opcode.

(finally) been defined: the serial number of objects in the
non-version after an abortVersion should be the last serial number of
the object before it was modified in the version.

This actually breaks the symmetry between serial numbers and
transaction ids, so it must be handled specially.

Specific changes include:

_setupDBs(): Updated documentation.

_finish(): Recognize the 'a' opcode, which is subtly different than
the 'o' opcode.  When the 'a' opcode is seen, we'll store both the
"live revision id" and the current transaction id in the _serials
table.  The former id is the serial number of the last change to the
non-version data.  If we're seeing the 'o' or 'x' opcodes, we'll save
None as the serial number, which tells the loop at the end of the
function that the object serial number and the transaction id are the
same (overwhelmingly, the most common case).

When serial# <> tid, we'll concatenate the two values together into
serial+tid as the value of the _serials's oid key.  So the value can
now be an 8-byte string (the common case), or a 16-byte string when
serial# <> tid.

_getSerialAndTid(): A new function which returns a 2-tuple of the
oid's current serial number and the tid of the last transaction that
modified the object.  Usually these are the same, but in the face of
abortVersion, they're not.

_getSerialAndTidMissingOk(): Like _getSerialAndTid() but if the oid
isn't in the _serials table, return (None, None) instead of raising a
KeyError.  Sometimes we need different table access semantics.

abortVersion(), commitVersion(), modifiedInVersion(), load(),
getSerial(), transactionalUndo(), history(), _dopack(): Use the new
_getSerialAndTid() access method to get the serial# and tid for the
last object modification, instead of accessing the _serials table
directly.

store(), _rootreachable(): Use _getSerialAndTidMissingOk()

and it is subtly different than the `o' opcode.  The serial number of
objects in the non-version after an abortVersion should be the last
serial number of the object before it was modified in the version.

This actually breaks the symmetry between serial numbers and
transaction ids, so it must be handled specially.

element if there isn't a prefix in the attribute name.

Fixes http://mail.zope.org/pipermail/zpt/2002-August/003697.html

The picklelog branch attempts to improve performance and reduce the
possibility of lock file exhaustion by optimistically writing pickle
and metadata tables at the store() call instead of during the
_finish() call.  store() will be bounded in the number of objects it
touches, but _finish() is unbounded, so we try to reduce the number of
database pages the latter might actually touch.

Other performance improvements are implemented based on hotshot
profiling.  The use of an extension module for Python 2.2 also speeds
up a critical loop.

This also implements the periodic checkpointing to improve recovery
times when the database is not cleanly shutdown.

doing a critical operation in C.  When fiddling the reference counting
we often need to

    1. unpack a 64bit string into a long
    2. add a delta value to the long
    3. pack the long into a 64bit string

_helper.incr() does these three steps in C.

This extension only works for Python 2.2, but Full.py has a Python
replacement for use with Python 2.1.

write_object(): uses the `x' opcode, which is very similar to `o'
except that the metadata is written optimistically during store() so
it shouldn't be written during _finish().

- Added BerkeleyConfig class for easier configuration of checkpointing
  policy and lock sizing.

- Added _lockstats() and _docheckpoint() helpermethods.

Checking why the tests are failing, i discovered a typo on the test. But unfortunately the tests are still failing. :(

returnf() function will definitely get called.  Otherwise, if we get
an exception during the test, we'll deadlock because it's returnf()
that releases the various test locks.

loadinc(): Pass the file pointer in as well as the mbox so that the
status printing will actually work.  I'm not passing the file in
/instead/ of the mbox, but I wasn't completely sure of the original
intent of the code.

Disabled tests that's probably broken during investigation.

If self->meth is something other than a PyFunction,
for example an ExtensionClass Method object, the blind cast breaks.  The
new code uses the string "(?)" if self->meth is not a function.

byte ranges because Acrobat Reader for Windows chokes when overlapping
ranges have been merged.

bug.  Before the fix, ZopeSecurityPolicy saw the aq_base global and thought
the module object was wrapped, leading to unexpected behavior.

in the error value.  It may provide extra info.

are not normally attributes of any object but acquisition wrappers, except in
one important case: if container is a module and that module happens to
import aq_base or aq_acquire from Acquisition, ZopeSecurityPolicy.validate()
does unintended things.  This made ModuleSecurityInfo declarations fail when
using the Python policy.

Now we no longer look at aq_base attributes, but rather the acquisition API,
which is what cAccessControl does.

- Show the most recent log entry first.

and lack of coverage in the code.

- Documented FORCE_PRODUCT_LOAD

- Changed FORCE_PRODUCT_LOAD so that if it is set, it determines
  whether products are installed regardless of whether ZEO_CACHE is
  set. This means that you can disable product installation by setting
  FORCE_PRODUCT_LOAD to an empty string even if you are not using a
  ZEO persistent cache.

- Fixed the help system and Zope Tutorial product to honor
  FORCE_PRODUCT_LOAD.

testrunner output.