Guarantee that the floating-point register stack is empty when moving

from C to Scheme and vice-versa.  This appears to fix a floating-point
stack overflow exception that has been observed under unix and OS/2.

diff --git a/v7/src/microcode/cmpauxmd/i386.m4 b/v7/src/microcode/cmpauxmd/i386.m4
index 012d138c36d36de7f50478538415e104606f6eed..72e04715a266542ecf292defe89f2da96774471d 100644 (file)
--- a/v7/src/microcode/cmpauxmd/i386.m4
+++ b/v7/src/microcode/cmpauxmd/i386.m4
@@ -1,6 +1,6 @@
 ### -*-Midas-*-
 ###
-###    $Id: i386.m4,v 1.35 1995/01/06 17:39:39 cph Exp $
+###    $Id: i386.m4,v 1.36 1995/02/26 03:34:36 cph Exp $
 ###
 ###    Copyright (c) 1992-95 Massachusetts Institute of Technology
 ###
@@ -518,6 +518,18 @@ IFOS2(`    OP(pop,l)       REG(eax)        # Pop struct return into registers
        jmp     IJMP(REG(eax))                          # Invoke handler
 
 define_c_label(interface_to_scheme)
+IF387(`
+       OP(cmp,l)       TW(IMM(0),EDR(i387_presence))
+       je      interface_to_scheme_proceed
+       ffree   ST(0)                                   # Free floating "regs"
+       ffree   ST(1)
+       ffree   ST(2)
+       ffree   ST(3)
+       ffree   ST(4)
+       ffree   ST(5)
+       ffree   ST(6)
+       ffree   ST(7)
+interface_to_scheme_proceed:')
 ifdef(`WINNT',
 `      mov     edi,_Free                               ; Free pointer = %edi
        sub     edi,_winnt_address_delta                ; as a scheme offset
@@ -582,6 +594,7 @@ IF387(`
        ffree   ST(4)
        ffree   ST(5)
        ffree   ST(6)
+       ffree   ST(7)
 interface_to_C_proceed:')
 
        OP(mov,l)       TW(REG(edx),REG(eax))           # Set up result