(There's really not a good reason for asynchat to be in
Python's standard lib.)

CVS (1.9).

as I hate line-continuation characters.

Remove done item from TO DO list (CDATA content support).

to a directory other than the ZEO package directory.

Add general support for CDATA element content, enabled by default for
<script> and <style> elements since those are the ones I found in HTML 4.01.

Added a test case to make sure <script> and <style> elements are handled
correctly.

set or cleared by specifying any true or false value.

(XXX I wonder if this should be recognized at compile time rather than
at interpretation time; oh well, this works.)

    more strict about matchnig only what's legal.  The expression ends
    up being a bit more complex, and needs additional checks to be
    done on what follows.

HTMLParser.check_for_whole_start_tag():  Helper method that uses
    locatestarttagend, performs the required additional checks, and
    determines whether we've actually found the end of the start tag,
    are at a buffer boundary, or have encountered an syntactical
    error.


HTMLParser.parse_starttag():  Use check_for_whole_start_tag() to see
    if we really have the start tag.

HTMLParseError.__init__():  Simplify assertion.


This should close ZPT(18).

Revise a test that requires "<tag attr=>" to be accepted -- it is not!

Add a check to ensure it raises an HTMLParseError.

Revise a test that requires "<tag attr=>" to be accepted -- it is not!

exceptions.

New tests which poke at transactionalUndo()
    checkSimpleTransactionalUndo
    checkTwoObjectUndo
    checkTwoObjectUndoAgain
    checkNotUndoable
    checkUndoInVersion

On the Full storage only, of course.

i.e. it undoes by writing new records.  This means undone transactions
can be redone.  It also has lots of other benefits.

- Made ZPT invoke DTML the normal way.

Also fix the error for a duplicate fill-slot (which was erroneously
reported as a duplicate slot *definition*).

Question arises, should we also complain about duplicate define-slot,
or is it okay to use define-slot with the same name more than once?
This would be similar to "#define X(a) (a+a)" in C; it does point out
that define-slot isn't a great name...

care of installing them when necessary (not for Python 2.0)

Now import asyncore and asynchat as top-level modules, rather than
from the medusa package. This makes it easier to use our custum
version for Python 1.5 and to use the standard versions for Python
2.x.

Note that the Python 2.1b1 and 2.1b2 releases had a broken asynchat
that causes these changes to break Zope. The Python cvs and earler and
later versions are OK.

care of installing them when necessary (not for Python 2.0)

checkCreateNoFile(), checkCreateWithFilename(),
checkCreateWithFileobj(): Make sure these tests close the file object
before they attempt to unlink the underlying file.y

checkCloseNoUnlink(), checkDel(): New tests in the CreateCommitLogTest
class.

CloseCommitLogTest: removed this class; its tests are folded into
CreateCommitLogTest.

(there's no DB object since we're poking at things at a lower level).

tearDown(): Make sure any outstanding transactions are aborted so as
not to retain modified objects connected to closed databases.  Second,
call self._close() to free any resources.

_close(): Close the DB object, which in turn closes the underlying
storage.

_finish(): In delete-a-version clause, use next_dup() instead of
next() to iterate over the currentVersions table with cursor set to
the vid of the version we're deleting.

abortVersion(): When cruising over the currentVersions table (and the
key set to vid of the version we're aborting), use next_dup() instead
of next().  Also, when curvid <> vid, raise StorageSystemError instead
of VersionError.  Same later on when we check curvid <> zero.

commitVersion(): Add sanity checking based on model documentation: if
the source version is the empty string, or src == dest, raise a
VersionCommitError.  Make the same changes we made previously to
abortVersion() to keep track of the oids in a set/dictionary instead
of a list to avoid duplicates.  Other fixes include, writing revid
instead of tid==self._serial to write_moved_object() and adding a
write_discard_version(svid) once the source version's been committed.

commitVersion() now works!

load(): Add another check to avoid finding an object revision in a
committed or aborted version if we can guess that version name.  It
now raises a VersionError if vid==zero and version is true.

Also, although we still raise VersionError later on, elaborate the
actual exception better in the error message.

    checkAbortVersionErrors
    checkModifyAfterAbortVersion
    checkCommitToNonVersion
    checkCommitToOtherVersion
    checkAbortOneVersionCommitTheOther

All are inhibited for Minimal storage.

(a.k.a. pickle pointer) is usually tid but not always!  In fact, it's
specifically not tid when we've aborted a version -- it'll be the
pointer to the pickle of the last non-version revision.  We were
actually setting up lrevid correctly, but writing tid to the metadata
record.  This fix writes lrevid instead.

storage has a failing abortVersion(), so it's mostly a checkpoint.

Not yet tested: commitVersion(), transactionalUndo(), undoLog().

Not yet implemented: history(), pack()

Specific changes:

    - Table txnOids renamed to txnoids everywhere.

    - Define a module global `zero' constant and use (via global)
      where ever appropriate (should be named ZERO but that's ugly).

    - _finish(): Only boost the refcount of objects referenced by the
      pickle when we've actually got a new pickle.  Also, add a note
      about the currentVersions table, which will contain multiple
      copies of vid/oid records.  It's easier to weed them out later
      than to weed them out here.  One of the places you need to weed
      them all out is at the end of this method, in the
      `delete-a-version' branch.

    - abortVersion(): Record oids found by scanning the
      currentVersions table in a set, to account for the possible
      multiple vid/oid records.  Fix some typos.  When calling
      write_nonversion_object() on the CommitLog, the prevrevid ought
      to be the revid of the current object revision, for undo
      purposes.

    - load(): Where we previously had a comment that asked whether we
      should assert that the version is empty (after we've checked
      everything else), now we really check that and raise a
      VersionError if not.

    - loadSerial(): Fix a typo.

    - store(): Inside the old-revision-id-is-true test, if the old vid
      is not zero and it is equal to the new vid, then the non-version
      revision of the object is the old non-version revision (added
      the else: clause).

    - transactionalUndo(): Implemented but not yet tested.

    - undo(): Removed.

    - versionEmpty(): Semantics are to return true if an unknown
      version is given.  As an extra bonus, if version is empty
      (i.e. the non-version) check the serials table.

common operation. ;)  Also, _dostore() automatically pickles the data
argument, since you don't really want anything else.

Lots of new tests:
    checkNonVersionModifiedInVersion()
    checkLoadSerial()
    checkVersionedStoreAndLoad()
    checkVersionedLoadErrors()
    checkConflicts()
    checkVersionLock()
    checkVersionEmpty()
    checkVersions()
    checkAbortVersion()

Skip for the Minimal storage:
    checkVersionedLoadErrors()
    checkVersionLock()
    checkVersionEmpty()
    checkVersions()
    checkAbortVersion()

Override for the Minimal storage (basically check for Unsupported exceptions):
    checkLoadSerial()
    checkVersionedStoreAndLoad()