Are __BOUNDED_POINTERS__ supported?

Are __BOUNDED_POINTERS__ supported?

[Steven Munroe]

Are __BOUNDED_POINTERS__ still a supported feature of glibc?

I've been looking a optimization of the string (strlen/strcmp/strcpy) 
functions for PPC64 and noticed several:

/* GKM FIXME: check high bounds.  */

Which was inherited from PPC32. Also that the usage of the 
__BOUNDED_POINTERS_ macros don't seem correct. For example 
sysdeps/powerpc/powerpc32/strchr.S contains:

L(missed):
	and.	rTMP1, rTMP1, rTMP2
	li	rRTN, 0
	STORE_RETURN_VALUE (rSTR)

which will generate:

	...
	li	rRTN, 0
	stw	rSTR,0(rRTN)

which will genrate a segfault. So bounded pointers are not working for 
pp32/ppc64.

Then I tried using gcc to generate some examples of bounded-pointers and 
bounds-checking using -fbounds-check -fbounded-pointers options. But 
there is no difference in the generated asm (with or without this options).

So are __BOUNDED_POINTERS__ still supported? And should I put effort 
into fixing the PPC32/PPC64 code?

Re: Are __BOUNDED_POINTERS__ supported?

[Ulrich Drepper]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Steven Munroe wrote:
> Are __BOUNDED_POINTERS__ still a supported feature of glibc?

The ppc side was never fully finished.  x86 should still basically work.
 It's not really supported but I am reluctant to remove the code since
it might come in handy for some error schecking scheme.  Feel free to
remove any BP code which you know won't work, though.

- -- 
- --------------.                        ,-.            444 Castro Street
Ulrich Drepper \    ,-----------------'   \ Mountain View, CA 94041 USA
Red Hat         `--' drepper at redhat.com `---------------------------
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.1 (GNU/Linux)

iD8DBQE+gJX82ijCOnn/RHQRApcSAJ9oJU+CNQJ9irkloEwIWmPIlPTRBACdF6PG
FZomgy1gpVVjxZSyUBmZc1o=
=EgOx
-----END PGP SIGNATURE-----