Re: master 9e94509 3/3: Generate info/dir directly from any org sources

["Basil L. Contovounesios" <contovob@tcd.ie>]

Eli Zaretskii <eliz@gnu.org> writes:

>> From: "Basil L. Contovounesios" <contovob@tcd.ie>
>> Cc: Mattias Engdegård <mattiase@acm.org>,  rgm@gnu.org,
>>   emacs-devel@gnu.org
>> Date: Sat, 06 Mar 2021 15:31:58 +0000
>> 
>>  texi_misc = $(shell ${MAKE} --no-print-directory -s -C doc/misc echo-sources)
>>  
>>  srcdir_doc_info_dir_inputs = \
>>    ${srcdir}/doc/emacs/emacs.texi \
>>    ${srcdir}/doc/lispintro/emacs-lisp-intro.texi \
>>    ${srcdir}/doc/lispref/elisp.texi \
>> -  $(addprefix ${srcdir}/doc/misc/,${texi_misc})
>> +  $(addprefix ${srcdir}/doc/misc/,$(filter %.texi %.org,${texi_misc}))
>
> This only works if the other output is interspersed "at word level",
> i.e. you don't have characters from one Make mixed up with those of
> another inside a single word.  Right?

Right.  (Though in practice it WFM™.)

> Why do we need to use $(shell) at all?

Ideally we shouldn't.  Even capturing the output of a sub-make recipe
e.g. with $(eval ...) would be better, since that would play better with
the jobserver.

> Can't we use $(wildcard) to
> the same effect without starting another process?  If the problem is
> that some *.texi files are generated, and we don't want them in the
> list, how about generating them in a subdirectory of doc/misc/?

Either sounds fine to me, but I don't understand well enough Glenn's
changes or how these recipes should work to be of much help with that.

Thanks,

-- 
Basil