The HTTP class is a backwards compatibility layer for the Python 1.5
API. (The only remaining use in the std library is xmlrpclib.)
The current change makes urllib issue HTTP/1.0 requests with
HTTPConnection, because is accesses HTTPResponse.fp directly instead
of using the read() method. Using fp directly interacts poorly with
persistent connections. There are probably better solutions than the
current one, but this is a start.
Add code to test_roundtrip() that figures out the encoding
from the first two lines of the file.
(We need to refactor this again to make it available to
all places that need this, e.g. linecache.py.)
In particular, watch out for comparing b"" to "". They're not equal,
but you can start at the code asking whether buf == "" for a long time
before realizing that it will never be True.
I'm not sure why I thought it should originally, but it introduces an
__del__() method on the response which cause the close() to be called
too soon using the HTTP compat class.
Also, remove some stale comments. The HTTPResponse calls makefile()
immediately, so there is no risk of it closing the socket.
If a makefile()-generated object is open and its parent socket is
closed, the parent socket should remain open until the child is
closed, too. The code to support this is moderately complex and
requires one extra slots in the socket object.
This change fixes httplib so that several urllib2net test cases pass
again.
Add SocketCloser class to socket.py, which encapsulates the
refcounting logic for sockets after makefile() has been called.
Move SocketIO class from io module to socket module. It's only use is
to implement the raw I/O methods on top of a socket to support
makefile().
Add unittests to test_socket to cover various patterns of close and
makefile.
In particular, add trivial implementations of .real, .imag and .conjugate()
to both, and add .numerator and .denominator to long.
Also some small optimizations (e.g. remove long_pos in favor of long_long).
error callback changes the bytes object in the exception the decoder might
use memory that's no longer in use. Change unicode_decode_call_errorhandler()
so that it fetches the adresses of the bytes array (start and end) from the
exception object and passes them back to the caller.
Reported by Joe Smith. This makes the CYGWIN tests pass;
it's a miracle it didn't fail on other platforms.
Seems like it was accidentally dropped (maybe a merge artifact?).
Jeremy Hylton
Guido van Rossum
Collin Winter
Georg Brandl
Walter Dörwald
Brett Cannon
Martin v. Löwis