Re: How could I help now?

[Ramón García]

> If you can provide a patch with that functionality I will integrate
> that, then we can move on to the next part.

Ok, I will work with that.

The current design does not depend on resolution of automatic
variables like context::<. However, it does depend on indirect
resolution. Such as

check-sum = ${shell md5sum address@hidden

referenced with

.OUT_OF_DATE = ${changed-value $@,$<.sum,${$<::check-sum}}

The first definition creates a variable check-sum which returns the
MD5 sum of the current file. The definition of .OUT_OF_DATE returns
true if the value of the check sum is different from the one when the
current target was last successfully compiled. This is nice because it
can be overriden for specific targets. For instance one can define
target variables:

%.variable: check-sum = ${$*}

so that the "check-sum" of CFLAGS.variable is just the value of CFLAGS.

And thus one can write
ai.o: ai.c CFLAGS.variable


If I understand well the issue is when an automatic variable is
invoked in := rule, independently of whether it is in atarget context
or not. If one writes

a.c: a.o
a.c: var := $<
a.c: defs.h

var would have exactly the same issue: its current value is not valid.