GitLab

also, commented out now-meaningless tests rather than tweaking them
to let them pass in a superficial manner (we'll discuss this).

against the master address in the config. error if they don't match, start transfer if they do

and reversed its interpretation
also removed the additional None check (it was checked twice)

this would have revealed the DataSourceClient_getUpdater bug.

it even though it's unusual.  (From code inspection, BIND 9 also seems to
accept it).  So I changed the behavior to accepting it, and updating the
test accordingly.

as expected.

corresponding documentation update.

wherever possible.  in particular, the XfrinConnection class is now
"db_file"-free.

removed now-unused code path, and adjusted existing test cases accordingly.

the AXFR changes until XfrinAXFREnd.finish_message().

via an exception.   added a test for that.

for non incremental update

to do_xfrin().

object is cleared as soon as possible after the xfrin session.

client to see it really works for an operational environment (and indeed
I found a bug in the Diff class).  To make it possible, also updated
how to set the data source client for XfrinConnection.

handle exceptions more correctly.  in particular it should have caught
various other exceptions from other modules (e.g. one from pydnspp/isc.dns)
without incorrectly assuming it uses Boost.Python (we don't use it any more).
This change fixes another problem in an existing test, so I also fixed
that test case, too.

supported multiple-message IXFR session.

the code so that the Diff obj is created in the RR processing on demand.

made minimal adjustment to support the IXFR state transitions.  Refactored
and updated tests for a simple IXFR test and sharing some of the existing
frameworks.

to _create_query().  we don't have to examine the resulting wire-format
binary at this level of tests.

XfrinConnection class and process_xfrin(), and only convert it to a string
when necessary.

clearly.  no behavior change in effect at this moment.

exception) classes.