Fix problem which only seems to occur on Core Duo: we were relying on unspecified behavior, and that is SHL setting the overflow flag with a shift count greater than one

library/compiler/x86/architecture.factor

@@ -27,10 +27,15 @@ M: cs-loc v>operand cs-loc-n cs-reg reg-stack ;
 : %alien-invoke ( symbol dll -- )
     2dup dlsym CALL rel-relative rel-dlsym ;
 
+: with-aligned-stack ( n quot -- )
+    #! On Linux, there is no requirement to align stack frames,
+    #! so this is mostly a no-op.
+    swap slip stack-reg swap ADD ; inline
+
 : compile-c-call* ( symbol dll args -- operands )
-    <reversed>
+    dup length cells [
-    [ [ PUSH ] each %alien-invoke ] keep
+        <reversed> [ PUSH ] each %alien-invoke
-    [ drop EDX POP ] each ;
+    ] with-aligned-stack ;
 
 GENERIC: push-return-reg ( reg-class -- )
 GENERIC: pop-return-reg ( reg-class -- )

library/compiler/x86/assembler.factor

@@ -352,6 +352,9 @@ M: operand CMP OCT: 071 2-operand ;
 : DIV  ( dst -- ) BIN: 110 t HEX: f7 1-operand ;
 : IDIV ( src -- ) BIN: 111 t HEX: f7 1-operand ;
 
+GENERIC: IMUL2 ( dst src -- )
+M: integer IMUL2 swap dup reg-code t HEX: 69 immediate-1/4 ;
+
 : CDQ HEX: 99 assemble-1 ;
 : CQO HEX: 48 assemble-1 CDQ ;
 

library/compiler/x86/intrinsics.factor

@@ -216,7 +216,7 @@ IN: compiler
     "y" operand "x" operand MOV
     ! Tag the value, since division cancelled tags from both
     ! inputs
-    "x" operand tag-bits SHL
+    "x" operand 1 tag-bits shift IMUL2
     ! Did it overflow?
     "end" get JNO
     ! There was an overflow, so make ECX into a bignum. we must