Re: libc 5.3.6 BREAKS accelx!

David Holland (dholland@hcs.harvard.edu)
Fri, 15 Mar 1996 20:57:27 -0500 (EST)

> I think it is great that we have a faster and better malloc for libc but
> I think it should not be at the expense of compatibility. I think that we
> should all still be able to write our own mallocs if we want to and not
> have libc break it. So if we can find a way to still use dl-malloc and
> have other mallocs coexist with it, that would be great.

You're missing a point, namely, that any replacement malloc that
doesn't implement valloc has always been broken; it's just that the
breakage is only visible with the new malloc.

Suggestion: all the places in libc that call valloc should free the
memory with vfree(), which is to be a new function.

This removes the problems with calling __libc_valloc directly. There
is still a cost in compatibility, this time probably with POSIX
(someone who has the relevant POSIX want to share what it says about
valloc, or tell us it doesn't?)

-- 
   - David A. Holland          | Number of words in the English language that
     dholland@hcs.harvard.edu  | exist because of typos or misreadings: 381