Re: [PATCH newlib] libc/include/malloc.h: Add prototype for GNU extension malloc_usable_size()

On Tue, 15 Mar 2022 at 11:44, Joel Sherrill <[email protected]> wrote:

>
>
> On Tue, Mar 15, 2022, 10:38 AM C Howland <[email protected]> wrote:
>
>> >
>> > ------------------------------
>> > *From:* Newlib <[email protected]>
>> on
>> > behalf of Joel Sherrill <[email protected]>
>> > *Sent:* Tuesday, March 15, 2022 11:16 AM
>> > *To:* [email protected] <[email protected]>
>> > *Subject:* [PATCH newlib] libc/include/malloc.h: Add prototype for GNU
>> > extension malloc_usable_size()
>> >
>> >
>> > This is not provided by the newlib malloc implementation but may
>> > be available in external implementations.
>> > ---
>> >  newlib/libc/include/malloc.h | 4 ++++
>> >  1 file changed, 4 insertions(+)
>> >
>> > diff --git a/newlib/libc/include/malloc.h b/newlib/libc/include/malloc.h
>> > index a9dc5bca6..e73095e1e 100644
>> > --- a/newlib/libc/include/malloc.h
>> > +++ b/newlib/libc/include/malloc.h
>> > @@ -137,6 +137,10 @@ extern void __malloc_lock(struct _reent *);
>> >
>> >  extern void __malloc_unlock(struct _reent *);
>> >
>> > +#if __GNU_VISIBLE
>> > +extern size_t malloc_usable_size(void *);
>> > +#endif
>> > +
>> >  /* A compatibility routine for an earlier version of the allocator.  */
>> >
>> >  extern void mstats (char *);
>> > --
>> > 2.24.4
>> >
>> >
>>      Sorry, but I fail to see why adding this makes sense.  If there's an
>> application which adds a function they should be editing the header file
>> when they add it.  (There's not even a stub being supplied here.)
>> Additionally, the Linux man page for the function says "The main use of
>> this function is for debugging and introspection."  That is, it is not
>> even
>> a general application utility, but of very limited use.  Either provide
>> the
>> function with the prototype or no prototype.
>>
>
> RTEMS provides its own malloc family implementation and we are providing
> it. I can wrap all that in an RTEMS conditional if you like.
>
> I have no experience with the malloc implementation in Newlib or any idea
> how to implement it.
>
> --joel
>

RTEMS wrapper is good.  (Sorry, but you didn't mention you were providing
it in RTEMS, although in hindsight I should have assumed so when wording my
response.)  No reason to try to retrofit it to the Newlib implementation.
Craig