Re: AM33/2.0 support for mn10300-elf

[Andrew Cagney <ac131313@redhat.com>]

Sigh, separate SIM and GDB patches might have been easier.

For the simulator, two refinements:

+void fpu_disabled_exception     PARAMS ((SIM_DESC, sim_cpu *, 
address_word));
The code needs to use strict ISO C.  It might be easier to just convert 
everything.

+  store_word (addr + 4, data.high);
+}
  #endif  /* not WITH_COMMON */

The MN10300 switched to sim/common long ago, the #if !WITH_COMMON code 
should be deleted.

they can both be considered pretty "obvious".

Other than that, its approved.

--

> +/* Fetch, and possibly build, an appropriate link_map_offsets structure
> +   for mn10300 linux targets using the struct offsets defined in <link.h>.
> +   Note, however, that link.h is not actually referred to in this file.
> +   Instead, the relevant structs offsets were obtained from examining
> +   link.h.  (We can't refer to link.h from this file because the host
> +   system won't necessarily have it, or if it does, the structs which
> +   it defines will refer to the host system, not the target.)  */
> +
> +struct link_map_offsets *
> +mn10300_linux_svr4_fetch_link_map_offsets (void)
> +{

This really belongs in mn10300-linux-tdep.c.  I also suspect that stuff 
is also missing from the patch (tm-mn10300.h?), as the function doesn't 
appear to be called?  Does this need an ABI sniffer (see osabi.[sh])?

> +  if (AM33_MODE == 2)

I'm guessing this is a macro that relies on global state.  New code 
instead accesses everything via parameters.  Here, something like:

	fi->arch->tdep->am33_mode

which, in GDB, is written as:

	gdbarch_tdep (get_frame_arch (fi))->am33_mode.

> +      get_frame_saved_regs (fi)[LAR_REGNUM]    = get_frame_base (fi) + offset + 4;

See: http://sources.redhat.com/gdb/current/ari/
The method get_frame_saved_regs() is obsolete.  Changes should be 
decreasing, not increasing that function's usage count :-/  The new code 
will need to be written in a way that avoids this method.  The best way 
of doing this is to convert it to convert that section of the code to 
the new unwind mechanism.

>      15, 16, 17, 18, 19, 20, 21, 22
> +    , 32, 33, 34, 35, 36, 37, 38, 39
> +    , 40, 41, 42, 43, 44, 45, 46, 47

GNU coding style?


+  set_gdbarch_fp0_regnum (gdbarch, 32);

Is this needed?  Adding a field to the am33's tdep struct would be 
better.  FP0_REGNUM is almost, but not quite deprecated.

sorry,
Andrew