On Wed, Jul 9, 2014 at 12:05 AM, Junio C Hamano [off-list ref] wrote:
Duy Nguyen [off-list ref] writes:
quoted
I wonder if we need to update_main_cache_tree() so many times in this
function. If I read the code correctly, all roads must lead to
update_main_cache_tree(0) in prepare_to_commit().
I think prepare-to-commit is too late; it does not want to know if
the index it was given to create a tree out of is the one that the
user will keep using after this invocation of "git commit" or just a
temporary one used for partial commit. The cache-tree rebuilt there
is what is recorded with commit_tree_extended() in cmd_commit(), but
if you are doing a partial commit, these generic code paths are
given a temporary index file on the filesystem to work on, and
cache-tree in that will not help user's later operation.
Right. Thanks for checking.
--
Duy