Re: [PATCH 1/3] Add read_cache_from() and discard_cache()

From: Junio C Hamano <junkio@cox.net>
Date: Jul 1, 2006, 18:51 UTC
Message-ID: <7v64ihdupr.fsf@assigned-by-dhcp.cox.net>
In-Reply-To: <Pine.LNX.4.63.0607011657460.29667@wbgn013.biozentrum.uni-wuerzburg.de>

Johannes Schindelin <Johannes.Schindelin@gmx.de> writes:

> True, I missed that one. But it is just a call to 
> cache_tree_free(active_cache_tree); in discard_cache(), right?

On the codepath to write out the new index file, calling cache_free_tree(&active_cache_tree) before write_cache() is all that should be needed. When "active_cache_tree == NULL", write_cache() would write out an index file without the cached tree information.

Currently not many things take advantage of cached tree information to optimize its operation. But I'd like to change that. For example, tree merges by read-tree should be able to take advantage of the fact that a cached tree read from the index and three trees being read all match for a subdirectory and do the merge of the directory without descending into it.

Show 11 quoted lines

>>  - index_timestamp is left as the old value in this patch when
>>    you switch cache using read_cache_from() directly.  I have a
>>    suspicion you may be bitten by "Racy Git" problem, especially
>>    because the operations are supposed to happen quickly thanks
>>    to the effort of you two ;-) increasing the risks that the
>>    file timestamp of the working tree file and the cached entry
>>    match.
>
> Yes. Again, just one line to discard_cache(), right?
>
> 	index_file_timestamp = 0;

This one I am not sure. Read the comment in ce_match_stat() and see the problematic sequence of events that this variable tries to help resolve applies to your use.