This is the mail archive of the
ecos-discuss@sources.redhat.com
mailing list for the eCos project.
RE: redefinition of `struct mallinfo' when building application
- To: <ecos-discuss at sources dot redhat dot com>
- Subject: RE: [ECOS] redefinition of `struct mallinfo' when building application
- From: "Ivan Jakab" <ivan at INTESIS dot hr>
- Date: Thu, 17 May 2001 17:49:55 +0200
- Cc: "Jonathan Larmour" <jlarmour at redhat dot com>
>
> Ivan Jakab wrote:
> >
> > How come I have no malloc.h in my eCos build? Did I miss
> something in the
> > configuration file? How come "struct mallinfo" is defined in
> stdlib.h? Sorry
> > for all this questions... They're just jumping from the top of
> my mind...
> > ;-)
>
> Yes, this came up internally recently. Here's what I said then:
>
> -=-=-=-
> The mallinfo() stuff could
> be moved to a separate malloc.h within isoinfra/ at any point we want,
> independently of the users of it.
>
> But there is the problem that mallinfo() is not in POSIX, nor in the
> upcoming revised POSIX standard incorporating some SVID stuff, nor at
> http://www.opengroup.org/onlinepubs/7908799/toc.htm. So I simply
> don't know
> what any standard requests, or whether it is, in fact, completely
> implementation defined. Certainly I believe linux uses <malloc.h> whereas
> BSD uses <sys/malloc.h> so the issue is not clear.
>
> Certainly we do not want to provide all the functions already in
> <malloc.h>. Only mallinfo(). Which is also okay for EL/IX level 1.
>
> So until I have accurate information, I don't want to provide a new header
> which we may have to obsolete in future, which is why, for now, it is in
> stdlib.h.
> -=-=-=-=-
>
OK. Thanx for the info...
Regards,
Ivan