Fix [8dd2807066] follow-up: two-argument 'max' had the same problem

Author: jan.nijtmans

generic/tclExecute.c

@@ -8812,7 +8812,7 @@ TclCompareTwoNumbers(
 	    w2 = (Tcl_WideInt)d2;
 	    goto wideCompare;
 	case TCL_NUMBER_BIG:
-	    Tcl_TakeBignumFromObj(NULL, value2Ptr, &big2);
+	    Tcl_GetBignumFromObj(NULL, value2Ptr, &big2);
 	    if (mp_isneg(&big2)) {
 		compare = MP_GT;
 	    } else {
@@ -8849,7 +8849,7 @@ TclCompareTwoNumbers(
 	    if (isinf(d1)) {
 		return (d1 > 0.0) ? MP_GT : MP_LT;
 	    }
-	    Tcl_TakeBignumFromObj(NULL, value2Ptr, &big2);
+	    Tcl_GetBignumFromObj(NULL, value2Ptr, &big2);
 	    if ((d1 < (double)WIDE_MAX) && (d1 > (double)WIDE_MIN)) {
 		if (mp_isneg(&big2)) {
 		    compare = MP_GT;
@@ -8871,7 +8871,7 @@ TclCompareTwoNumbers(
     break;
 
     case TCL_NUMBER_BIG:
-	Tcl_TakeBignumFromObj(NULL, valuePtr, &big1);
+	Tcl_GetBignumFromObj(NULL, valuePtr, &big1);
 	switch (type2) {
 	case TCL_NUMBER_INT:
 	    compare = mp_cmp_d(&big1, 0);
@@ -8898,7 +8898,7 @@ TclCompareTwoNumbers(
 	    Tcl_InitBignumFromDouble(NULL, d2, &big2);
 	    goto bigCompare;
 	case TCL_NUMBER_BIG:
-	    Tcl_TakeBignumFromObj(NULL, value2Ptr, &big2);
+	    Tcl_GetBignumFromObj(NULL, value2Ptr, &big2);
 	bigCompare:
 	    compare = mp_cmp(&big1, &big2);
 	    mp_clear(&big1);

tests/expr.test

@@ -7460,6 +7460,10 @@ test expr-bug-8dd2807066 {Bug 8dd2807066} {
     expr {max(abs(-668336881543038127783364011867))}
 } 668336881543038127783364011867
 
+test expr-bug-8dd2807066-1 {Bug 8dd2807066, more elaborate} {
+    expr {max(abs(-668336881543038127783364011867),0)}
+} 668336881543038127783364011867
+
 # Bug e3dcab1d14 TODO: Need to work out a test case that fails
 # without tcl_precision, which has been eliminated in 9.0