[3.2] Bubble media queries

[xzyfer]

This PR implements bubbling and merging of media queries.

Fixes #185. Specs added sass/sass-spec#215.

To get this over the line I enlisted the help of @hugogiraudel to buff up our @media bubbling spec coverage.

TODO

• hoisting
• media_level_4
• media_wrapper_selector
• mixin
• merge_no_repeat
• selector_wrapper_media
• maintain the correct indentation for nested output mode

As a side affect of these fixes the following tests an outstanding @extends spec now also passes.

Bonus

This the cssize visitor added will also for the basis of @at-root!!!

---

This PR is ready to go for 3.2!

[xzyfer]

@mgreter any ideas why the AUTOTOOLS=yes COVERAGE=no BUILD=shared would be failing?

I'm not familiar with autotools, and it onl fails on the shared build. This is the only thing holding this PR up.

https://travis-ci.org/sass/libsass/jobs/45832632

[QuLogic]

You have fallback_impl defined as inline in the .cpp file, but declared without inline in the .hpp file. The inline definition is used in the .cpp, but since the .hpp doesn't have the definition, it's undefined for anything that uses it, e.g, the fallback method.

In short, drop the inline and it works , though I'm uncertain why it doesn't affect eval.?pp... If you check the symbols, then you see that the one from Eval doesn't exist:

$ nm .libs/libsass.so | c++filt | grep 'fallback_impl'
000000000008af80 T Sass::Contextualize::fallback_impl(Sass::AST_Node*)
0000000000161be0 T Sass::To_C::fallback_impl(Sass::AST_Node*)
00000000000cd6a0 T Sass::Cssize::fallback_impl(Sass::AST_Node*)
00000000000b3ae0 W Sass::Expand::fallback_impl(Sass::AST_Node*)

so I think it's unused and then optimized out.

[xzyfer]

Success!! Thanks for the find @QuLogic!

I'll clean up this PR tomorrow.

[coveralls]

Coverage increased (+0.25%) when pulling c79ef9f on xzyfer:feat/bubble-media-queries-2 into d679128 on sass:master.

[sebastian-steinmann]

Any ETA on getting this into a release?

Stuck with slow ruby-sass because of this bug.

[xzyfer]

This will land in 3.2 likely some time in Feb.