Max's series to fix some problems around filters consists of 42 patches.
I'm sure that we didn't find all bugs around filters, and that filters
would be a constant source of bugs in future, as during developing new
feature nobody will consider all possible cases of dealing with filters
(OK, somebody will, but it's hard).
So, I'm thinking about starting some deprecations which will help us to
simplify all the picture at least in not far future. So, I'd really want
to deprecate implicit filters, ->file child based filters (move all
filters to use backing child), filename based interfaces (use node-names).
Most simple thing is implicit filters, so let's start from them.
drive-mirror don't support filter-node-name, so I propose to deprecate
it at all, together with drive-backup, instead of adding support of
filter-node-name, what do you think?
Vladimir Sementsov-Ogievskiy (2):
qapi: deprecate drive-mirror and drive-backup
qapi: deprecate implicit filters
qemu-deprecated.texi | 11 +++++++++++
qapi/block-core.json | 10 ++++++++--
qapi/transaction.json | 2 +-
include/block/block_int.h | 10 +++++++++-
blockdev.c | 20 ++++++++++++++++++++
5 files changed, 49 insertions(+), 4 deletions(-)
libvir-list mailing list