On 31 Jan 2022 14:58, Jon Turney wrote:
> On 28/01/2022 07:58, Mike Frysinger wrote:
> > --- /dev/null
> > +++ b/newlib/libm/Makefile.inc
> > @@ -0,0 +1,26 @@
> > +info_TEXINFOS += %D%/libm.texi
> > +%C%_libm_TEXINFOS = %D%/targetdep.tex $(LIBM_CHEWOUT_FILES)
> > +
> > +LIBM_CHEWOUT_FILES =
> > +
> > +LIBM_CHAPTERS =
> > +
> > +%D%/libm.dvi: %D%/targetdep.tex $(LIBM_CHEWOUT_FILES)
> > +
> > +%D%/targetdep.tex: $(LIBM_CHAPTERS)
> > +	$(AM_V_GEN)cat $^ > $@.tmp
> > +	$(AM_V_at)$(SHELL) $(newlib_basedir)/../move-if-change $@.tmp $@
> > +	$(AM_V_at)touch $@
> 
> We carefully only update $@ if it's changed, then touch it anyhow?
> 
> This seems like a transcription error, since originally we were touching 
> a timestamp file, or perhaps leftover from testing?

you're right of course.  i saw other places that weren't using the stamp
idiom, and then when rewriting these rules i remembered those places and
thought i wouldn't bother deploying the stamp pattern here since it was
already doing it wrong.  but i misremembered which ones were wrong.

> The same pattern occurs later, as well.

which one do you mean ?  the .xml ones ?  those currently aren't using the
stamp idiom.  or do you just mean the multiple places where targetdep.tex
is generated ?

> > --- a/newlib/libm/common/Makefile.am
> > +++ b/newlib/libm/common/Makefile.am
> >   
> > -CHAPTERS =
> > +LIBM_CHEWOUT_FILES =
> > +LIBM_CHAPTERS =
> > +include ./Makefile.inc
> 
> Is this boilerplate really needed before including the makefile 
> fragment, since '+=' should work even if the variable isn't already defined?

yes, automake enforces it

ideally we'll kill all libm/*/Makefile.am files at which point this boiler
plate isn't needed since there will just be the one set in the top-level
libm/Makefile.am.  i'm holding off on doing that clean up until i can finish
landing my current batch of changes.
-mike