HTTP APPEND

[John Arbash Meinel]

Vincent LADEUIL wrote:

...

> So I guess the question is: do we trust the lock enough to
> prevent corruption ?

I think what would make sense would be to check the lock when switching
from coping file contents (text knits) to copying inventories.

In terms of code, that would be in bzrlib/fetch.py specifically around
line 136:
                pp.next_phase()
                self._fetch_weave_texts(revs)

-- Check here if the lock is still held
                pp.next_phase()
                self._fetch_inventory_weave(revs)
                pp.next_phase()
                self._fetch_revision_texts(revs)
                self.count_copied += len(revs)


That should limit possible corruption, and still give reasonable
performance, since it is only 1 round trip. And will keep us from
writing inventories that claim things we can't be sure of.

John
=:->

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 254 bytes
Desc: OpenPGP digital signature
Url : https://lists.ubuntu.com/archives/bazaar/attachments/20060830/a5f103e5/attachment.pgp