summary refs log tree commit diff
path: root/sysdeps/m68k
diff options
context:
space:
mode:
authorAndreas Schwab <schwab@suse.de>2004-10-09 21:18:03 +0000
committerAndreas Schwab <schwab@suse.de>2004-10-09 21:18:03 +0000
commit4e448be5887f151ec16974e17ae8c2ebf4cb7c47 (patch)
treef7b7ae2a4c36cbe08cde286a287830f71cf26f21 /sysdeps/m68k
parent800c2344f3a77e7bedb15e3ed248429d10236d12 (diff)
downloadglibc-4e448be5887f151ec16974e17ae8c2ebf4cb7c47.tar.gz
glibc-4e448be5887f151ec16974e17ae8c2ebf4cb7c47.tar.xz
glibc-4e448be5887f151ec16974e17ae8c2ebf4cb7c47.zip
Remove __THROW from inline definitions.
Diffstat (limited to 'sysdeps/m68k')
-rw-r--r--sysdeps/m68k/fpu/bits/mathinline.h42
1 files changed, 21 insertions, 21 deletions
diff --git a/sysdeps/m68k/fpu/bits/mathinline.h b/sysdeps/m68k/fpu/bits/mathinline.h
index 1e43e43047..acbac47aac 100644
--- a/sysdeps/m68k/fpu/bits/mathinline.h
+++ b/sysdeps/m68k/fpu/bits/mathinline.h
@@ -1,5 +1,5 @@
 /* Definitions of inline math functions implemented by the m68881/2.
-   Copyright (C) 1991,92,93,94,96,97,98,99,2000,2002, 2003
+   Copyright (C) 1991,92,93,94,96,97,98,99,2000,2002, 2003, 2004
      Free Software Foundation, Inc.
    This file is part of the GNU C Library.
 
@@ -110,7 +110,7 @@
 #if defined __USE_MISC || defined __USE_ISOC99
 # define __inline_mathop(func, op)			\
   __inline_mathop1(double, func, op)			\
-  __inline_mathop1(float, __CONCAT(func,f), op)		\
+  __inline_mathop1(float, __CONCAT(func,f), op)	\
   __inline_mathop1(long double, __CONCAT(func,l), op)
 #else
 # define __inline_mathop(func, op)			\
@@ -118,7 +118,7 @@
 #endif
 
 #define __inline_mathop1(float_type,func, op)				      \
-  __m81_defun (float_type, func, (float_type __mathop_x)) __THROW	      \
+  __m81_defun (float_type, func, (float_type __mathop_x))		      \
   {									      \
     float_type __result;						      \
     __asm("f" __STRING(op) "%.x %1, %0" : "=f" (__result) : "f" (__mathop_x));\
@@ -175,7 +175,7 @@ __inline_mathop(trunc, intrz)
    for the function names.  */
 
 #define __inline_functions(float_type, s)				  \
-__m81_defun (float_type, __CONCAT(__floor,s), (float_type __x))	__THROW	  \
+__m81_defun (float_type, __CONCAT(__floor,s), (float_type __x))	  \
 {									  \
   float_type __result;							  \
   unsigned long int __ctrl_reg;						  \
@@ -191,7 +191,7 @@ __m81_defun (float_type, __CONCAT(__floor,s), (float_type __x))	__THROW	  \
   return __result;							  \
 }									  \
 									  \
-__m81_defun (float_type, __CONCAT(__ceil,s), (float_type __x)) __THROW	  \
+__m81_defun (float_type, __CONCAT(__ceil,s), (float_type __x))	  	  \
 {									  \
   float_type __result;							  \
   unsigned long int __ctrl_reg;						  \
@@ -217,7 +217,7 @@ __inline_functions(long double,l)
 #ifdef __USE_MISC
 
 # define __inline_functions(float_type, s)				  \
-__m81_defun (int, __CONCAT(__isinf,s), (float_type __value)) __THROW	  \
+__m81_defun (int, __CONCAT(__isinf,s), (float_type __value))	  	  \
 {									  \
   /* There is no branch-condition for infinity,				  \
      so we must extract and examine the condition codes manually.  */	  \
@@ -227,7 +227,7 @@ __m81_defun (int, __CONCAT(__isinf,s), (float_type __value)) __THROW	  \
   return (__fpsr & (2 << 24)) ? (__fpsr & (8 << 24) ? -1 : 1) : 0;	  \
 }									  \
 									  \
-__m81_defun (int, __CONCAT(__finite,s), (float_type __value)) __THROW	  \
+__m81_defun (int, __CONCAT(__finite,s), (float_type __value))	  	  \
 {									  \
   /* There is no branch-condition for infinity, so we must extract and	  \
      examine the condition codes manually.  */				  \
@@ -238,7 +238,7 @@ __m81_defun (int, __CONCAT(__finite,s), (float_type __value)) __THROW	  \
 }									  \
 									  \
 __m81_defun (float_type, __CONCAT(__scalbn,s),				  \
-	     (float_type __x, int __n))	__THROW				  \
+	     (float_type __x, int __n))					  \
 {									  \
   float_type __result;							  \
   __asm ("fscale%.l %1, %0" : "=f" (__result) : "dmi" (__n), "0" (__x));  \
@@ -255,7 +255,7 @@ __inline_functions(long double,l)
 #if defined __USE_MISC || defined __USE_XOPEN
 
