This is the mail archive of the
libc-hacker@sourceware.cygnus.com
mailing list for the glibc project.
Re: malloc debug functions
- To: roland@frob.com
- Subject: Re: malloc debug functions
- From: Wolfram Gloger <wmglo@dent.med.uni-muenchen.de>
- Date: 1 Sep 1998 10:45:13 -0000
- CC: libc-hacker@gnu.org, mcculley@cuspy.com
- References: <199808311707.NAA11127@baalperazim.frob.com>
> > Gene McCulley <mcculley@cuspy.com> writes:
> >
> > > The 2.0.6 version NEWS file mentions malloc_object_allocated_size,
> > > malloc_find_object_address, and malloc_walk. The ChangeLog mentions
> > > the creation of malloc-size.c and malloc-walk.c, but they don't
> > > exist.
>
> That is obvious. What should also be obvious is that such a
> user-incompatible change should not be made without warning and
> explanation. At the very least, their disappearance should be in noted
> NEWS. If it is feasible to implement them in the new malloc, then they
> should be added.
For the new malloc:
malloc_object_allocated_size() already exists -- it's just called
malloc_usable_size() now (maybe an alias could be added ?).
malloc_find_object_address() would be extremely slow if implemented,
and could not work for mmap()ed chunks, as far as I can see.
Similarly for malloc_walk() -- it could not work properly if the heap
is fragmented by foreign sbrk() calls, or for mmap()ed chunks.
My question is: Were these functions available in a widely used glibc
release such that it would be worth to spend time on implementing
these functions for the new releases ? I would actually prefer having
a new stand-alone GNU malloc release with these functions (the last
one I have, from late 1994, doesn't seem to have them).
Regards,
Wolfram.
--
`Surf the sea, not double-u three...'
wmglo@dent.med.uni-muenchen.de