Fix deletion of duplicate rpaths

[carlocab]

Deferring field population to the end seems to break rpath deletion of
duplicated rpaths.

Closes #356.

---

This might not be the ideal solution, but I think duplicated rpaths happen rarely enough for the performance hit to not be that bad.

Also, getting this into Homebrew/brew would unblock Homebrew/homebrew-core#77263.

[carlocab]

I've added a test for the deletion of duplicate rpaths. To do so, I've checked in a binary libdupe.dylib into your test fixtures and amended the rpath deletion test to also delete the duplicated /usr/lib rpath:

❯ otool -l test/bin/x86_64/libdupe.dylib | rg -A2 LC_RPATH
          cmd LC_RPATH
      cmdsize 24
         path /usr/lib (offset 12)
--
          cmd LC_RPATH
      cmdsize 24
         path /usr/lib (offset 12)

I probably could've amended your Makefile to also generate duplicate rpaths in one of your existing test fixtures, but I was worried about breaking your existing tests.

[woodruffw]

Thanks for the fix! It's strange that deferring the population breaks deletion of duplicates, so that's something I should take a closer look at. This LGTM and I agree that the performance hit probably isn't too severe, so I'll merge now.

[woodruffw]

Cut with 2.5.1, thanks again!

[carlocab]

Thanks for the fix!

No problem! Thanks for your work on ruby-macho. I wish I had a better fix for you, but a lot of this stuff is way over my head at the moment.

It's strange that deferring the population breaks deletion of duplicates, so that's something I should take a closer look at.

Yup, I agree.

Cut with 2.5.1, thanks again!

Can we get that into Homebrew/brew?

[woodruffw]

Can we get that into Homebrew/brew?

Yeah; it should be as simple as brew vendor-gems; I think there's a vendor-gems workflow that can be used to run that update directly on the CI. I might not have permissions to do that anymore, though 🙂

[carlocab]

Oh, yes, I was vaguely aware of that. Just never used vendor-gems before. Ruby is still mostly voodoo to me 😄