update pointer to js to sockjs-0.3.4. Also fix doctype and input
Fix for R16B's removal of parameterized modules
start xmerl before cowboy
SockJS actually depends on xmerl application as well
Give proper attribution to Michael Gebetsroither and Ivan Blinkov
- change rebar to actually see real errors - update cowboy_echo example to cowboy 0.8 - cowboy_handler terminate now has 3 arguments instead of 2 from: terminate(_Req, _Service) -> to: terminate(_Reason, _Req, _Service) -> - use cowboy_req:peer for ip/port of client - update cowboy_test_server.erl to cowboy 0.8 - update README.md - update usage of cowboy_req:transport to cowboy_req:get for cowboy 0.8 - cowboy 0.8 has all cookies lower cased (fixes JsessionidCookie.test_xhr)
- WebsocketHixie76 support has been dropped (from Cowboy). - Fixes #49 Thanks to the following people, for the collective patches that made this possible: Michael Gebetsroither, Yurii Rashkovskii, Ivan Blinkov.
from: terminate(_Req, _Service) -> to: terminate(_Reason, _Req, _Service) ->
Copy OTP's paramaterized module parse-transform (https://github.com/erlang/pmod_transform) into the source tree; add compiler pragmas to files that require parameterized modules to use the parse-transform (sockjs_pmod_pt). The module implementing the parse-transform has been renamed, as per OTP's suggestion, to avoid name clashes.
…5 seconds Timeout mechanism is broken with regard to websockets. cowboy process might be overloaded by incoming messages (And session not being able to pick them up quickly enough). If that happens, it's quite likely that session process will receive a timeout after 5 seconds of not having the 'reply' call. This seems to be less of a big deal for non-ws transports as they don't neccesairly come via a signle bottleneck process - receiver is different than sender.
Should be atom
Cowboy now passes SockJS protocol tests, updated docs