From: Hugo van der Sanden <hv@crypt.org>
Date: Sun, 20 Aug 2000 07:30:46 +0000 (+0100)
Subject: UTF8 concat fixes.
X-Git-Url: http://git.shadowcat.co.uk/gitweb/gitweb.cgi?a=commitdiff_plain;h=15bb2692a112e58a65a3e260753db74fb4044704;p=p5sagit%2Fp5-mst-13.2.git

UTF8 concat fixes.

Subject: [PATCH @6713] Re: [ID 20000815.006] latest patched perl core dumps
Message-Id: <200008200630.HAA18053@crypt.compulink.co.uk>

p4raw-id: //depot/perl@6719
---

diff --git a/pp_hot.c b/pp_hot.c
index 1b5f278..0a0c084 100644
--- a/pp_hot.c
+++ b/pp_hot.c
@@ -158,14 +158,11 @@ PP(pp_concat)
             /* Set TARG to PV(left), then add right */
             U8 *l, *c, *olds = NULL;
             STRLEN targlen;
+	    s = (U8*)SvPV(right,len);
             if (TARG == right) {
-                /* Need a safe copy elsewhere since we're just about to
-                   write onto TARG */
-		olds = (U8*)SvPV(right,len);
-                s = (U8*)savepv((char*)olds);
+		/* Take a copy since we're about to overwrite TARG */
+		olds = s = (U8*)savepvn((char*)s, len);
 	    }
-            else
-                s = (U8*)SvPV(right,len);
             l = (U8*)SvPV(left, targlen);
             if (TARG != left)
                 sv_setpvn(TARG, (char*)l, targlen);
@@ -175,14 +172,14 @@ PP(pp_concat)
             targlen = SvCUR(TARG) + len;
             if (!right_utf) {
                 /* plus one for each hi-byte char if we have to upgrade */
-                for (c = s; *c; c++)  {
+                for (c = s; c < s + len; c++)  {
                     if (*c & 0x80)
                         targlen++;
                 }
             }
             SvGROW(TARG, targlen+1);
             /* And now copy, maybe upgrading right to UTF8 on the fly */
-            for (c = (U8*)SvEND(TARG); *s; s++) {
+            for (c = (U8*)SvEND(TARG); len--; s++) {
                  if (*s & 0x80 && !right_utf)
                      c = uv_to_utf8(c, *s);
                  else
diff --git a/t/op/append.t b/t/op/append.t
index d115146..afaf6a1 100755
--- a/t/op/append.t
+++ b/t/op/append.t
@@ -2,7 +2,7 @@
 
 # $RCSfile: append.t,v $$Revision: 4.1 $$Date: 92/08/07 18:27:36 $
 
-print "1..3\n";
+print "1..13\n";
 
 $a = 'ab' . 'c';	# compile time
 $b = 'def';
@@ -19,3 +19,38 @@ $_ = $a;
 $_ .= $b;
 print "#3\t:$_: eq :abcdef:\n";
 if ($_ eq 'abcdef') {print "ok 3\n";} else {print "not ok 3\n";}
+
+# test that when right argument of concat is UTF8, and is the same
+# variable as the target, and the left argument is not UTF8, it no
+# longer frees the wrong string.
+{
+    sub r2 {
+	my $string = '';
+	$string .= pack("U0a*", 'mnopqrstuvwx');
+	$string = "abcdefghijkl$string";
+    }
+
+    r2() and print "ok $_\n" for qw/ 4 5 /;
+}
+
+# test that nul bytes get copied
+{
+    my($a, $ab) = ("a", "a\000b");
+    my($u, $ub) = map pack("U0a*", $_), $a, $ab;
+    my $t1 = $a; $t1 .= $ab;
+    print $t1 =~ /b/ ? "ok 6\n" : "not ok 6\t# $t1\n";
+    my $t2 = $a; $t2 .= $ub;
+    print $t2 =~ /b/ ? "ok 7\n" : "not ok 7\t# $t2\n";
+    my $t3 = $u; $t3 .= $ab;
+    print $t3 =~ /b/ ? "ok 8\n" : "not ok 8\t# $t3\n";
+    my $t4 = $u; $t4 .= $ub;
+    print $t4 =~ /b/ ? "ok 9\n" : "not ok 9\t# $t4\n";
+    my $t5 = $a; $t5 = $ab . $t5;
+    print $t5 =~ /b/ ? "ok 10\n" : "not ok 10\t# $t5\n";
+    my $t6 = $a; $t6 = $ub . $t6;
+    print $t6 =~ /b/ ? "ok 11\n" : "not ok 11\t# $t6\n";
+    my $t7 = $u; $t7 = $ab . $t7;
+    print $t7 =~ /b/ ? "ok 12\n" : "not ok 12\t# $t7\n";
+    my $t8 = $u; $t8 = $ub . $t8;
+    print $t8 =~ /b/ ? "ok 13\n" : "not ok 13\t# $t8\n";
+}