Modify options for darwin platforms

[qnighy]

It is known that we need a different set of options for the darwin platforms as seen in:

• https://github.com/protocolbuffers/protobuf/blob/v4.0.0-rc2/kokoro/release/ruby/macos/ruby/ruby_build_environment.sh#L27-L50
• https://github.com/sparklemotion/nokogiri/blob/v1.11.0.rc3/scripts/setup-osx-native-builders#L34-L39

This patch backports the modifications above, and also fixes #177.

[kou]

cross-ruby.rake is for building Windows binaries by Mingw-w64.
Why do you want to use it for macOS?

[qnighy]

cross-ruby.rake is for building Windows binaries by Mingw-w64.
Why do you want to use it for macOS?

In fact, I'm not sure about the exact reason why google-protobuf, grpc, or nokogiri uses rake-compiler for non-Windows platforms. I made this patch only because I wanted to easily reproduce build environments for google-protobuf.

Nevertheless, I thought it's a common usecase because rake-compiler-dock provides builds for Linux.

[kou]

I wanted to easily reproduce build environments for google-protobuf.

How about using https://github.com/protocolbuffers/protobuf/blob/v4.0.0-rc2/kokoro/release/ruby/macos/ruby/ruby_build_environment.sh#L27-L50 instead of using raw rake-compiler?

Nevertheless, I thought it's a common usecase because rake-compiler-dock provides builds for Linux.

I'm not sure about it's a common use case but cross-ruby.rake isn't for it.

[qnighy]

How about using https://github.com/protocolbuffers/protobuf/blob/v4.0.0-rc2/kokoro/release/ruby/macos/ruby/ruby_build_environment.sh#L27-L50 instead of using raw rake-compiler?

Yeah, that's what I'm doing now. And it took a while to figure out we need the following additional patch:

-    'LDFLAGS=-pipe -s',
+    'LDFLAGS=-pipe',

So I made this PR to share the knowledge with whoever needs it.

I'm not sure about it's a common use case but cross-ruby.rake isn't for it.

I thought rake-compiler-dock shares philosophy with rake-compiler (because they're in the same org), but is it not?

[kou]

Philosophy...?
I'm just the current maintainer not the original author but I don't have philosophy of this project... What do you want to say by it? "You should accept this patch"?
Anyway, we don't list the feature in our feature overview: https://github.com/rake-compiler/rake-compiler#feature-overview

If we add support for building different version Rubies for the current platform, we need to redesign and rename the current cross-ruby.task. It's not "cross-compiling".

We can't accept this patch because this approach is too ad-hoc. MINGW_TARGET =~ /darwin/ is strange. MinGW is "Minimalist GNU for Windows". It's not related to Darwin.

[qnighy]

What do you want to say by it? "You should accept this patch"?

Depends on the answer. It's fine if rake-compiler is totally for MinGW only and rake-compiler-dock's Linux builds are only working by chance. Or if Linux is supported but darwin is out of support, then that would be fine too, although unfortunate to me. I wanted to clarify which of them are supported.

If we add support for building different version Rubies for the current platform, we need to redesign and rename the current cross-ruby.task. It's not "cross-compiling".

I agree that the current use-case is not a cross-compilation. However, I think supporting darwin hosts or darwin targets does sometimes imply cross-compilation because darwin runs on several ISAs.

We can't accept this patch because this approach is too ad-hoc.

I'm confused again. Then is there a chance such a change is accepted if cross-ruby.rake is refactored in a non-ad-hoc way? Like renaming MINGW_* into some other names and supporting both .exe and non-.exe files...

[larskanis]

I'm currently adding support for MacOS to rake-compiler-dock. It mostly works already - I expect to have it ready when ruby-3.0.0 comes out. First I aim to support x86_64-apple-darwin, maybe later arm64-apple-darwin as cross targets.