[PATCH] Fix printing functions with complex return on PPC64

[PATCH] Fix printing functions with complex return on PPC64

[Tiago Stürmer Daitx]

This patch fix the current PPC64 return value code to properly handle complex 
types by passing the target type of the COMPLEX value. It also makes use of a
variable to hold TYPE_TARGET_TYPE (valtype) for better readability and reuse.

Previously two testcases from gdb.base/callfuncs.exp and another two from
gdb.base/varargs.exp would fail with GDB Interal Error due to improperly
calling convert_typed_floating with a COMPLEX type instead of using the
target type.

Excerpt from a diff between PPC64 run log before and after the fix:

  -FAIL: gdb.base/callfuncs.exp: call inferior func with struct - returns float _Complex (GDB internal error)
  -FAIL: gdb.base/callfuncs.exp: call inferior func with struct - returns double _Complex (GDB internal error)

  -FAIL: gdb.base/varargs.exp: print find_max_float_real(4, fc1, fc2, fc3, fc4) (GDB internal error)
  -FAIL: gdb.base/varargs.exp: print find_max_double_real(4, dc1, dc2, dc3, dc4) (GDB internal error)
  +FAIL: gdb.base/varargs.exp: print find_max_float_real(4, fc1, fc2, fc3, fc4)

The existing FAIL is due to the complex arguments and will be fixed by another 
patch series.

No regression was detected on PPC (32 bits) testcases.

Regards,
Tiago Daitx


gdb/Changelog

2012-11-01  Tiago Stürmer Daitx  <tdaitx@linux.vnet.ibm.com>

        * ppc-sysv-tdep.c (ppc64_sysv_abi_return_value): Set correct
        type on float conversion for complex type.
---
diff --git a/gdb/ppc-sysv-tdep.c b/gdb/ppc-sysv-tdep.c
index 48b4765..0ffeab9 100644
--- a/gdb/ppc-sysv-tdep.c
+++ b/gdb/ppc-sysv-tdep.c
@@ -1920,11 +1920,13 @@ ppc64_sysv_abi_return_value (struct gdbarch *gdbarch, struct value *function,
 	      gdb_byte regval[MAX_REGISTER_SIZE];
 	      struct type *regtype =
 		register_type (gdbarch, tdep->ppc_fp0_regnum);
+	      struct type *target_type;
+	      target_type = check_typedef (TYPE_TARGET_TYPE (valtype));
 	      if (writebuf != NULL)
 		{
 		  convert_typed_floating ((const bfd_byte *) writebuf +
-					  i * (TYPE_LENGTH (valtype) / 2),
-					  valtype, regval, regtype);
+					  i * TYPE_LENGTH (target_type), 
+					  target_type, regval, regtype);
 		  regcache_cooked_write (regcache,
                                          tdep->ppc_fp0_regnum + 1 + i,
 					 regval);
@@ -1936,8 +1938,8 @@ ppc64_sysv_abi_return_value (struct gdbarch *gdbarch, struct value *function,
                                         regval);
 		  convert_typed_floating (regval, regtype,
 					  (bfd_byte *) readbuf +
-					  i * (TYPE_LENGTH (valtype) / 2),
-					  valtype);
+					  i * TYPE_LENGTH (target_type),
+					  target_type);
 		}
 	    }
 	}

Re: [PATCH] Fix printing functions with complex return on PPC64

[Ulrich Weigand]

Tiago Daitx wrote:

> gdb/Changelog
> 
> 2012-11-01  Tiago Stürmer Daitx  <tdaitx@linux.vnet.ibm.com>
> 
>         * ppc-sysv-tdep.c (ppc64_sysv_abi_return_value): Set correct
>         type on float conversion for complex type.

This is OK.

Bye,
Ulrich

-- 
  Dr. Ulrich Weigand
  GNU Toolchain for Linux on System z and Cell BE
  Ulrich.Weigand@de.ibm.com

Re: [PATCH] Fix printing functions with complex return on PPC64

[Sergio Durigan Junior]

On Wednesday, January 23 2013, Ulrich Weigand wrote:

> Tiago Daitx wrote:
>
>> gdb/Changelog
>> 
>> 2012-11-01  Tiago Stürmer Daitx  <tdaitx@linux.vnet.ibm.com>
>> 
>>         * ppc-sysv-tdep.c (ppc64_sysv_abi_return_value): Set correct
>>         type on float conversion for complex type.
>
> This is OK.

