cgen-ibld.in 64-bit extract bug

cgen-ibld.in 64-bit extract bug

[Richard Henderson]

CGEN_INSN_INT is usually an unsigned int, while valuep is a long.
Under these conditions, the masking and sign-extension done at the
end of the function is incorrect unless sizeof(int)==sizeof(long).

Ok?


r~


	* cgen-ibld.in (extract_normal): Make value and mask a long.

Index: cgen-ibld.in
===================================================================
RCS file: /cvs/cvsfiles/devo/opcodes/cgen-ibld.in,v
retrieving revision 1.15
diff -c -p -d -r1.15 cgen-ibld.in
*** cgen-ibld.in	2001/07/01 02:49:14	1.15
--- cgen-ibld.in	2001/07/17 05:32:18
*************** extract_normal (cd, ex_info, insn_value,
*** 415,421 ****
  #endif
       long *valuep;
  {
!   CGEN_INSN_INT value, mask;
  
    /* If LENGTH is zero, this operand doesn't contribute to the value
       so give it a standard value of zero.  */
--- 415,421 ----
  #endif
       long *valuep;
  {
!   long value, mask;
  
    /* If LENGTH is zero, this operand doesn't contribute to the value
       so give it a standard value of zero.  */

Re: cgen-ibld.in 64-bit extract bug

[Frank Ch. Eigler]

Hi -

rth wrote:
: CGEN_INSN_INT is usually an unsigned int, while valuep is a long.
: [...]
: Ok?
: 	* cgen-ibld.in (extract_normal): Make value and mask a long.

Sure, that, or changing CGEN_INSN_INT to (unsigned?) long instead.
Are you dealing with a 64-bit host or a 64-bit target?

- FChE
-- 
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.0.6 (GNU/Linux)
Comment: For info see http://www.gnupg.org

iD8DBQE7VB22VZbdDOm/ZT0RAkKaAJ9x+V4WihDaexCN8TVt2REA5ZJoSACffzF2
OrN1dGa4LiPdDJX11IFs6/U=
=0eH1
-----END PGP SIGNATURE-----

Re: cgen-ibld.in 64-bit extract bug

[Richard Henderson]

On Tue, Jul 17, 2001 at 07:12:54AM -0400, Frank Ch. Eigler wrote:
> Sure, that, or changing CGEN_INSN_INT to (unsigned?) long instead.

Whichever you prefer.

> Are you dealing with a 64-bit host or a 64-bit target?

64-bit host, 32-bit target (as it happens).


r~