This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: mudflap problem
- From: fche at redhat dot com (Frank Ch. Eigler)
- To: Doug Graham <dgraham at nortel dot com>
- Cc: gcc at gcc dot gnu dot org
- Date: 03 Feb 2005 15:58:13 -0500
- Subject: Re: mudflap problem
- References: <41FA851F.9080802@ll.mit.edu> <20050128183620.GQ22921@redhat.com><41FA9E91.4030801@ll.mit.edu> <20050128205705.GR22921@redhat.com><20050201185504.GT8746@nortelnetworks.com>
Doug Graham <dgraham@nortel.com> writes:
> [...]
> > That's right. Like the other old bounded-pointers gcc patch (Hi Gary!),
> > the sourceforge-hosted bounds-checker patch uses "fat" pointers for
> > some local pointer variables. This enables this particular check.
>
> Do you mean the patches at http://sourceforge.net/projects/boundschecking?
> If so, I don't think they use fat pointers for anything.
You could be right. I based my comment on a glance at documentation
embedded within the latest patch.
> But they are able to detect a lot of problems that mudflap can't,
> because they 1) add padding between objects, and 2) check all
> pointer arithmetic. [...]
Right: all those extra function calls into the runtime pay for the
extra checking capability.
> I'm trying to figure out how to say this in a way that doesn't make
> me sound ungrateful, but IMO, mudflap just won't be very useful
> until it incorporates a feature like [adding gaps / fattening up
> local pointers]. [...]
It's a fair observation. BTW, how does mudflap's performance compare
to the boundchecker (not having the latter built lately)?
- FChE