Should this makefile really hang make?

[Bryan Ischo]

I'm trying to use .SECONDEXPANSION to create pattern rules for which the 
prerequisites list is a function of the target.  But things don't work 
exactly as I expect.  Here is an example of a makefile that causes make 
to apparently enter an infinite loop:

.SECONDEXPANSION:
all: foo.a
%.a: $$(@:%.a=%.o)

I can't understand why make would go into an infinite loop with the 
above.  Shouldn't it be using the pattern rule to generate this rule:

foo.a: $(@:%.a=%.o)

and then doing a second expansion of $(@:%.a=%.o) with $@ equal to foo.a?

And if so, wouldn't that result in a prerequisite of foo.o?

Here is another example:

.SUFFIXES:
.PRECIOUS: %.c %.o

all: foo.a

%.c:
    echo > $@

%.o: %.c
    gcc -o $@ -c $<

.SECONDEXPANSION:
%.a: $$(patsubst %.a,%.o,$$@)
    ar cr $@ $^

This makefile produces this output:

echo > %.c
gcc -o %.o -c %.c
ar cr foo.a %.o

So make is using the pattern rule for %.a when figuring out how to build 
foo.a, but in the substitution that it does to determine the 
prerequisites it is using % instead of foo.  I don't get it.

Thanks,
Bryan