Add AppVeyor for testing on Win

[YanLobat]

It seems there are some troubles with path resolving on Windows
import BPage from 'b:b-page';

It will eat Win slashes and path to this block won't be resolved.

[awinogradov]

You can try and write your results here. But we use nodejs api for path building and we believe it have to work right.

[YanLobat]

@awinogradov I wrote here, cause I've tried already :)
I am in proccess of understanding right now. If I find anything, I'll make PR.

[awinogradov]

Ok) we can wait)

[YanLobat]

В общем решил этот ишью для себя, захотел сделать PR и обнаружил, что у меня версия 0.0.2 :)
Код немного изменился, но проблема была примерно в этом месте.
Path нодовский, если ты на винде берет виндовый путь, потом в эти обратные слэши еще экранирование добавляется и в итоге в конечный require (который получается из подобного рода записей: import BPage from 'b:b-page';) доходит "обгрызанный" путь. А сам CommonJS как верно заметили юзает нормальные слэши(юниксовые), поэтому пришлось ввернуть следующую штуку перед возвращением значения .replace(/\\/g, "/");

Так что, наверное, можно закрывать, но на будущее иметь в виду :)

[Yeti-or]

Let's keep this issue, and add CI for windows

[Guria]

adding .replace(/\\/g, "/"); here works for me. Why it still not fixed?

Should I make a PR with this change?

[Guria]

Hmm. I think "required-path" package is more suitable place to fix it.
cc @Yeti-or

[Guria]

Ok. Investigated a bit more.
In fact both nodejs require and webpack require works well with windows separators in path.
Even with mixed ones like: require('./components\\App\\App.js');

The problems starts when we replacing node with requires using generators.
It returns wrong string, like require('./components\App\App.js')

I see now 3 potential places to put fix:

• make required-path return always nix separators (see Fix windows separators in pathStr Yeti-or/required-path#3)
• make replace on output of required-path call
• make replace in generators that outputs require statements.

I like first one. The only problem I see here, is that purpose of required-path package will be a bit spoiled, since it's output works well now for any OS.
On the other side it seems that it's a generators job to prepare a valid output here.

What do you think guys?

[Guria]

One more suggestion: Yeti-or/required-path#3 (comment)

[awinogradov]

Still waiting for PR ;)

[Guria]

still waiting an answer from @Yeti-or