John Wheeler <wheelerw...@runbox.com> writes:
> On 5/9/22 16:10, Jean Abou Samra wrote:
>>
>> Random idea: how about taking a difference approach and doing this
>> in LY_DEFINE instead? You can do something along the lines
>> of
>>
>> scm_set_object_property_x (the_resulting_scm_procedure,
>> ly_symbol2scm ("c++-location"),
>> scm_cons (ly_string2scm (__FILE__),
>> to_scm (__LINE__)));
>>
>> and then implement 'make tags' via 'out/bin/lilypond -ddump-tags',
>> where -ddump-tags is an option that iterates over the lily module
>> and outputs the tags file by reading the object property (somewhat
>> similar to the autogeneration of the Internals in
>> scm/document-functions.scm). That means this doesn't need to maintain
>> a duplicated version of the name mangling logic. Is the etags
>> format simple enough to write?
>
> The TAGS file structure is simple enough, and I agree having
> only one place to maintain name mangling logic is good.
>
> But, I am not following you on the reference to
> 'out/bin/lilypond -ddump-tags'. Is -ddump-tags a command line option
> to the lilypond executable?
Not yet. That's what "and then implement ..." is about, even if it's
worded a bit unfortunately in a manner suggesting that the LilyPond part
may already be there.
--
David Kastrup
