Re: toolchain/38315: make: ?= should resolve LHS

["Simon J. Gerraty" <sjg%crufty.net@localhost>]

> I am not sure about calling VarFind twice.  What happens in this case:

Revise patch (below) avoids calling VarFind twice,
but makes no difference to your test case - though what did you
want/expect in the output? @echo 'V.$$ = ${V.$$}' ?

If I make your test case:

DOLLAR = $$
LBRACE = {
RBRACE = }
V.$ = FOO
V.$$ = foo
V.$$${LBRACE}DOLLAR${RBRACE} = bar
V.${DOLLAR} ?= baz

all:
        @echo '${V.$$}'

gmake outputs 'foo'
freebsd make outputs ''
netbsd make (with[out] patch) outputs 'FOO'

Not sure what to do about that one.
For the original test case

gmake,freebsd make and ours with the patch output 'OK'
without the patch we get 'bug'

Index: var.c
===================================================================
RCS file: /cvsroot/src/usr.bin/make/var.c,v
retrieving revision 1.127
diff -u -p -r1.127 var.c
--- var.c       15 Feb 2008 21:29:50 -0000      1.127
+++ var.c       27 Mar 2008 15:53:10 -0000
@@ -872,9 +872,15 @@ Boolean
 Var_Exists(const char *name, GNode *ctxt)
 {
     Var                  *v;
+    char          *cp;
 
-    v = VarFind(name, ctxt, FIND_CMD|FIND_GLOBAL|FIND_ENV);
-
+    if ((cp = strchr(name, '$')) != NULL) {
+       cp = Var_Subst(NULL, name, ctxt, FALSE);
+    }
+    v = VarFind(cp ? cp : name, ctxt, FIND_CMD|FIND_GLOBAL|FIND_ENV);
+    if (cp != NULL) {
+       free(cp);
+    }
     if (v == (Var *)NIL) {
        return(FALSE);
     } else {