Re: What's cooking in git.git (Jul 2019, #03; Fri, 12)

[Junio C Hamano <gitster@pobox.com>]

Matthew DeVore <matvore@comcast.net> writes:

> On Fri, Jul 12, 2019 at 02:02:52PM -0700, Junio C Hamano wrote:
>> * md/list-objects-filter-combo (2019-06-28) 10 commits
>>  - list-objects-filter-options: make parser void
>>  - list-objects-filter-options: clean up use of ALLOC_GROW
>>  - list-objects-filter-options: allow mult. --filter
>>  - strbuf: give URL-encoding API a char predicate fn
>>  - list-objects-filter-options: make filter_spec a string_list
>>  - list-objects-filter-options: move error check up
>>  - list-objects-filter: implement composite filters
>>  - list-objects-filter-options: always supply *errbuf
>>  - list-objects-filter: put omits set in filter struct
>>  - list-objects-filter: encapsulate filter components
>> 
>>  The list-objects-filter API (used to create a sparse/lazy clone)
>>  learned to take a combined filter specification.
>> 
>>  There is a bit of interaction with cc/multi-promisor topic, whose
>>  conflict resolution I have no confidence in X-<.  Extra sets of
>>  eyes are appreciated.
>> 
>
> Sorry for the delay. I was on vacation and then catching up for a week after I
> got back. I uploaded a merged commit here:
>
> https://github.com/matvore/git/tree/filts
>
> And the merged file itself (only this one had conflicts) is here:
>
> https://github.com/matvore/git/blob/filts/list-objects-filter.c

Thanks.  I fetched the 'filts' branch and found:

 (1) master..filts~1 matches the copy I have above exactly (modulo
     my sign-off and committer identity, of course);

 (2) if I merge cc/multi-promisor on top of filts~1 using the machinery
     I use to rebuild 'pu' every day, I get more-or-less the same result
     as your 'filt' branch (modulo formatting and minor comments).

So it does look like the conflict resolution I have been carrying is
something you would agree on, which is a good news ;-)  Thanks.

> I'll comment on the conflicts:
> ...
> md/list-objects-filter-combo changed the contract of this function such that an
> attempt to combine filter specs will terminate with BUG rather than return an
> error. All the callers already check filter_options.choice, so this is still
> fine (it particular, I double-checked partial_clone_get_default_filter_spec and
> its call site at builtin/fetch.c:1524)

OK, thanks for being careful.

>
>>   	/*
>>   	 * Record the initial filter-spec in the config as
>>   	 * the default for subsequent fetches from this remote.
>>   	 */
>> ++<<<<<<< md/list-objects-filter-combo
>>  +	core_partial_clone_filter_default =
>>  +		xstrdup(expand_list_objects_filter_spec(filter_options));
>>  +	git_config_set("core.partialclonefilter",
>>  +		       core_partial_clone_filter_default);
>> ++||||||| merged common ancestors
>> ++	core_partial_clone_filter_default =
>> ++		xstrdup(filter_options->filter_spec);
>> ++	git_config_set("core.partialclonefilter",
>> ++		       core_partial_clone_filter_default);
>> ++=======
>> + 	filter_name = xstrfmt("remote.%s.partialclonefilter", remote);
>> + 	git_config_set(filter_name, filter_options->filter_spec);
>> + 	free(filter_name);
>> + 
>> + 	/* Make sure the config info are reset */
>> + 	promisor_remote_reinit();
>> ++>>>>>>> cc/multi-promisor
>>   }
>>   
>>   void partial_clone_get_default_filter_spec(
>
> md/list-objects-filter-combo used the expand_list_objects_filter_spec function
> to expand the filter spec string rather than get it directly. So the merged
> result simply applies that alteration to cc/multi-promisor.
>
> I checked whether callers to this function (partial_clone_register) would ever
> give a null filter_options (or a non-null with a NULL filter_spec) and both
> calls are guarded by "if (filter_options.choice)" so filter_options.filter_spec
> should also be set.

Good.  This was the part I was most unsure about.

Thanks.