> The new test,
> 
>         oodate = (gn->cmtime == 0 || Arch_LibOODate (gn) ||
>                  (gn->cmtime == 0 && (gn->type & OP_DOUBLEDEP)));

The original test:

	oodate = (Arch_LibOODate (gn) || 
		  (gn->cmtime == 0 && (gn->type & OP_DOUBLEDEP)));

worked fine for a.out, but Arch_LibOODate() did not DTRT on ELF if the
lib did not exists.  That's now fixed, but I was asked to short
circuit the call in Make_OODate() anyway.  The first cut of that was
1.26 was wrong.

> looks like a complicated way of writing
> 
>         oodate = (gn->cmtime == 0 || Arch_LibOODate (gn));

Yes I think you are right.

> There's a bug hiding in there somewhere... :-)

Maybe not a bug, but a waste of the compiler's time.

--sjg