fast-import mirroring a foreign repository

[Ian Clatworthy]

Matt McClure wrote:
> What's the best way to mirror a foreign repository with `bzr
> fast-import`?  When I looked into it several months ago, it seemed
> that `bzr fast-import` required the entire fast-import stream,
> starting at the point of the original import, to make subsequent
> incremental imports work.  Is that still the case?
> 

Hi Matt,

Yes, that's still the case. I do have plans for adding an --append
option to the fast-import command that treats the data as an incremental
data set but that's not yet done. If you'd like to have a go at adding
that, it should be pretty simple for someone like yourself.

FWIW, I've spent 15-16 hours over the weekend fixing numerous fastimport
issues (most not yet merged). My current focus is getting full imports
working in all the various edge cases. Following the first law of data
migration, edge cases seem to be the norm in most large data sets. :-(

On the bright side of things, fastimport is gaining a lot more tests.
Speed is looking pretty good too in my latest branch, e.g. the first 35k
revisions of Python 3.0 import in 10-15 minutes before it falls over
with the same bug (I think) that Frits reported a few hours ago -
https://bugs.launchpad.net/bzr-fastimport/+bug/417238.

I'll post some more details as soon as things are working to my
satisfaction - hopefully a few hours from now.

Ian C.