Re: quoting $$ (or something completely different?)

[Philip Guenther]

On 4/1/06, Tom Bachmann <address@hidden> wrote:
> ther is a more serious problem I did not see, yet: while the example
> makefile is obviously stupid, a real makefile includes another file
> between the two assignments, so a more complete definition of foo would be
>
> define FOO
> CFLAGS := foo bar baz foo2
> include OTHERFILE
> $(eval $(call SET,CFLAGS))
> endef
>
> now it begins to make sense (to me), as OTHERFILE can make assignments
> like CFLAGS += -DFOO etc. It would be very annoying to always have to
> write $(eval ...).

No need.  Just put the 'include' in FOO inside an $(eval):
-----
define SET
$$(NAME)_$(1) := $$($(1))
endef

define FOO
$(eval CFLAGS := foo bar baz foo2
$(call SET,CFLAGS))
endef

$(call FOO,foo/bar)
----

(I realized after replying before that FOO only needs a single $(eval))


> What still confuses me is the following: I can write the initial
> sequence and everything works fine, so how can I just transform it into
> a "funtion"?

By putting the entire body of FOO inside an $(eval)


You later wrote:
>this seems to work and I guess I understand why it does, but I'm still
>wondering if the second parameter for set can be avoided.

You need $(NAME) to be expanded in a context where $1 is set to the
argument to FOO.  That can only be done inside FOO itself.  So it
either needs to appear textually in FOO (such as in a second argument
to SET) or be expanded with $(eval) inside FOO, as seen above.


Philip Guenther