Content length handling

[amouat]

Closes #514.

I moved all logic for setting Content-Length out of write_response and into dispatch.

I expanded the tests in lib/tests/head_handling.rs.

[amouat]

Test failure is due to out-of-sync libraries. The fix I used is to bump rusqlite to "0.13" in examples/raw_sqlite/Cargo.toml. I didn't include the fix in this PR as it's a different issue.

[SergioBenitez]

It's failing because of rust-lang/cargo#4902.

[SergioBenitez]

This changeset is much more complicated than it needs to be. All that needs to happen is that the body needs to be stripped using Response.strip_body() if the user's handler responded successfully to a HEAD request. The code already does the rest. This should be a line or two of changes.

[amouat]

I'm new to Rust and know very little of Rocket internals, so there may well be a better way of doing this. I'm afraid I can't see it though, and I did play with a few approaches.

I'll take you through my logic. In order to correctly meet the HTTP spec, we need to:

1. Set the content-length on HEAD requests to the same as the GET equivalent.
2. Set the content-length on empty body requests to 0 (I'm not sure if this is essential)
3. Do not set the content-length for streaming responses
4. Strip the body from HEAD requests

In the original code, the body is stripped before the content-length is set. To correctly set the length, we can either move the code that sets the content-length so it happens before the strip, or save the content-length somewhere so we know what to set it to. I felt it was simpler to move the content-length code to dispatch.

The final match branch can probably be simplified to _ at the loss of some explicitness. It could also be changed to an if/else, but I think the match is cleaner.

I wonder if it would be possible to automatically generate HEAD routes at start-up rather than handle them in dispatch if they aren't set. This would simplify dispatch and get rid of the need to cache the method type for this patch.

[SergioBenitez]

Unfortunately I don't have a lot of time at the moment, so I went ahead and took care of it, @amouat. Apologies for being a bit short on my previous response.

[amouat]

No worries, I had assumed strip_body set the size to 0, which was wrong.