[PATCH] move lock_read() into "bzr log"

[John Arbash Meinel]

Robert Widhopf-Fenk wrote:
> Also with the earlier changes in the revspec date: the
> current code still reads the knit (index) twice.
> 
> The attached patch removes the two lock_read() calls and adds
> a single lock_read() earlier in the call stack.
> 
> 
> 

Locks are re-entrant, so I would leave the lower level locks, in case 
some other code is calling them directly.

We probably should factor out all of the 'if revision is None' stuff 
into a helper function, since I think it gets done repeatedly. And now 
all the other functions need to do the same thing (any time we parse a 
date stamp).

Maybe something like:

def convert_revision_spec_to_revnos(branch, revision_spec):

And then we would also want

def convert_revision_spec_to_revision_ids(branch, revision_spec):


At the very least -1 on removing the existing locks. It is fine to add 
one earlier, though.

John
=:->