Re: [PATCH v10 00/11] malloc hooks removal

[Stafford Horne <shorne@gmail.com>]

On Sat, Oct 9, 2021 at 8:24 AM DJ Delorie <dj@redhat.com> wrote:
>
> Stafford Horne <shorne@gmail.com> writes:
> > Is the issue in malloc/malloc-debug.c?  There is a big if for
> > GLIBC_2_0 -> GLIBC_2_34.
> > It seems from GLIBC_2_35 nothing is available any longer.  Is that
> > correct, that is what is happening.
>
> Yes.  Old ports get compatibility symbols that do nothing.  New ports
> get no symbols since there are no old apps that would have been linked
> against them.
>
> If yours is the first new port since we made this change, there might be
> some debugging needed...

I think maybe this is the issue.  The first #if is not terminated, and
it keeps the include of mtrace.c/mcheck.c from being included.  Those
includes should provide the real implementations of mtrace/muntrace to
be overloaded with the LD_PRELOAD.

I am not sure where we should end the if though.  I can test something.

#if SHLIB_COMPAT (libc_malloc_debug, GLIBC_2_0, GLIBC_2_34)
/* Support only the glibc allocators.  */
extern void *__libc_malloc (size_t);
extern void __libc_free (void *);
extern void *__libc_realloc (void *, size_t);
extern void *__libc_memalign (size_t, size_t);
extern void *__libc_valloc (size_t);
extern void *__libc_pvalloc (size_t);
extern void *__libc_calloc (size_t, size_t);

....

#include "mcheck.c"
#include "mtrace.c"
#include "malloc-check.c"

#if SHLIB_COMPAT (libc_malloc_debug, GLIBC_2_0, GLIBC_2_24)
extern void (*__malloc_initialize_hook) (void);
compat_symbol_reference (libc, __malloc_initialize_hook,
                         __malloc_initialize_hook, GLIBC_2_0);
#endif