This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
Re: [PATCH v2 1/2] Add nmalloc and nrealloc
- From: Rich Felker <dalias at aerifal dot cx>
- To: Florian Weimer <fweimer at redhat dot com>
- Cc: Paul Eggert <eggert at cs dot ucla dot edu>, OndÅej BÃlka <neleai at seznam dot cz>, libc-alpha at sourceware dot org
- Date: Mon, 4 Nov 2013 15:01:30 -0500
- Subject: Re: [PATCH v2 1/2] Add nmalloc and nrealloc
- Authentication-results: sourceware.org; auth=none
- References: <20131031164614 dot GA28117 at domone dot podge> <20131031202932 dot 8942074699 at topped-with-meat dot com> <20131103165411 dot GA27987 at domone dot podge> <5276AB5B dot 8020208 at cs dot ucla dot edu> <52775E1C dot 6080901 at redhat dot com>
On Mon, Nov 04, 2013 at 09:43:08AM +0100, Florian Weimer wrote:
> On 11/03/2013 09:00 PM, Paul Eggert wrote:
>
> >I would avoid macros whose arguments are types; in this
> >case they're not worth the extra confusion caused by the fact that
> >they're not function-like. Just have the callers invoke nmalloc,
> >with no NMALLOC macro. It's easier and less error-prone to
> >read and write code like this:
> >
> > new_global = nmalloc (new_nalloc, sizeof *new_global);
> >
> >than code like this:
> >
> > new_global = NMALLOC (new_nalloc, struct link_map *);
>
> The downside is that
>
> new_global = nmalloc (new_nalloc, sizeof new_global);
>
> type-checks (and may even be correct on some architectures), but
> this fails reliably at compile time:
>
> new_global = NMALLOC (new_nalloc, struct link_map);
>
> I understand the reservations about magic macros, though.
What's worse still, if the macro takes a type and passes it to sizeof,
a caller might inadvertently pass sizeof(type) itself, causing the
macro to evaluate sizeof(sizeof(type)), i.e. sizeof(size_t), instead
of the desired size.
Rich