Bazaar focus for 2.1 and 2.2
Jelmer Vernooij
jelmer at canonical.com
Fri Dec 18 01:49:40 GMT 2009
On Thu, 2009-12-17 at 19:41 -0600, John Arbash Meinel wrote:
> >> This would make the memory leaks less of
> >> a problem, and it should make the scheduling of code imports a bit
> >> fairer, since large branches would not keep the system busy for a long
> >> time.
> >
> > It would also need some work so that we don't publish an import until
> > the import actually finishes, I think that would be quite confusing.
Is it really problematic if an import is behind on the foreign branch
while it is being filled in, rather than being empty as is the case now?
> >> The overhead of resuming an existing import should be relatively small.
> >
> > Yeah. Let's talk about this in Wellington if we don't get to it before
> > then :-)
> I should note that *imports* should be different, and checkpointing
> every so often is a good thing. (InterDifferingSerializer writes to disk
> every 100 revs, triggering an autopack every 1000 revs or so.)
Perhaps we could factor out some of that logic into a helper class that
is shared by the different fetcher implementations in foreign branch
plugins, fastexport/fastimport, etc. This sort of logic might be
specific to the target repository format, which may or may not need
regular repacking, flushing, etc.
Ian and I looked at bringing in one of the classes from fastimport that
handled batch imports into bzrlib, but we never got around to actually
writing the tests and proposing it for merging.
Cheers,
Jelmer
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 835 bytes
Desc: This is a digitally signed message part
Url : https://lists.ubuntu.com/archives/ubuntu-distributed-devel/attachments/20091218/ca09a3c0/attachment.pgp
More information about the ubuntu-distributed-devel
mailing list