 # define __inline_functions(float_type, s)				  \
-__m81_defun (int, __CONCAT(__isnan,s), (float_type __value)) __THROW	  \
+__m81_defun (int, __CONCAT(__isnan,s), (float_type __value))	  	  \
 {									  \
   char __result;							  \
   __asm("ftst%.x %1\n"							  \
@@ -275,7 +275,7 @@ __inline_functions(long double,l)
 #ifdef __USE_ISOC99
 
 # define __inline_functions(float_type, s)				  \
-__m81_defun (int, __CONCAT(__signbit,s), (float_type __value)) __THROW	  \
+__m81_defun (int, __CONCAT(__signbit,s), (float_type __value))	  	  \
 {									  \
   /* There is no branch-condition for the sign bit, so we must extract	  \
      and examine the condition codes manually.  */			  \
@@ -285,13 +285,13 @@ __m81_defun (int, __CONCAT(__signbit,s), (float_type __value)) __THROW	  \
   return (__fpsr >> 27) & 1;						  \
 }									  \
 									  \
-__m81_defun (float_type, __CONCAT(__scalbln,s),				  \
-	     (float_type __x, long int __n)) __THROW			  \
+  __m81_defun (float_type, __CONCAT(__scalbln,s),			  \
+	     (float_type __x, long int __n))				  \
 {									  \
   return __CONCAT(__scalbn,s) (__x, __n);				  \
 }									  \
 									  \
-__m81_defun (float_type, __CONCAT(__nearbyint,s), (float_type __x)) __THROW \
+__m81_defun (float_type, __CONCAT(__nearbyint,s), (float_type __x))	  \
 {									  \
   float_type __result;							  \
   unsigned long int __ctrl_reg;						  \
@@ -305,7 +305,7 @@ __m81_defun (float_type, __CONCAT(__nearbyint,s), (float_type __x)) __THROW \
   return __result;							  \
 }									  \
 									  \
-__m81_defun (long int, __CONCAT(__lrint,s), (float_type __x)) __THROW	  \
+__m81_defun (long int, __CONCAT(__lrint,s), (float_type __x))		  \
 {									  \
   long int __result;							  \
   __asm ("fmove%.l %1, %0" : "=dm" (__result) : "f" (__x));		  \
@@ -314,7 +314,7 @@ __m81_defun (long int, __CONCAT(__lrint,s), (float_type __x)) __THROW	  \
 									  \
 __m81_inline float_type							  \
 __m81_u(__CONCAT(__fma,s))(float_type __x, float_type __y,		  \
-			   float_type __z) __THROW			  \
+			   float_type __z)				  \
 {									  \
   return (__x * __y) + __z;						  \
 }
@@ -331,7 +331,7 @@ __inline_functions (long double,l)
 # define __inline_functions(float_type, s)				\
 __m81_inline void							\
 __m81_u(__CONCAT(__sincos,s))(float_type __x, float_type *__sinx,	\
-			      float_type *__cosx) __THROW		\
+			      float_type *__cosx)			\
 {									\
   __asm ("fsincos%.x %2,%1:%0"						\
 	 : "=f" (*__sinx), "=f" (*__cosx) : "f" (__x));			\
@@ -351,14 +351,14 @@ __inline_functions (long double,l)
 /* Note that there must be no whitespace before the argument passed for
    NAME, to make token pasting work correctly with -traditional.  */
 # define __inline_forward_c(rettype, name, args1, args2)	\
-extern __inline rettype __attribute__((__const__))	\
-name args1 __THROW					\
-{							\
-  return __CONCAT(__,name) args2;			\
+extern __inline rettype __attribute__((__const__))		\
+  name args1							\
+{								\
+  return __CONCAT(__,name) args2;				\
 }
 
 # define __inline_forward(rettype, name, args1, args2)	\
-extern __inline rettype name args1 __THROW		\
+extern __inline rettype name args1			\
 {							\
   return __CONCAT(__,name) args2;			\
 }