Index: src/atomic.h
==================================================================
--- src/atomic.h
+++ src/atomic.h
@@ -280,11 +280,11 @@
 #if !defined(OF_HAVE_THREADS)
 	return ++*p;
 #elif defined(OF_HAVE_GCC_ATOMIC_OPS)
 	return __sync_add_and_fetch(p, 1);
 #elif defined(OF_X86_64_ASM) || defined(OF_X86_ASM)
-	uint32_t i;
+	int32_t i;
 
 	__asm__ __volatile__ (
 	    "xorl	%0, %0\n\t"
 	    "incl	%0\n\t"
 	    "lock\n\t"
@@ -351,11 +351,11 @@
 #if !defined(OF_HAVE_THREADS)
 	return --*p;
 #elif defined(OF_HAVE_GCC_ATOMIC_OPS)
 	return __sync_sub_and_fetch(p, 1);
 #elif defined(OF_X86_64_ASM) || defined(OF_X86_ASM)
-	uint32_t i;
+	int32_t i;
 
 	__asm__ __volatile__ (
 	    "xorl	%0, %0\n\t"
 	    "decl	%0\n\t"
 	    "lock\n\t"
@@ -387,11 +387,11 @@
 		    "movl	%2, %0\n\t"
 		    "movl	%0, %%eax\n\t"
 		    "orl	%1, %0\n\t"
 		    "lock\n\t"
 		    "cmpxchg	%0, %2\n\t"
-		    "jne	0\n\t"
+		    "jne	0b"
 		    : "=&r"(i)
 		    : "r"(i), "m"(*p)
 		    : "eax", "cc"
 		);
 # ifdef OF_X86_64_ASM
@@ -401,11 +401,11 @@
 		    "movq	%2, %0\n\t"
 		    "movq	%0, %%rax\n\t"
 		    "orq	%1, %0\n\t"
 		    "lock\n\t"
 		    "cmpxchg	%0, %2\n\t"
-		    "jne	0\n\t"
+		    "jne	0b"
 		    : "=&r"(i)
 		    : "r"(i), "m"(*p)
 		    : "rax", "cc"
 		);
 # endif
@@ -433,11 +433,11 @@
 	    "movl	%2, %0\n\t"
 	    "movl	%0, %%eax\n\t"
 	    "orl	%1, %0\n\t"
 	    "lock\n\t"
 	    "cmpxchg	%0, %2\n\t"
-	    "jne	0\n\t"
+	    "jne	0b"
 	    : "=&r"(i)
 	    : "r"(i), "m"(*p)
 	    : "eax", "cc"
 	);
 
@@ -463,11 +463,11 @@
 		    "movl	%2, %0\n\t"
 		    "movl	%0, %%eax\n\t"
 		    "andl	%1, %0\n\t"
 		    "lock\n\t"
 		    "cmpxchg	%0, %2\n\t"
-		    "jne	0\n\t"
+		    "jne	0b"
 		    : "=&r"(i)
 		    : "r"(i), "m"(*p)
 		    : "eax", "cc"
 		);
 # ifdef OF_X86_64_ASM
@@ -477,11 +477,11 @@
 		    "movq	%2, %0\n\t"
 		    "movq	%0, %%rax\n\t"
 		    "andq	%1, %0\n\t"
 		    "lock\n\t"
 		    "cmpxchg	%0, %2\n\t"
-		    "jne	0\n\t"
+		    "jne	0b"
 		    : "=&r"(i)
 		    : "r"(i), "m"(*p)
 		    : "rax", "cc"
 		);
 # endif
@@ -509,11 +509,11 @@
 	    "movl	%2, %0\n\t"
 	    "movl	%0, %%eax\n\t"
 	    "andl	%1, %0\n\t"
 	    "lock\n\t"
 	    "cmpxchg	%0, %2\n\t"
-	    "jne	0\n\t"
+	    "jne	0b"
 	    : "=&r"(i)
 	    : "r"(i), "m"(*p)
 	    : "eax", "cc"
 	);
 
@@ -539,11 +539,11 @@
 		    "movl	%2, %0\n\t"
 		    "movl	%0, %%eax\n\t"
 		    "xorl	%1, %0\n\t"
 		    "lock\n\t"
 		    "cmpxchg	%0, %2\n\t"
-		    "jne	0\n\t"
+		    "jne	0b"
 		    : "=&r"(i)
 		    : "r"(i), "m"(*p)
 		    : "eax", "cc"
 		);
 # ifdef OF_X86_64_ASM
@@ -553,11 +553,11 @@
 		    "movq	%2, %0\n\t"
 		    "movq	%0, %%rax\n\t"
 		    "xorq	%1, %0\n\t"
 		    "lock\n\t"
 		    "cmpxchg	%0, %2\n\t"
-		    "jne	0\n\t"
+		    "jne	0b"
 		    : "=&r"(i)
 		    : "r"(i), "m"(*p)
 		    : "rax", "cc"
 		);
 # endif
@@ -585,11 +585,11 @@
 	    "movl	%2, %0\n\t"
 	    "movl	%0, %%eax\n\t"
 	    "xorl	%1, %0\n\t"
 	    "lock\n\t"
 	    "cmpxchgl	%0, %2\n\t"
-	    "jne	0\n\t"
+	    "jne	0b"
 	    : "=&r"(i)
 	    : "r"(i), "m"(*p)
 	    : "eax", "cc"
 	);
 
@@ -619,11 +619,11 @@
 	__asm__ __volatile__ (
 	    "lock\n\t"
 	    "cmpxchg	%2, %3\n\t"
 	    "sete	%b0\n\t"
 	    "movzbl	%b0, %0"
-	    : "=&d"(r), "+a"(o)	/* use d instead of r due to gcc bug */
+	    : "=&d"(r), "+a"(o)	/* use d instead of r to avoid a gcc bug */
 	    : "r"(n), "m"(*p)
 	    : "cc"
 	);
 
 	return r;
@@ -652,11 +652,11 @@
 	__asm__ __volatile__ (
 	    "lock\n\t"
 	    "cmpxchg	%2, %3\n\t"
 	    "sete	%b0\n\t"
 	    "movzbl	%b0, %0"
-	    : "=&d"(r), "+a"(o)	/* use d instead of r due to gcc bug */
+	    : "=&d"(r), "+a"(o)	/* use d instead of r to avoid a gcc bug */
 	    : "r"(n), "m"(*p)
 	    : "cc"
 	);
 
 	return r;
@@ -685,11 +685,11 @@
 	__asm__ __volatile__ (
 	    "lock\n\t"
 	    "cmpxchg	%2, %3\n\t"
 	    "sete	%b0\n\t"
 	    "movzbl	%b0, %0"
-	    : "=&d"(r), "+a"(o)	/* use d instead of r due to gcc bug */
+	    : "=&d"(r), "+a"(o)	/* use d instead of r to avoid a gcc bug */
 	    : "r"(n), "m"(*p)
 	    : "cc"
 	);
 
 	return r;