Age | Commit message (Collapse) | Author |
|
Currently marked as experimental because it can't be tweaked
(just enabled/disabled) and because it is not documented yet.
|
|
This would prevent the client from knowing whether the request
was received fully without the connection closing first.
|
|
The {data, IsFin, Data} uses IsFin to indicate whether this
is the last chunk, while chunked transfer-encoding uses the
length of Data, and ends when it is 0. We must therefore not
send chunks with empty data.
|
|
The stream handlers can be specified using the protocol
option 'stream_handlers'. It defaults to [cowboy_stream_h].
The cowboy_stream_h module currently does not forward the
calls to further stream handlers. It feels like an edge
case; usually we'd want to put our own handlers between
the protocol code and the request process. I am therefore
going to focus on other things for now.
The various types and specifications for stream handlers
have been updated and the cowboy_stream module can now
be safely used as a behavior. The interface might change
a little more, though.
This commit does not include tests or documentation.
They will follow separately.
|
|
|
|
|
|
This commit changes Cowboy to follow RFC6585.
|
|
|
|
|
|
|
|
Correct expected return type from `no_call` to `undefined` in
if_modified_since when last_modified callback is not defined. Add an
http_SUITE test to catch regressions.
|
|
Includes refactoring of the related code to avoid repetition.
|
|
|
|
|
|
After the switch to Websocket, we are no longer in a request/response
scenario, therefore a lot of the cowboy_req functions do not apply
anymore.
Any data required from the request will need to be taken from Req
in init/2 and saved in the handler's state.
|
|
|
|
|
|
The option for enabling Websocket compression has been
renamed. Previously it was shared with HTTP compression,
now it's specific to Websocket. The new option is named
'websocket_compress'.
|
|
Removes some duplicate code from cowboy_http.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Now named read_part/read_part_body, with a verb indicating action.
|
|
This is taken care of at the protocol level now.
|
|
|
|
This is a large commit. The cowboy_req interface has largely
changed, and will change a little more. It's possible that
some examples or tests have not been converted to the new
interface yet. The documentation has not yet been updated.
All of this will be fixed in smaller subsequent commits.
Gotta start somewhere...
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Now that we have a public map they are pretty much pointless.
|
|
Along with more cowboy_req tests.
This commit also removes cowboy_req:url/1 and cowboy_req:host_url/1
in favor of the much more powerful new set of functions.
|
|
This internal function is no longer necessary.
|
|
|
|
|
|
|
|
Previously only DATA was sent, with missing HEADERS.
|
|
Error reporting for connection processes has been improved,
using a similar proc_lib hack as was done for the stream
processes.
|
|
|
|
Note that this commit has currently only been tested on Linux.
It might be incomplete for other platforms.
|
|
|
|
The path segments . and .. are now removed according to the
rules found in RFC3986.
The path segments are now percent-decoded using the correct
algorithm (the one in RFC3986 and not the "query string" one).
|
|
The If-Modified-Since and If-Unmodified-Since headers are
only used when If-None-Match or If-Match were not found,
respectively. The latter are preferred by the standard
and the former is only there for compatibility with older
clients.
|
|
Was badly implemented previously.
|