Re: AW: [Xenomai-help] undefinedreference __xnarch_xchg_called_with_bad_pointer

[Heikki Lindholm <holindho@domain.hid>]

[-- Attachment #1: Type: text/plain, Size: 760 bytes --]

Philippe Gerum kirjoitti:
> Heikki Lindholm wrote:
> 
>> Uh. I guess the compiler isn't killing the *_bad_pointer function as 
>> dead code when debug is enabled (although it is dead-code.)
> 
> 
> The optimizer is disabled by --enable-debug so that we don't get strange 
> jumps while tracing with GDB, I guess that's why the dead code is not 
> eliminated. What about just killing this check, since the same code is 
> already validated when compiling it as a kernel module? (except 
> asm-uvm/system.h, but that's not important).

Yep. That seems like the way to go. I don't see much to gain from the 
check. Of course, paranoid way would be to add the __bad_ptr() function 
with abort() or something inside. Quick patch attached for the former case.

- hl

[-- Attachment #2: xeno-atomic-badptr.patch --]
[-- Type: text/plain, Size: 550 bytes --]

--- atomic.h.orig	2006-02-28 21:01:59.000000000 +0200
+++ atomic.h	2006-02-28 21:01:34.000000000 +0200
@@ -142,12 +142,6 @@
 }
 #endif
 
-/*
- * This function doesn't exist, so you'll get a linker error
- * if something tries to do an invalid xchg().
- */
-extern void __xnarch_xchg_called_with_bad_pointer(void);
-
 static __inline__ unsigned long
     __xchg(volatile void *ptr, unsigned long x, unsigned int size)
 {
@@ -159,7 +153,6 @@
 	return __xchg_u64(ptr, x);
 #endif
     }
-    __xnarch_xchg_called_with_bad_pointer();
     return x;
 }