This is the mail archive of the mailing list for the GCC project.

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]


On Fri, May 30, 2003 at 08:14:50PM +0200, Andreas Jaeger wrote:
> ALLOCATE_TRAMPOLINE is not used anymore:
> ./config/d30v/d30v.h:   which the trampoline is for.  Normally (when `ALLOCATE_TRAMPOLINE' is not
> ./config/d30v/d30v.h:/* #define ALLOCATE_TRAMPOLINE(FP) */
> ./doc/tm.texi:@findex ALLOCATE_TRAMPOLINE
> ./doc/tm.texi:@item ALLOCATE_TRAMPOLINE (@var{fp})
> ./doc/tm.texi:@code{ALLOCATE_TRAMPOLINE} is not defined), the stack slot for the
> ./function.c:#ifdef ALLOCATE_TRAMPOLINE
> ./function.c:  tramp = ALLOCATE_TRAMPOLINE (fp);
> Ok to commit the appended patch if bootstrap on i686-linux-gnu passes?
> Andreas
> 2003-05-30  Andreas Jaeger  <>
> 	* function.c (trampoline_address): Remove ALLOCATE_TRAMPOLINE
> 	usage.
> 	* doc/tm.texi (Trampolines): Remove ALLOCATE_TRAMPOLINE.
> 	* config/d30v/d30v.h: Remove traces of ALLOCATE_TRAMPOLINE.

Actually ALLOCATE_TRAMPOLINE has never worked properly.  I spent some time in
the 1990-ish time frame (when I was MIPS maintiainer at OSF) trying to get it
to work, and never could track down all of the hidden assumptions to it.  Which
is somewhat of a shame, since there are systems out there that you just can't
make the stack executable, or the instruction space is in a completely
different address space than the data space.  Also, there have been exploits
over the years that relied on smashing the stack, and changing the return
address to the stack to get executable code.  When I did the d30v.h port, I
made a complete scan through the tm.texi file for all documented macros, and
put them into d30v.h.

Michael Meissner

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]