Hi,
Philip Martin wrote:
>>Jon Foster wrote:
>
>>>2) I'd like to reduce further the number of locks that are taken.
>>> This patch still errs on the side of locking too much, which gives
>>> correct output but is not as fast as possible. It locks much less
>>> than before, though. This looks like it will require careful
>>> thought and testing - which is one reason to write the test-cases
>>> first. (This is the FIXME in the patch).
>
>
> When anchor and target are different you may want to take a
> non-recursive lock on the anchor, and then take a target lock using
> anchor as the associated lock. The current patch can be committed
> without this.
>
Yes, that's the general idea. However, I think there are a few
special-cases to be aware of. I want to find all these special-cases
and write tests for them before I start changing the code too much.
I'm not going to be online this weekend, so I'm unlikely to even start
this until the weekend after. (Unless someone else does it first).
Kind regards,
Jon Foster
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Thu Mar 11 00:20:58 2004