The new text wasn't more accurate than the text it
replaced.  I'll add more words, but on the 3.5 branch
(where it belongs), and then merge that to the trunk.

tpc_abort():  Repair fubar'ed "repaired" logic.

tpc_abort():  Release the commit lock no matter what.

now is txn.savepoint(optimistic=True).

Port from 2.7 branch.

Collector 1900.

send_reply(), return_error():  Stop trying to catch an exception that doesn't
exist, when marshal.encode() raises an exception.  Jeremy simplified the
marshal.encode() half of this about 3 years ago, but apparently forgot to
change ZEO/zrpc/connection.py to match.

transactions for users defined in a non-root acl_users folder. Zope logs a
acl_users path together with a username (separated by a space) and this
previous fix failed to take this into account. Fixed by taking anything after
the last space in the compared path out of the equation.

now activates obj.

Re-add the zeoctl module to ZEO, and rewrite runzeo so
it can run as a main program (the scripts created by
mkzeoinstance.py expect to be able to do this).

http://mail.zope.org/pipermail/zope3-dev/2005-September/015605.html

argument from addBeforeCommitHook().

This adds .pop() methods to BTrees and Buckets.

Alas, zpkg changed since the branch was made, and it
wasn't (obviously) possible to build the branch anymore,
so much of this code didn't actually appear on the
branch.  The C code was changed to simplify and purge
non-ANSI gcc extensions, and the new test cases beefed
up a little.

in 3.8 (it didn't get released in ZODB 3.5; will be
released in ZODB 3.6).

happened in 3.6a1 (3.5a7 become 3.5 final, because 3.5a7 had
been released with Zope 3.2c1).

    
  http://permalink.gmane.org/gmane.comp.web.zope.zope3/13884

ZConfig-using zpkg; this will help keep the build-support stuff out of the
top-level directory of the checkout (such is the plan, at least)

This adds .setdefault() methods to BTrees and Buckets.

- note the zpkg command line that should be used
- remove Persistence from the resource map
- remove SETUP.cfg now that it is not needed

update_from_seq():  If PySequence_Check(seq) returns true,
go on to check for the existence of "iteritems" when deciding
whether the input is or is not "a sequence".  Alas,
PySequence_Check() does return true for PersistentMapping and
PersistentDict instances, so that they couldn't be used as
arguments to BTree/Bucket construction or update().

This is nasty type-sniffing, but it was before too.  There's
no way to make such stuff bulletproof, so I'm settling for
incremental improvement.