Re: sindex() and strstr() //Re: To Paul D. Smith: "make -j" as agrea ttool for SMP and cluster parallelization

[Markus Mauhart]

"Paul D. Smith" <address@hidden> wrote ...
>
> I'm pretty sure there _ARE_ places in make which invoke sindex() with
> length values which are less than EOS; in fact I think I even thought
> I'd provided an example but I can't find it at the moment.  I'd have to
> look at it again to be sure though.

My experience is restricted to using "find sindex/subst_expand in files"
in the current sources, and I may missinterpret the few results.
OTOH I have a theory explaining why your intuition and longterm experince
could fail ... that cases you have in mind dont exist anymore ... one
candidate for me is function variable_expand_string() which calls
sindex() indirectly through subst_expand() ... currently here the
pattern gets an artificial EOS, not necessary for later subst_expand()
->sindex(), but necessary for prior find_percent(pattern) ... IMHO it
could be that older make versions didnt use that EOS-forcing find_percent()
[maybe they used "find_char_unquote (pattern, '%', ':', 0);)" or such ? ]
and therefore could supply EOS-less patterns into subst_expand()/sindex().
Anyway, I'm looking forward to your final results.


Regards,
Markus.