Create fake S3 library for tests.

[google]

Create fake S3 library for tests.

[lavelle]

For this, I'm thinking that a good approach would be to create a WebDAV to S3 client wrapper. We already have a local WebDAV server running for the unit tests there, so the fake S3 client could pull in data from that over WebDAV and then convert it into the same format as the real API. We'll need a way of loading filesystem data from a remote server and WebDAV is as good as any.

This means that both tests will be loading data from the same filesystem, so we can share tests across them. For example, "expect the first item in the directory to have name foo.txt" or "expect foo.txt to have type text/plain" are dependent on specific contents of the test filesystem.

I've already extracted the WebDAV test server out to its own directory on the WebDAV branch.

Thoughts @jmuk?

[jmuk]

As we briefly talked about this offline, using existing webdav local server sounds fine.
But keep in mind that S3 may have weird file names which can't appear in a normal file system (like a file name like with "///foo"). I assume that those exceptional ones can be addressed by other unittests of util.js or something like that without having s3 server.

[lavelle]

Yeah, good point. We can cover those exceptional cases with a few separate tests. The test server is now merged, so I'll proceed with this.