[RFC] [PATCH] (was: speeding up commits)

[Martin Pool]

On  9 Dec 2005, duchier at ps.uni-sb.de wrote:
> In the interest of getting some early feedback, here is my current
> patch for the BzrCherrypickMetadata idea (on top of my prelude patch).
> This version implements preliminary support for "commit" and "add"; do
> not try any other inventory-modifying commands as they will certainly
> result in corrupted metadata.
> 
> I'd appreciate if those of you who (unlike me) actually know what you
> are doing could comment and/or advise.
> 
> Please, ignore the _xxx inventory caching hack: it is awaiting a
> better solution through improved transaction objects.

I haven't run it, just looked at the code.

This increases the number of ways in which the working inventory behaves
differently from a recorded inventory.  That's fine, but makes me wonder
if we shouldn't have two classes (or groups of classes), rather than one
which sometimes behaves one way, sometimes another.

I would like to have some new tests for this functionality.  For example
after a commit check that the working inventory shows the right
predecessors, and also after a merge.

This tends to speak of 'heads', which is the git term and not a bad word
for it, but a new use here.  It might fit better to say 'new parents' or
'pending parents'. 

-- 
Martin
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: Digital signature
Url : https://lists.ubuntu.com/archives/bazaar/attachments/20051213/dcabdb87/attachment.pgp