Re: [PATCH] Don't run personality syscall at configure time; don't check it at all

[Pedro Alves <pedro@palves.net>]

On 2021-05-12 7:24 p.m., Tom Tromey wrote:
>>>>>> "Pedro" == Pedro Alves <pedro@palves.net> writes:
> 
> Pedro> "man 2 personality" says:
> 
> Pedro>        This system call first appeared in Linux 1.1.20 (and thus first
> Pedro>        in a stable kernel release with Linux 1.2.0); library support
> Pedro>        was added in glibc 2.3.
> 
> Pedro> ...
> 
> Pedro>        ADDR_NO_RANDOMIZE (since Linux 2.6.12)
> Pedro>               With this flag set, disable address-space-layout randomization.
> 
> Pedro> glibc 2.3 was released in 2002.
> Pedro> Linux 2.6.12 was released in 2005.
> 
> It turns out that this caused the build to fail on RH ES5.

Wow, talk about ancient.  Sorry about the breakage.  That's no longer maintained by RH,
so I don't think we/RH really care about it.  But if you have to support it, fine with
me to keep GDB building there, especially as it doesn't block anything or is complicated.

> 
> Address space randomization works fine there, and 'personality' exists
> -- but ADDR_NO_RANDOMIZE is only defined in <linux/personality.h>, not
> <sys/personality.h>.
> 
> I have a patch to work around this by restoring this check:
> 
> Pedro> -  AC_CHECK_DECLS([ADDR_NO_RANDOMIZE],,, [#include <sys/personality.h>])
> 
> ... and the corresponding define in nat/linux-personality.c.
> 
> (I tried just including the linux/ header but it doesn't declare
> 'personality'.)

Would it work to just include both?

Otherwise, do we really need the configure check?  Wouldn't just:

#ifndef ADDR_NO_RANDOMIZE
# define ADDR_NO_RANDOMIZE 0x0040000
#endif

work the same?

> 
> I don't know if anybody else cares about this old system.  If so, let me
> know and I can send the patch.  Otherwise I may not bother.

Up to you.

Thanks,
Pedro Alves