Jeff King <peff@xxxxxxxx> writes:
> On Mon, Apr 17, 2017 at 07:27:16PM -0700, Junio C Hamano wrote:
>
>> > @@ -81,7 +82,7 @@ static struct option builtin_config_options[] = {
>> > OPT_GROUP(N_("Other")),
>> > OPT_BOOL('z', "null", &end_null, N_("terminate values with NUL byte")),
>> > OPT_BOOL(0, "name-only", &omit_values, N_("show variable names only")),
>> > - OPT_BOOL(0, "includes", &respect_includes, N_("respect include directives on lookup")),
>> > + OPT_BOOL(0, "includes", &respect_includes_opt, N_("respect include directives on lookup")),
>>
>> It would be more in line with what the log message advertised if you
>> did
>>
>> static struct config_options config_options = {
>> -1, /* .respect_includes: unspecified */
>> };
>>
>> OPT_BOOL(0, "includes", &config_options.respect_includes, N_("...")),
>>
>> no?
>
> I think I like the split between the option-value here and the "final"
> value that goes into config_options.respect_includes. Because we
> actually munge it later based on the given-config value anyway.
>
> So I agree this makes the diff larger than it might need to be, but I
> think the end result is a bit nicer.
Yeah, I didn't see the end result was a single bit (unsigned :1).
This separation is OK.