Pathname double backslashes on windows

[ricardocasares]

I'm having problems with a small app on express that makes requests to JIRA API using scoped. The problem is that, on windows, the app crashes trying to parse JSON from an empty response.

I don't know why the response comes empty, but I'm suspicious about the fact that, on windows, scoped builds the uri path with double backslashes as noted here:

Good URI on Linux: https://gist.github.com/ricardocasares/9486926#file-response-on-linux-L120

Bad URI on Windows: https://gist.github.com/ricardocasares/9486926#file-response-on-windows-L120

[ricardocasares]

I think I found the problem in the scope index.coffee, at line 105, when using node's path.join(). I think path.join() on windows uses backslashes.

I replaced path.join(p,suffix) for p + "/" + suffix and now it works fine on windows. But I don't think this is much of an elegant solution, or if it breaks something else.

[technoweenie]

Maybe url.resolve instead? Though it might mess people up if if the suffix starts with a /.

[ricardocasares]

I think url.resolve does the job. I don't know what do you mean by messing people if the suffix starts with a /? url.resolve seems to handle it.

[technoweenie]

url resolve will replace the path if the suffix has a leading slash.

Example:

$ node
> path.join("https://foo.com/bar", "/foo")
'https:/foo.com/bar/foo'
> url.resolve("https://foo.com/bar", "/foo")
'https://foo.com/foo'

Looks like the scoped client code checks for a leading slash already though. I think I'd prefer to use url.resolve since that's a builtin function for handling this though.

[ricardocasares]

Hmm, also url.resolve replaces the last segment of the url

> url.resolve('https://foo.com/bar','baz')
'https://foo.com/baz'
> url.resolve('https://foo.com/bar/qux','baz')
'https://foo.com/bar/baz'

[ricardocasares]

how should we handle this?

[ricardocasares]

If we append a last / to the end of @options.pathname, we get the addition to the path, plus the path rewrite when a leading / comes in:

> url.resolve('http://foo.com/bar/','baz')
'http://foo.com/bar/baz'
> url.resolve('http://foo.com/bar/','/baz')
'http://foo.com/baz'

[technoweenie]

We could add a trailing / to the prefix.