.. with regard to:
- not crashing with decode_compressed_response==True
- expected cassette content for body string
- expected response content, i.e. proper decompression
Format with line length 110 to match flake8
make black part of linting check
Update travis spec for updated black requirements
Add diff output for black on failure
update changelog
* fix typo in pytest.mark.xskip
Change xskip by skipif marker as xskip is an unknown pytest marker.
* fix FileModeWarning
This fix the following warning:
FileModeWarning: Requests has determined the content-length for this request using the binary size of the file: however, the file has been opened in text mode (i.e. without the 'b' flag in the mode). This may lead to an incorrect content-length. In Requests 3.0, support will be removed for files in text mode.
* fix waring "calling yaml.load() without Loader=... is deprecated"
This fix the following warning:
YAMLLoadWarning: calling yaml.load() without Loader=... is deprecated, as the default Loader is unsafe. Please read https://msg.pyyaml.org/load for full details.
* fix collections.abc deprecation warning in python 3.7.
* update Flask dependency in order to get rid of the Request.is_xhr warning
This fix the following warning:
DeprecationWarning: 'Request.is_xhr' is deprecated as of version 0.13 and will be removed in version 1.0. The 'X-Requested-With' header is not standard and is unreliable. You may be able to use 'accept_mimetypes' instead.
Matching on bodies uses urllib.parse.parse_qs(), which fails to handle
UTF-8+URLEncoded POST bodies when the input is `bytes` rather than `str`,
causing matching to fail..
Fixed this by always doing decode('ascii') on URLEncoded POST bodies first.
test_xmlrpclib was failing with "can't pickle thread.lock objects" on Windows.
Other small issues were related to backslashes in paths and different line endings.
I wasn't emulating the stateful file-object in my response stub,
so urllib3 wasn't decompressing gzipped bodies properly. This
should fix that problem.
Thanks @bryanhelmig for the motivation to dig into this.
This commit changes the whole core internal flow of requests.
Now, requests are actually physically made lazily when a response
is requested. This allows the entire request to be sent at once.
Otherwise, it would be impossible to compare whether requests have
already been recorded, since httplib.send() allows you to effectively
stream requests over HTTP.