I've committed the patch for Tiago, with fixes in the ChangeLog entry
(wrong date, spaces instead of TABs).

     http://sourceware.org/ml/gdb-cvs/2013-01/msg00150.html

Thanks,

-- 
Sergio

Re: [PATCH] Fix printing functions with complex return on PPC64

[Jan Kratochvil]

On Sat, 17 Nov 2012 16:59:39 +0100, Tiago Stürmer Daitx wrote:
> --- a/gdb/ppc-sysv-tdep.c
> +++ b/gdb/ppc-sysv-tdep.c
> @@ -1925,7 +1925,8 @@ ppc64_sysv_abi_return_value (struct gdbarch *gdbarch, struct value *function,
>  		{
>  		  convert_typed_floating ((const bfd_byte *) writebuf +
>  					  i * (TYPE_LENGTH (valtype) / 2),
> -					  valtype, regval, regtype);
> +					  TYPE_TARGET_TYPE (valtype), regval,
> +					  regtype);
>  		  regcache_cooked_write (regcache,
>                                           tdep->ppc_fp0_regnum + 1 + i,
>  					 regval);
> @@ -1938,7 +1939,7 @@ ppc64_sysv_abi_return_value (struct gdbarch *gdbarch, struct value *function,
>  		  convert_typed_floating (regval, regtype,
>  					  (bfd_byte *) readbuf +
>  					  i * (TYPE_LENGTH (valtype) / 2),
> -					  valtype);
> +					  TYPE_TARGET_TYPE (valtype));
>  		}
>  	    }
>  	}

OK for commit.

You could also replace TYPE_LENGTH (valtype) / 2 there
by TYPE_LENGTH (TYPE_TARGET_TYPE (valtype)) to make it more clear
(use some variable for TYPE_TARGET_TYPE (valtype); not required.


Thanks,
Jan

[PATCH] Fix printing functions with complex return on PPC64

[Tiago Stürmer Daitx]

This patch fix the current PPC64 return value code to properly handle complex 
types by passing the target type of the COMPLEX value.

Previously two testcases from gdb.base/callfuncs.exp and another two from
gdb.base/varargs.exp would fail with GDB Interal Error due to improperly
calling convert_typed_floating with a COMPLEX type instead of using the
target type.

Excerpt from a diff between PPC64 run log before and after the fix:

  -FAIL: gdb.base/callfuncs.exp: call inferior func with struct - returns float _Complex (GDB internal error)
  -FAIL: gdb.base/callfuncs.exp: call inferior func with struct - returns double _Complex (GDB internal error)

  -FAIL: gdb.base/varargs.exp: print find_max_float_real(4, fc1, fc2, fc3, fc4) (GDB internal error)
  -FAIL: gdb.base/varargs.exp: print find_max_double_real(4, dc1, dc2, dc3, dc4) (GDB internal error)
  +FAIL: gdb.base/varargs.exp: print find_max_float_real(4, fc1, fc2, fc3, fc4)

The existing FAIL is due to the complex arguments and will be fixed by another 
patch series.

No regression was detected on PPC (32 bits) testcases.

Regards,
Tiago Daitx


gdb/Changelog

2012-11-01  Tiago Stürmer Daitx  <tdaitx@linux.vnet.ibm.com>

        * ppc-sysv-tdep.c (ppc64_sysv_abi_return_value): Set correct
        type on float conversion for complex type.
---
 gdb/ppc-sysv-tdep.c |    5 +++--
 1 files changed, 3 insertions(+), 2 deletions(-)

diff --git a/gdb/ppc-sysv-tdep.c b/gdb/ppc-sysv-tdep.c
index a221b70..1a0f74d 100644
--- a/gdb/ppc-sysv-tdep.c
+++ b/gdb/ppc-sysv-tdep.c
@@ -1925,7 +1925,8 @@ ppc64_sysv_abi_return_value (struct gdbarch *gdbarch, struct value *function,
 		{
 		  convert_typed_floating ((const bfd_byte *) writebuf +
 					  i * (TYPE_LENGTH (valtype) / 2),
-					  valtype, regval, regtype);
+					  TYPE_TARGET_TYPE (valtype), regval,
+					  regtype);
 		  regcache_cooked_write (regcache,
                                          tdep->ppc_fp0_regnum + 1 + i,
 					 regval);
@@ -1938,7 +1939,7 @@ ppc64_sysv_abi_return_value (struct gdbarch *gdbarch, struct value *function,
 		  convert_typed_floating (regval, regtype,
 					  (bfd_byte *) readbuf +
 					  i * (TYPE_LENGTH (valtype) / 2),
-					  valtype);
+					  TYPE_TARGET_TYPE (valtype));
 		}
 	    }
 	}
-- 
1.7.1