Re: secondary expansion changes

[Paul Smith]

On Thu, 2009-09-24 at 16:47 +0200, Boris Kolpackov wrote:
> It should print an empty line and currently prints 'foo.o'.

Gah!  Order-only prereqs were basically not implemented at all for
secondary expansion in implicit rules.  Somehow during my changes that
entire functionality was just dropped on the floor.

Should be fixed now; thanks for the test case.  Let me know if you still
see the other issue.


I also found a bad memory leak (that I think was there with 3.81 as
well) when dealing with chains of implicit rules.  If you have a lot of
implicit rule chaining this COULD have been a problem; in one of the
regression tests I saw 6K leaked (and this are small, targeted tests).


There are still some memory leaks (no errors) that appear to be struct
dep's not getting freed; I'm trying to find those.  It sure would be
nice if valgrind had an option to jump into the debugger when it finds
leaks, and not just errors... make is so recursive and data-driven that
knowing what leaked from just a stack trace is tough.  If I could just
examine that memory it would be a lot easier.

Anyway.

Please give it another whirl.