Currently, when the chunk size was an invalid hex number (i.e.
contained a non-HEX octet), hyper would halt its parsing at the
last valid hex digit and report this as the chunk size.
For example, the following would be the chunk sizes reported for
some strings:
"X" => 0
"1X" => 1
"aY" => 10
"a;a" => 10 (only this should be correct!)
This fix makes it so that any chunk size, which cannot be parsed
as a valid hex number, causes an IoError. Additionally, the case
where the chunk size is followed by a chunk extension is also
handled (since the ";" token is a valid delimiter between the
chunk size and, as such, along with a LWS octet, represents a
valid terminal character of the chunk size).
Regression tests for the `get_chunk_size` function and included.
- Some stray deriving -> derive changes
- use::{mod} -> use::{self}
- fmt.write -> fmt.write_str
This does not catch the last case of fmt.write_str in the
Show impl of a Header Item. This will need to be changed
separately.
- Includes ergonomic traits like IntoUrl and IntoBody, allowing easy
usage.
- Client can have a RedirectPolicy.
- Client can have a SslVerifier.
Updated benchmarks for client. (Disabled rust-http client bench since it
hangs.)
Instead, you can use an instance of a NetworkConnector with
`Request::with_connector`. This allows overloading of the NetworkStream
constructors, so that it is easy to modify how an `HttpStream` is
created, while still relying on the rest of the stream implementation.
BREAKING CHANGE
rfc2616 does not specify the maximum length of *Reason Phrase* so it's
a good practice to handle even unreasonably long input.
16 char. buffer is not enough to correctly handle even the common `304
Moved Permanently`. Increase buffer size to more realistic 32. Also, up
to 128 more characters will be read and ignored, providing even greater
versatility without increasing memory usage.
Issue #163
- Rename read_until_space() to read_token_until_space()
- Check if the byte is part of the token in read_token_until_space()
- Return HttpMethodError if the token is empty
- Remove the unnecessary is_valid_method() function
Also adds an EmptyWriter, used for GET and HEAD requests,
which will return an io::ShortWrite error if the user ever tries
to write to a GET or HEAD request.
Closes#77
The old parser used a manually unrolled state machine
and was broken due to upstream rust issues with match
statements.
The new parser is a read into a stack-allocated buffer
followed by a single match on the contents of that
buffer.
This significantly improves the benchmarks for method
reads by almost 30%, in addition to working around
the upstream rust issues with reordering match blocks.
