[test] Enhance testcases to be run on windows

[ChrisAlderson]

I had a few issues when running the test cases on Windows.

• [test] Windows systems do not have gid or uid
• [test] Adjust assumption of filesize for Windows eol
• [test-refactor] Use rimraf to remove test fixtures on all systems

The rest of the commits were for adding AppVeyor CI. If we don't want AppVeyor to test against Windows I can create a new PR.

[DABH]

Awesome! I gave this a shot on Windows and am seeing one of the tests failing

  1) Logger, ExceptionHandler Custom exitOnError function does not exit:
     Uncaught Unknown assertation failure occured, assumed `[]` to deeply equal `[ 'Ignore this error' ]`
  C:\winston\test\log-exception.test.js
                                    v
      66.       assume(child.killed).false();
      67.       assume(stdout).deep.equals(['Ignore this error']);
      68.       child.kill();

but this seems likely unrelated to your change. Are you also seeing this test fail?

[ChrisAlderson]

Sometimes I do, I'm unsure where this error comes from because when I run the tests a second time the error is gone. This is also why I added the appveyor.yml configuration.

[indexzero]

iirc npx is cross-platform. Either approach is 🆗with me.

[ChrisAlderson]

Could be, but npx was introduced with [email protected] so this won't work with node@6.

[indexzero]

Ah yes the "fun" of test matricies ^_^ ... technically we could do npm i -g [email protected], but it's a lot of time to add to each test run for little-to-no gain. Carry on 👍 🙏

[indexzero]

Can we get the value based on the platform then assert it? If we get 4096 on Windows it's a bug and if we get 4100 on *nix it's a bug. e.g.:

// Either 4096 on Unix or 4100 on Windows because of the eol
const expected = process.platform === 'win32' ? 4100 : 4096;
assume(stats.size).equals(expected);

[indexzero]

Would like to see the stat change suggested implemented before merge, but once that's done you have my 👍

... thanks for taking the time to add this @ChrisAlderson!

[ChrisAlderson]

Added the change you suggested @indexzero, also merged with master to fix the testcases in #1318.

[ChrisAlderson]

@DABH Think the error from log-exception was introduced by me in #1286, because of the handleExceptions and unhandleExceptions methods not being refactored correctly.

I'll open another PR to solve this issue.