Windows Compat

[schneems]

• Path separators are platform dependent. We should use a cross platform API when building them.

[luislavena]

Here the version of Ruby is hardcoded, so will always use the same version of Ruby, even that you have different versions in the matrix defined beneath.

Take a look at my sample test package gem:

https://github.com/luislavena/test-ruby-c-extension/blob/master/appveyor.yml

Please note that make in this case is available thanks to rake-compiler and DevKit integration.

If you need just make, take a look to this thread in AppVeyor:

http://help.appveyor.com/discussions/problems/2504-what-is-the-right-way-to-include-the-ruby-devkit

Hope that helps.

[schneems]

Interesting, so you're saying that Ruby devkit comes with a GCC version of make. In this case the thing I'm having problems with is compiling libass and not the ruby bindings. The project doesn't use rake-compiler. I'm not sure if that's the best way to move forwards. It's in the ext sub dir here https://github.com/schneems/sassc-ruby/tree/master/ext.

The makefile they're using doesn't work for nmake https://github.com/sass/libsass/blob/3ae9a2066152f9438aebaaacd12f39deaceaebc2/Makefile. Which is where I got stuck. They are tested on windows https://github.com/sass/libsass/blob/3ae9a2066152f9438aebaaacd12f39deaceaebc2/appveyor.yml however that is greek to me, I have no idea what's going on other than they're not using nmake.

[xzyfer]

For reference this is where the LibSass Windows compilation happens

      if ($env:Compiler -eq "mingw") {
        mingw32-make -j4 sassc
      } else {
        msbuild /m:4 /p:Configuration=$env:Config sassc\win\sassc.sln
      }

[luislavena]

That is using a version of mingw just downloaded (in previous step).

I no longer have a Windows machine or involved with RubyInstaller but
DevKit bundled with AppVeyor should be able to compile libsass without
issues.

Sorry for top posting, sent from mobile.

On Jul 28, 2016 02:47, "Michael Mifsud" [email protected] wrote:

For reference this is where the LibSass Windows compilation happens

  if ($env:Compiler -eq "mingw") {
    mingw32-make -j4 sassc
  } else {
    msbuild /m:4 /p:Configuration=$env:Config sassc\win\sassc.sln
  }

—
You are receiving this because you commented.
Reply to this email directly, view it on GitHub
#1 (comment),
or mute the thread
https://github.com/notifications/unsubscribe-auth/AAAQVq8pCBAZ27jxIEFkTT3O9d2Pl7HXks5qZ_wGgaJpZM4JUuzO
.

[schneems]

Got it building and I believe FFI is finding and using the DLL correctly, however there's a segfault right now:

https://ci.appveyor.com/project/schneems/sassc-ruby/build/17/job/lsdrbpbpr4s4vf9k#L173

C:/projects/sassc-ruby/lib/sassc/engine.rb:39: [BUG] Segmentation fault
ruby 2.0.0p648 (2015-12-16) [i386-mingw32]
-- Control frame information -----------------------------------------------
c:0027 p:---- s:0126 e:000125 CFUNC  :compile_data_context
c:0026 p:0284 s:0122 e:000121 METHOD C:/projects/sassc-ruby/lib/sassc/engine.rb:39
c:0025 p:0034 s:0111 e:000108 METHOD C:/projects/sassc-ruby/test/functions_test.rb:174
c:0024 p:0011 s:0103 e:000102 METHOD C:/projects/sassc-ruby/test/functions_test.rb:92
c:0023 p:0028 s:0100 e:000099 BLOCK  C:/Ruby200/lib/ruby/gems/2.0.0/gems/minitest-5.5.1/lib/minitest/test.rb:108
c:0022 p:0008 s:0098 e:000097 METHOD C:/Ruby200/lib/ruby/gems/2.0.0/gems/minitest-5.5.1/lib/minitest/test.rb:206
c:0021 p:0007 s:0094 e:000093 BLOCK  C:/Ruby200/lib/ruby/gems/2.0.0/gems/minitest-5.5.1/lib/minitest/test.rb:105
c:0020 p:0019 s:0092 e:000091 METHOD C:/Ruby200/lib/ruby/gems/2.0.0/gems/minitest-5.5.1/lib/minitest/test.rb:258
c:0019 p:0007 s:0088 e:000087 BLOCK  C:/Ruby200/lib/ruby/gems/2.0.0/gems/minitest-5.5.1/lib/minitest/test.rb:104
c:0018 p:0034 s:0086 e:000085 METHOD C:/Ruby200/lib/ruby/gems/2.0.0/gems/minitest-5.5.1/lib/minitest.rb:327
c:0017 p:0035 s:0079 E:0013f8 METHOD C:/Ruby200/lib/ruby/gems/2.0.0/gems/minitest-5.5.1/lib/minitest/test.rb:278
c:0016 p:0007 s:0073 E:001d44 METHOD C:/Ruby200/lib/ruby/gems/2.0.0/gems/minitest-5.5.1/lib/minitest/test.rb:103
c:0015 p:0029 s:0070 e:000069 METHOD C:/Ruby200/lib/ruby/gems/2.0.0/gems/minitest-around-0.3.2/lib/minitest/around/unit.rb:11
c:0014 p:0012 s:0066 e:000065 METHOD C:/Ruby200/lib/ruby/gems/2.0.0/gems/minitest-5.5.1/lib/minitest.rb:774
c:0013 p:0019 s:0060 e:000058 METHOD C:/Ruby200/lib/ruby/gems/2.0.0/gems/minitest-5.5.1/lib/minitest.rb:301

i'm getting the same thing (in my VM) locally. This is my first time trying to fix a native extension (let alone one on windows). So i'm not sure the best way to debug the segfault. It could be a problem with the compilation, or a problem with how FFI is binding to the dll, or a problem with the ruby code, i'm not sure how to move forwards to narrow down possibilities. Any good resources on where to get started debugging extensions with FFI?

[saper]

As far as I understand Ruby for Windows DevKits are based on MinGW and Visual Studio should not be used due to various incompatibilities related to the runtime C++ support. Can you try switching to make and the Unix-style build (which is also supported by libsass) and see if that fixes the segmentation fault?

[schneems]

So it does look like I should switch away from msbuild. I looked around for other ruby gems that work on windows with a native extension and use FFI. I found sqlite3-ruby from @tenderlove which looks like a good blueprint. He is using the rake-compiler gem which seems like most people are leveraging that toolchain.

Development for me is insanely frustrating and slow on a Windows VM. I just finished the provisioning process through Heroku to get a IRL windows laptop. I should have physical hardware soon. I'm out on vacation next week. After that I can come back and I think i'm going to start with basics by making a "hello world" native extension that works on windows and mac/linux (using sqlite3-ruby as a blueprint) then bring that experience back to this project.

Thanks all for the help so far. It's been harder than expected, but i've learned more as well.