General question regarding interrupting bzr operations

[Nicholas Allen]

> We certainly *could*. And I know of another system where "update" was,
> "revert and save as a changeset, apply changes, apply changeset". It had
> the problem that it was *really* slow at doing all of that.
>
> Now, we are quite a bit faster at a lot of those steps. But there is the
> simple fact that you need to stat the entire tree to find changes, and
> then do a diff on any changed file to generate the patch, then update,
> then throw away the patch.
>
> It probably isn't a huge amount of overhead. But there is *some*
> overhead involved.
>
> Modulo the Windows locked files bug, it really is a short window of time
> where anything might be inconsistent. I certainly would rather spend the
> effort fixing the bug first, before we decide that saving a patch is needed.
Yes it would be better to fix the bug. Perhaps it could be nice to have 
this as an option for the really paranoid who are willing to accept a 
degrading of performance for some added security...

Cheers,

Nick