Bump to a customised version of resemble for node 8+.

[alexjeffburke]

No description provided.

[alexjeffburke]

This PR restores the ability to install the plugin on (at least) node 8.

[coveralls]

Coverage increased (+0.3%) to 90.909% when pulling 96cb3f9 on alexjeffburke:master into 701e784 on unexpectedjs:master.

[coveralls]

Coverage increased (+0.3%) to 90.909% when pulling 96cb3f9 on alexjeffburke:master into 701e784 on unexpectedjs:master.

[coveralls]

Coverage increased (+0.3%) to 90.909% when pulling 96cb3f9 on alexjeffburke:master into 701e784 on unexpectedjs:master.

[papandreou]

Just as a sanity check I wanted to see the current version fail first, so I went rm -fr node_modules && npm install && npm test on the current master with node 8.9.3 -- and it worked:

> unexpected-resemble@3.1.0 test /home/andreas/work/unexpected-resemble
> mocha ./documentation/**/*.md ./test/**/*.js && npm run lint



  to-resemble
    ✓ example #1 (documentation/assertions/any/to-resemble.md:4:1) should fail with the correct error message (205ms)

  index
    ✓ example #1 (documentation/index.md:23:1) should fail with the correct error message (38ms)
    ✓ example #2 (documentation/index.md:45:1) should succeed (45ms)

  unexpected-resemble
    ✓ should succeed when the comparison is successful (58ms)
    ✓ should fail when the comparison is unsuccessful (137ms)
    ✓ should compare images provided as data: urls
    with mismatchPercentage in odd camel case
      ✓ should succeed when the comparison is successful (59ms)
      ✓ should fail when the comparison is unsuccessful (174ms)


  8 passing (746ms)

Could we take it from the top? What exactly did you see when it failed initially?

[alexjeffburke]

Hmm we’re you on linux by chance? I’m pretty sure the issues start when you’re compiling with clang... I could reliably get it to fail until canvas v1.6.3 which has a fix in it specifically for that situation on macOS. Trying to think what else might be relevant - I was using node 8.9.1 so I think it was a thing. I noticed the travis job forces the use of gcc and am wondering if that masked it in CI.

[papandreou]

Yeah, I'm on linux, I guess you're not?

[alexjeffburke]

That’ll be it, it’s the compiler - my understanding of the the as-is version of nan is it’s unhappy with clang.

[alexjeffburke]

And sorry, to make sure it’s captured - I was putting this together on macOS 10.12 :)

[papandreou]

Thanks, I've published 3.2.0 with this change. I bisected and found out that 7c54916 was what broke the browser build, so I also reverted that.

Tried getting rid of the resemble module and also tried upgrading resemblejs to the newest version. Both attempts caused the test suite to fail right and left, so...

[alexjeffburke]

Haha after my brief encounter I second the above :P Thanks for looking into it & getting the change out!