rails again. I hate it. It kills me. It drains my soul.
But this time it’s an external dependency that’s got my panties in a wad.
This project uses travis for CI. Travis recently auto-updated its Ruby setup to use bundler 2. Bundler 2 doesn’t support older versions of Ruby, but travis still allows you to declare them as test targets.
Okay, fine. Fuck you, Ruby. But that’s all manageable without wasting too much time, right?
Except I do my testing in a docker container to avoid the horrific, painful dependency bullshit Ruby forces on my system. Yes, even with rvm and some of its better plugins. rvm adds its own pain points to any process. Plus, there are more dependencies for this project than just Ruby’s, so rvm only helps so much. When you have to have just the right version of Image Magick and Libre Office… well.
So there I am, with my well-crafted Dockerfile that uses the base Ruby image, and I can’t for the life of me get this fucking Bundler 2 shit to work. Why not? Because Ruby fucking sucks.
The base Ruby image uses bundler 1, and puts it into a location you can’t
really remove without basically manually deleting its files. As in, rm -rf
/usr/local/lib/ruby/gems/2.5.0/blah/etc
. And probably a dozen tertiary
locations I’m not about to try and figure out. And, yes, jackasses of the world,
I did try a proper gem uninstall, even using the -i
flag.
Even after installing a later version in the location defined by the image as
the “right” place, bundler 1 is still used. Even directly executing the
bundler 2 binary, which lives in /usr/local/bundler
, and double-checking that
I’m executing the correct one, it won’t work. Even though Bundler is supposed
to actually read the Gemfile.lock to see which version of itself to execute, it
fucking will not work. It just runs Bundler 1, period.
So I start digging into the binary’s code. Because this is Ruby – these aren’t
compiled standalone binaries. And what do you suppose I find? Some crazy-ass
bullshittery that is apparently the cool way to do gems. This binary is actually
asking the Ruby gem system to decide which version of itself to use. So no matter
which version’s binary you call, you’re getting whatever Gem.activate_bin_path
returns.
I couldn’t figure out a way to force the issue no matter what I tried. So I rebuilt my Dockerfile using debian as a base. I installed Ruby manually. I installed bundler manually. And you know what? It all works fine.
How the fuck did Ruby get so absurd? Why are there so many goddamn points of failure now? I have an easier time wrangling node dependencies. And I fucking hate node.
And yet somehow, somebody has convinced the library world (as in libraries that have books and put up exhibits online and stuff) that Ruby on Rails is a good idea. All I see is money being thrown away while we debug absurd issues that simply don’t happen in Python, Go, or even node.
At least it’s not PHP….
I also despise Ruby, and tend to think it was written by a bunch of drunk frat boys who have limited talent even when sober.
Better late than never, but I have to respond: why are you so mean to drunken frat boys? Surely they wouldn’t unleash something as monstrous as Ruby.
Skill issue, I’m afraid.
Sad, but true! Not sure how Ruby devs got so bad at programming in general, but it does seem like a major skill issue. Poor bastards don’t even realize how bad they are.