Re: [RFC 00/14] convert dir.c to take an index parameter

[Jeff Hostetler <git@jeffhostetler.com>]

On 5/8/2017 1:12 PM, Brandon Williams wrote:
> On 05/06, Junio C Hamano wrote:
>> Brandon Williams <bmwill@google.com> writes:
>>
>>> One of the things brought up on the list in the past few days has been
>>> migrating away from using the index compatibility macros.  One of the issues
>>> brought up in that thread was how simply doing that conversion doesn't
>>> eliminate the reliance on global state (specifically the_index).  If one day we
>>> want to have a 'repository object' passed around then we first need to convert
>>> different subsystems to be prepared to handle that.  This series provides a
>>> first step, converting the code in dir.c to take a 'struct index_state' and
>>> using that instead of implicitly using 'the_index'.
>>
>> Very nicely done (I only skimmed "dir.c" in the end result and didn't
>> go through the changes with fine toothed comb, though).
>>
>> I would have done this without the first step and then instead had a
>> final patch that only inserts a single
>>
>>     #define NO_THE_INDEX_COMPATIBILITY_MACROS
>>
>> at the beginning of dir.c once everybody in dir.c loses the
>> reference to all "cache" macros at the end, if I were doing this
>> series, but it is a personal taste.
>>
>> The resulting dir.c does not even refer to the_index, which is very
>> nice.
>
> Thanks! I'm glad there's a few people who see this as a positive change.

Agreed.  This looks like a nice start.
Jeff