Revert one hunk of change 32034 that had the possibility of being buggy

(the sprintf "%c" code will work correctly when the SV is UTF-8).
Audit all the rest for UTF-8 correctness, and force SvUTF-8_off() in
utf8.c to ensure correctness. (The string is reset to "", so this will
not be a behaviour change.)

p4raw-id: //depot/perl@32040

diff --git a/dump.c b/dump.c
index bdaf41a..0d7fd34 100644 (file)
--- a/dump.c
+++ b/dump.c
@@ -219,8 +219,10 @@ Perl_pv_escape( pTHX_ SV *dsv, char const * const str,
     const char * const end = pv + count; /* end of string */
     octbuf[0] = esc;
 
-    if (!flags & PERL_PV_ESCAPE_NOCLEAR) 
+    if (!flags & PERL_PV_ESCAPE_NOCLEAR) {
+           /* This won't alter the UTF-8 flag */
            sv_setpvn(dsv, "", 0);
+    }
     
     if ((flags & PERL_PV_ESCAPE_UNI_DETECT) && is_utf8_string((U8*)pv, count))
         isuni = 1;
@@ -279,8 +281,13 @@ Perl_pv_escape( pTHX_ SV *dsv, char const * const str,
             sv_catpvn(dsv, octbuf, chsize);
             wrote += chsize;
        } else {
-           const char string = (char) c;
-            sv_catpvn(dsv, &string, 1);
+           /* If PERL_PV_ESCAPE_NOBACKSLASH is set then bytes in the range
+              128-255 can be appended raw to the dsv. If dsv happens to be
+              UTF-8 then we need catpvf to upgrade them for us.
+              Or add a new API call sv_catpvc(). Think about that name, and
+              how to keep it clear that it's unlike the s of catpvs, which is
+              really an array octets, not a string.  */
+            Perl_sv_catpvf( aTHX_ dsv, "%c", c);
            wrote++;
        }
         if ( flags & PERL_PV_ESCAPE_FIRSTCHAR ) 

diff --git a/regcomp.c b/regcomp.c
index b302561..f649188 100644 (file)
--- a/regcomp.c
+++ b/regcomp.c
@@ -9674,7 +9674,17 @@ clear_re(pTHX_ void *r)
 STATIC void
 S_put_byte(pTHX_ SV *sv, int c)
 {
-    if (isCNTRL(c) || c == 255 || !isPRINT(c))
+    /* Our definition of isPRINT() ignores locales, so only bytes that are
+       not part of UTF-8 are considered printable. I assume that the same
+       holds for UTF-EBCDIC.
+       Also, code point 255 is not printable in either (it's E0 in EBCDIC,
+       which Wikipedia says:
+
+       EO, or Eight Ones, is an 8-bit EBCDIC character code represented as all
+       ones (binary 1111 1111, hexadecimal FF). It is similar, but not
+       identical, to the ASCII delete (DEL) or rubout control character.
+       ) So the old condition can be simplified to !isPRINT(c)  */
+    if (!isPRINT(c))
        Perl_sv_catpvf(aTHX_ sv, "\\%o", c);
     else {
        const unsigned char string = (unsigned char) c;

diff --git a/utf8.c b/utf8.c
index c665a41..45b17b1 100644 (file)
--- a/utf8.c
+++ b/utf8.c
@@ -2136,6 +2136,7 @@ Perl_pv_uni_display(pTHX_ SV *dsv, const U8 *spv, STRLEN len, STRLEN pvlim, UV f
     const char *s, *e;
 
     sv_setpvn(dsv, "", 0);
+    SvUTF8_off(dsv);
     for (s = (const char *)spv, e = s + len; s < e; s += UTF8SKIP(s)) {
         UV u;
          /* This serves double duty as a flag and a character to print after