Arvados

21700-remove-rvm @ 24e3397d88f64e7f71b4378ea93cb3e60a1b7f5c

developer-run-tests: #4181 (this ran on the previous commit but the latest commit is in the "obvious trivial fix" category so I hope this is good enough)

build-packages-multijob: #4174

The actual removing RVM part of this was relatively simple. What got complicated was coaxing Bundler to work with a wider variety of Ruby versions and setups (e.g., different sets of gems already installed across different distros). On the whole I think these were good bugs to tackle: I think the arvados-api-server package is more likely to work on more systems now. And now I think our build+deploy process is better established to write into a Docker image later. But admittedly I didn't plan to have to tackle so much when I started work on the ticket.

• All agreed upon points are implemented / addressed.
  • Yes
• Anything not implemented (discovered or discussed during work) has a follow-up story.
  • N/A
• Code is tested and passing, both automated and manual, what manual testing was done is described
  • See above
• Documentation has been updated.
  • All references to RVM have been removed. Not sure if this warrants an upgrade note? If it does, it would basically say: "If you previously installed with RVM, you'll need to make sure the Ruby you want is the first ruby on the system $PATH. Our package won't find it automatically anymore." I am skeptical anybody actually installed that way but this would cover our bases I guess.
• Behaves appropriately at the intended scale (describe intended scale).
  • No change in scale
• Considered backwards and forwards compatibility issues between client and server.
  • No change
• Follows our coding standards and GUI style guidelines.
  • Mostly N/A since we don't have a shell style guide, but yes to the extent it makes sense

Now that we don't use $bundler anywhere, I suppose this gem🙄 can also be deleted from run-tests.sh:

        if test -d /var/lib/arvados-arvbox/ ; then
            # Inside arvbox, use bundler-installed binstubs.  The                                                                                                         
            # system bundler and rail's own bin/bundle refuse to work.                                                                                                    
            # I don't know why.                                                                                                                                           
            bundle=binstubs/bundle
        fi

The bit where we use awk to implement "bundle package --no-install" ourselves is so awful that I tried to research what we might be doing wrong, and spent some time trying to trick bundler into doing the thing the docs say it does, and -- success! -- now I hate bundler more than I hate the awk script.

So, LGTM

Tom Clegg wrote in #note-3:

The bit where we use awk to implement "bundle package --no-install" ourselves is so awful that I tried to research what we might be doing wrong, and spent some time trying to trick bundler into doing the thing the docs say it does, and -- success! -- now I hate bundler more than I hate the awk script.

Yup, I also went through these same seven stages of grief. I didn't want to write it. I tried everything you did. I eventually decided that I hate the awk script, and it's the least worst option.