fix for overload/stringfy and pp_regcomp

(bug found by code inspection)

The code in pp_regcomp to make a mortal copy of the pattern string
in the case of overload or utf8+'use bytes', missed the case of overload
with utf8. Fix this and at the same time simplify the code.

2 files changed, 47 insertions, 23 deletions

diff --git a/lib/overload.t b/lib/overload.t
index 4be12603d6..b9316207ea 100644
--- a/lib/overload.t
+++ b/lib/overload.t
@@ -48,7 +48,7 @@ package main;
 
 $| = 1;
 BEGIN { require './test.pl' }
-plan tests => 4983;
+plan tests => 4995;
 
 use Scalar::Util qw(tainted);
 
@@ -2186,6 +2186,45 @@ fresh_perl_is
       'Error message when sub stub is encountered';
 }
 
+{
+    # check that the right number of stringifications
+    # and the correct un-utf8-ifying happen on regex compile
+    package utf8_match;
+    my $c;
+    use overload '""' => sub { $c++; $_[0][0] ? "^\x{100}\$" : "^A\$"; };
+    my $o = bless [0], 'utf8_match';
+
+    $o->[0] = 0;
+    $c = 0;
+    ::ok("A" =~  "^A\$",	"regex stringify utf8=0 ol=0 bytes=0");
+    ::ok("A" =~ $o,		"regex stringify utf8=0 ol=1 bytes=0");
+    ::is($c, 1,			"regex stringify utf8=0 ol=1 bytes=0 count");
+
+    $o->[0] = 1;
+    $c = 0;
+    ::ok("\x{100}" =~ "^\x{100}\$",
+				"regex stringify utf8=1 ol=0 bytes=0");
+    ::ok("\x{100}" =~ $o,	"regex stringify utf8=1 ol=1 bytes=0");
+    ::is($c, 1,			"regex stringify utf8=1 ol=1 bytes=0 count");
+
+    use bytes;
+
+    $o->[0] = 0;
+    $c = 0;
+    ::ok("A" =~  "^A\$",	"regex stringify utf8=0 ol=0 bytes=1");
+    ::ok("A" =~ $o,		"regex stringify utf8=0 ol=1 bytes=1");
+    ::is($c, 1,			"regex stringify utf8=0 ol=1 bytes=1 count");
+
+    $o->[0] = 1;
+    $c = 0;
+    ::ok("\xc4\x80" =~ "^\x{100}\$",
+				"regex stringify utf8=1 ol=0 bytes=1");
+    ::ok("\xc4\x80" =~ $o,	"regex stringify utf8=1 ol=1 bytes=1");
+    ::is($c, 1,			"regex stringify utf8=1 ol=1 bytes=1 count");
+
+
+}
+
 { # undefining the overload stash -- KEEP THIS TEST LAST
     package ant;
     use overload '+' => 'onion';
diff --git a/pp_ctl.c b/pp_ctl.c
index d4709b4f33..213f0638fa 100644
--- a/pp_ctl.c
+++ b/pp_ctl.c
@@ -205,28 +205,13 @@ PP(pp_regcomp)
 	    if (PL_op->op_flags & OPf_SPECIAL)
 		PL_reginterp_cnt = I32_MAX; /* Mark as safe.  */
 
-	    if (DO_UTF8(tmpstr)) {
-		assert (SvUTF8(tmpstr));
-	    } else if (SvUTF8(tmpstr)) {
-		/* Not doing UTF-8, despite what the SV says. Is this only if
-		   we're trapped in use 'bytes'?  */
-		/* Make a copy of the octet sequence, but without the flag on,
-		   as the compiler now honours the SvUTF8 flag on tmpstr.  */
-		STRLEN len;
-		const char *const p = SvPV(tmpstr, len);
-		tmpstr = newSVpvn_flags(p, len, SVs_TEMP);
-	    }
-	    else if (SvAMAGIC(tmpstr)) {
-		/* make a copy to avoid extra stringifies */
-		tmpstr = newSVpvn_flags(t, len, SVs_TEMP | SvUTF8(tmpstr));
-	    }
-
-	    /* If it is gmagical, create a mortal copy, but without calling
-	       get-magic, as we have already done that. */
-	    if(SvGMAGICAL(tmpstr)) {
-		SV *mortalcopy = sv_newmortal();
-		sv_setsv_flags(mortalcopy, tmpstr, 0);
-		tmpstr = mortalcopy;
+	    if ((SvUTF8(tmpstr) && IN_BYTES)
+		    || SvGMAGICAL(tmpstr) || SvAMAGIC(tmpstr))
+	    {
+		/* make a temporary copy; either to avoid repeating
+		 * get-magic, or overloaded stringify, or to convert to bytes */
+		tmpstr = newSVpvn_flags(t, len, SVs_TEMP |
+					    (IN_BYTES ? 0 : SvUTF8(tmpstr)));
 	    }
 
 	    if (eng)