llvm 10.0.0

[smorimoto]

• Have you followed the guidelines for contributing?
• Have you checked that there aren't other open pull requests for the same formula update/change?
• Have you built your formula locally with brew install --build-from-source <formula>, where <formula> is the name of the formula you're submitting?
• Is your test running fine brew test <formula>, where <formula> is the name of the formula you're submitting?
• Does your build pass brew audit --strict <formula> (after doing brew install <formula>)?

---

[smorimoto]

The test succeeds, but there is a problem. I don't know what approach to take in brew to avoid this problem, so I would like to hear the opinion.

λ brew audit --strict llvm
llvm:
  * python modules have explicit framework links
    These python extension modules were linked directly to a Python
    framework binary. They should be linked with -undefined dynamic_lookup
    instead of -lpython or -framework Python.
      /usr/local/opt/llvm/lib/python2.7/site-packages/lldb/_lldb.so

[BrewTestBot]

• New formulae in homebrew/core should not have a bottle do block

[SMillerDev]

Please leave LLVM@9 out for now. You could make a new pull request for it if needed.

[Bo98]

Can you reduce the diff on llvm? You moved blocks around for no reason and actually in ways that make the audit fail:

==> brew audit llvm --online
==> FAILED
Error: 3 problems in 1 formula detected
llvm:
  * C: 19: col 3: `stable` (line 19) should be put before `pour_bottle?` (line 6)
  * C: 69: col 3: `head` (line 69) should be put before `keg_only` (line 11)

Also oclgrind will need a revision bump.

[timmyjose]

Please expedite this! :-)

[SMillerDev]

Please consider hosting a copy in a personal tap if you want this done fast, which is very easy to do: https://docs.brew.sh/How-to-Create-and-Maintain-a-Tap

If you want it done well you'll have to wait for maintainers to have time to review the changes and for CI to test it.

[Bo98]

oclgrind might just need a revision bump rather than llvm@9, unless you tried it out locally and it didn't work.

[timmyjose]

@SMillerDev It was a friendly suggestion. I very well know that it takes time to review and test - no need to get snarky.

[Bo98]

• Remove Aliases/llvm@9. You don't need an alias to something of the same name.

Error: Formula duplicating alias: /usr/local/Homebrew/Library/Taps/homebrew/homebrew-core/Aliases/llvm@9

• It looks like oclgrind does in fact need llvm@9, but still keep the revision bump as well since it will be needed to ensure users have the right dependency.

• The test of llvm@9 fails. Here's what I think will fix it:

[smorimoto]

Thank you for the review. I applied that.

[bcardiff]

Thanks @imbsky for this PR!

The current Formula/crystal.rb will also need depends_on "llvm@9". In the next version it will have support for llvm 10. Should that be fixed in this pr as Formula/oclgrid.cr or in a separate one?

[Bo98]

Probably this one. Can the test be improved to pick up those cases better? It seems like the current one never fails if the LLVM version is incompatible. What action breaks under LLVM 10?

[bcardiff]

I will try to improve them for some smoke testing. Because right now the issue appears when building from source probably or running the compiler specs that are to heavy for the CI.

[Bo98]

If it only fails to build from source, I could potentially add llvm to the list of formulae which test building dependents from source, if that doesn't end up being too costly time-wise.

[iMichka]

llvm@9:
  * Versioned formulae in homebrew/core should use `keg_only :versioned_formula`
  * Formulae should not have a `HEAD` spec

@imbsky can you please run brew audit llvm@9 --online --new-formula llvm@9 and brew audit --online llvm before pushing any changes? Running this build blocks our CI for 5 to 6 hours each time, so it would be great if audit errors are fixed before pushing. Thanks.

[smorimoto]

I'm sorry, I made some unnecessary commits, so please cancel them manually if necessary. I will try to test it in advance from next time. The only remaining problem is the following. Does anyone have a solution?

λ brew audit --online llvm
llvm:
  * python modules have explicit framework links
    These python extension modules were linked directly to a Python
    framework binary. They should be linked with -undefined dynamic_lookup
    instead of -lpython or -framework Python.
      /usr/local/opt/llvm/lib/python2.7/site-packages/lldb/_lldb.so
Error: 1 problem in 1 formula detected

[smorimoto]

The last commit should pass the test.

[smorimoto]

It seems that the build was aborted for some reason.

[Bo98]

It's already requeued. There were some CI issues earlier.

[smorimoto]

I wasn't used to Jenkins so I didn't know how to check it. Thank you!

[smorimoto]

All tests have passed!

[smorimoto]

@SMillerDev @Bo98 @iMichka Could you check this?

[Bo98]

I think this is good. I checked the bottle earlier and nothing seems to be missing. I'm going to check crystal in a moment and then I'll merge this.

[smorimoto]

@Bo98 Thank you!

[smorimoto]

When I built the Crystal compiler locally with brew's LLVM package (10.0.0), there seemed to be no problems.

[bcardiff]

Lucky us! It is usually a problem. If something arises, I will fix it later, we should be near to release the next version with llvm 10 support though.

[Bo98]

Thanks!

[smorimoto]

@Bo98 Thank you, too!

[timmyjose]

Thank you to @imbsky and all the folks who worked on this - I had built llvm+lld+clang 10 on my own in the interim, but it took around 40GB (!) on my local machine, so this is a very welcome release. Cheers!