[Axiom-developer] Re: [Gcl-devel] Re: gcl-2.6.7 and FC4

[Camm Maguire]

Greetings!

Here is what we've been doing since early FC:

int
main(int argc, char **argv, char **envp) {

#ifdef NEED_NONRANDOM_SBRK
#if SIZEOF_LONG == 4
        if (!syscall(SYS_personality,PER_LINUX32))
#else
        if (!syscall(SYS_personality,PER_LINUX))
#endif
          execvp(argv[0],argv);
#endif


so either:

1) configure is not detecting the situation and defining
   NEED_NONRANDOM_SBRK
2) the syscall syntax is changed
3) the problem lies somewhere else

it would appear.

Take care,


Juho Snellman <address@hidden> writes:

> <address@hidden> wrote:
> > GCL-2.6.7 fails to build on FC4.
> > I'm trying to debug it now but I thought I'd mention it in case
> > anyone already knows why.
> 
> FC4 randomizes the brk (and a bunch of other memory regions) more
> agressively than the exec-shield patches in previous Fedora kernels.
> Also the old methods of marking the executable as non-randomizable
> don't work any more. To test whether this is the reason, try building
> under setarch ("setarch i386 -R make"). 
> 
> The only working method of disabling the randomization seems to be
> doing something like this at startup:
> 
>     {
>        long pers = personality(-1);
>        /* 0x40000 aka. ADDR_NO_RANDOMIZE */
>        if (!(pers & 0x40000)) {
>            personality(pers | 0x40000);
>            execve(argv[0], argv, envp);
>        }
>     }   
> 
> See the sbcl-devel thread "Memory randomization problems coming" for
> some more information on this.
> 
> -- 
> Juho Snellman
> 
> 
> 
> _______________________________________________
> Gcl-devel mailing list
> address@hidden
> http://lists.gnu.org/mailman/listinfo/gcl-devel
> 
> 
> 

-- 
Camm Maguire                                            address@hidden
==========================================================================
"The earth is but one country, and mankind its citizens."  --  Baha'u'llah