frame_dummy and the 68000.

[Detlef Voss]

Hi!

Chris Johns wrote:
> 
> Hi,
> 
> Getting ELF C++ static constructors to run on a 68302 I am getting an
> illegal instruction exception in frame_dummy (crtstuff.c in gcc
> sources). I have marked the code which is not supported on the 68000. I
> built the compiler but I see it is in the RPMs as well.
> 
> Is this a multilib bug in gcc ?
> 
> Does a patch exist ?
> 
> In :
[snip: assemblerstuff]
>   3c:   4a8a            tstl %a2
> >>>>>>>>>>>             ^^^^^^^^ is not valid on a 68000
> 

While porting RTEMS to a funny Dragonball VZ based screenphone I run
into the same trouble. The default processor is the 68020, but this
change has obviously a side effect and the lib lacks support for the
68000. As no one will use such an expensive processor, try to change it
back! ;)
I fiddled in gcc/config/m68k/m68k-none.h
There You will find something like 

#define TARGET_CPU_DEFAULT M68K_CPU ...

change it to M68K_CPU_m68000 and build the gcc again.
This is obviously no good solution, but it should work.

The next ´tools related´ problem I ran into was gdb & the old HP stub
(found in the efi68k stuff). After running into an breakpoint, the
programmcounter will point _behind_ the trap instruction, so continue
will fail. But as I did not apply the rtems patch to gdb I do not know
if this is already fixed. So I would be interested in knowing if this is
still a problem on the 68000.

BRgd
	Detlef 

======================================================================
Computers are like airconditioners.
They stop working properly if you open WINDOWS!