Barry Pederson <bp@barryp.org> [Thu, 10 Dec 2009 23:43:40 -0600] rev 173
The serialization layer was making hardcoded assumptions about how
short strings were encoded (utf-8). Add 'encoding' parameters
to AMQPWriter.write_shortstr() and AMQPReader.read_shortstr()
so the encoding can be controlled at a higher level. By default
though those two methods act as before. Thanks to 'angrybaldguy'
for bringing this up. (Copied from main branch)
barryp@host-145-114-107-208.midco.net [Mon, 05 Oct 2009 09:38:02 -0500] rev 172
Minor tweak to ensure the Channel class is included in pydocs, no real change
in the client's behavior.
barryp@host-145-114-107-208.midco.net [Fri, 28 Aug 2009 22:53:35 -0500] rev 171
Sync up CHANGES file with version 0.6.1
barryp@host-145-114-107-208.midco.net [Sun, 16 Aug 2009 22:17:21 -0500] rev 170
Close TCP sockets when there's an error connecting, to keep file descriptors
from piling up. Thanks todmitriy.samovskiy@... for pointing this out.
barryp@host-145-114-107-208.midco.net [Sun, 16 Aug 2009 10:23:28 -0500] rev 169
Fix minor typo in docs
barryp@host-145-114-107-208.midco.net [Sat, 15 Aug 2009 23:29:29 -0500] rev 168
denis.bilenko@... pointed out a couple try/except blocks that catch too much,
potentially preventing stop operations like KeyboardInterrupt from working.
barryp@host-145-114-107-208.midco.net [Sat, 15 Aug 2009 18:31:14 -0500] rev 167
Make it so that the upper levels see the exceptions raised by the lower levels,
instead of nonsense exceptions like:
TypeError: 'NoneType' object is not iterable
when something like a socket closure happens.
barryp@host-145-114-107-208.midco.net [Sat, 15 Aug 2009 13:37:12 -0500] rev 166
Get rid of the __del__ methods in Connection and Channel classes, which
allows them to be garbage collectible if references to them are lost, because
they form a cycle, and Python won't GC cyclical references that have __del__
methods.
Users really should use the .close() methods or 'with' statements
to make sure things are cleaned up properly. But failing that, this
change will at least ensure there's not a big leak including a TCP socket.
barryp@host-145-114-107-208.midco.net [Sat, 15 Aug 2009 13:33:56 -0500] rev 165
Add unittests provided by majek04@... checking that objects from this
module are garbage-collectable if references to them are dropped.
barryp@macbook.home [Mon, 23 Feb 2009 14:32:11 -0600] rev 164
Add some documentation about using 'with' statements.
barryp@macbook.home [Mon, 23 Feb 2009 14:13:09 -0600] rev 163
Add support for using Connection and Channel objects in Python 'with'
statements.
barryp@macbook.home [Mon, 23 Feb 2009 13:03:13 -0600] rev 162
Document the recent changes and change the version number so as not to be confused
with the original 0.6 version.
barryp@macbook.home [Mon, 23 Feb 2009 13:02:16 -0600] rev 161
Fix problem with dangling reference loops after Connections close. Thanks
to majek04@... for pointing this out.
barryp@macbook.home [Mon, 23 Feb 2009 12:27:12 -0600] rev 160
Make sure any while loops in the transport module raise IOError exceptions
if a socket can't be read from or written to. i Thanks majek04@... for
pointing this out.
barryp@macbook.home [Mon, 19 Jan 2009 11:03:04 -0600] rev 159
Restore the thread-related code that was temporarily removed for 0.6.
barryp@macbook.home [Mon, 19 Jan 2009 10:42:46 -0600] rev 158
Added tag 0.6 for changeset dbf98c6e962a
barryp@macbook.home [Mon, 19 Jan 2009 10:42:24 -0600] rev 157
Temporarily remove the thread-related development code, so the rest of
the improvements can be packaged as a new release.
barryp@macbook.home [Thu, 15 Jan 2009 20:14:02 -0600] rev 156
Polish up the docs a bit, including a warning to be sure to
close Connection or Channel objects otherwise you could
lose messages when the program ends.
barryp@macbook.home [Thu, 15 Jan 2009 19:41:53 -0600] rev 155
Have the Connection initializer actually behave the way the docs
have implied for a long time - supplying default values for host
(localhost) userid and password (guest, guest).
barryp@macbook.home [Thu, 15 Jan 2009 19:07:44 -0600] rev 154
Combine the two different SSL transports, turns out they're not that different.