… back. This already worked for most adapters, but had an issue for the HTTP client adapter. Note that I had to change the line breaks in the webmock server response to \r\n (which matches the HTTP spec, I believe) in order to get curb and patron to parse the headers correctly.
A few fixes that improve APIs used by VCR
…erministic in Ruby 1.8.7-p357
Added support for Excon
Allow WebMock.disable_net_connect! to accept a RegExp as :allow Parameter
em_http_request_1_x: use bytesize of message instead of its length for content-length header
Fixed a Rack::Lock bug around failing to close the response body
…d uri, or RequestStub
Weirdly, it appears that on 1.8.7, Array#delete will return the object you pass to it rather than the object in the array! O = Struct.new(:a, :b) i1, i2 = O.new(3, 5), O.new(3, 5) list = [i1] deleted = list.delete(i2) deleted.equal?(i1) # => true on 1.9.2, false on 1.8.7 deleted.equal?(i2) # => false on 1.9.2, true on 1.8.7
request. Previously, the httpclient and typhoeus adapters would use different request_signature instances for the global_stub hook and the after_requets hook. Using the same instance makes it more consistent and enables a new VCR feature.
Previously it just used the definition in Object#== which only returned true for the same object instance.
This is important for tools like VCR that use global stubs for more than just returning a stubbed response--for consistency, it needs to always have the stub block invoked so it can manage before/after request hooks.
Previously it didn't fully support this (but didn't prevent it either) and it had weird ambiguities. This cleans the code up a bit, too.
…r content-length header In rubies 1.8.x this will have no effect whatsoever In rubies 1.9.x this fixes problem of receiver's message getting truncated (since some utf-8 characters are more than 1 byte long)
…h Curb =< 0.7.15
net_http.rb: StubSocket needs read_timeout for aws-sdk