From: Nicholas Clark <nick@ccl4.org>
Date: Fri, 11 Jul 2008 19:42:14 +0000 (+0000)
Subject: In Perl_sv_utf8_upgrade_flags(), don't assume that the SV is well
X-Git-Url: http://git.shadowcat.co.uk/gitweb/gitweb.cgi?a=commitdiff_plain;h=4612962a4858c44dd33e3b80c5cbd3f836ae0808;p=p5sagit%2Fp5-mst-13.2.git

In Perl_sv_utf8_upgrade_flags(), don't assume that the SV is well
formed with a trailing '\0'. And do assume that bytes_to_utf8() does.

p4raw-id: //depot/perl@34128
---

diff --git a/sv.c b/sv.c
index 0f6903c..6ac3408 100644
--- a/sv.c
+++ b/sv.c
@@ -3166,13 +3166,21 @@ Perl_sv_utf8_upgrade_flags(pTHX_ register SV *const sv, const I32 flags)
 	    const U8 ch = *t++;
 	    /* Check for hi bit */
 	    if (!NATIVE_IS_INVARIANT(ch)) {
-		STRLEN len = SvCUR(sv) + 1; /* Plus the \0 */
+		STRLEN len = SvCUR(sv);
+		/* *Currently* bytes_to_utf8() adds a '\0' after every string
+		   it converts. This isn't documented. It's not clear if it's
+		   a bad thing to be doing, and should be changed to do exactly
+		   what the documentation says. If so, this code will have to
+		   be changed.
+		   As is, we mustn't rely on our incoming SV being well formed
+		   and having a trailing '\0', as certain code in pp_formline
+		   can send us partially built SVs. */
 		U8 * const recoded = bytes_to_utf8((U8*)s, &len);
 
 		SvPV_free(sv); /* No longer using what was there before. */
 		SvPV_set(sv, (char*)recoded);
-		SvCUR_set(sv, len - 1);
-		SvLEN_set(sv, len); /* No longer know the real size. */
+		SvCUR_set(sv, len);
+		SvLEN_set(sv, len + 1); /* No longer know the real size. */
 		break;
 	    }
 	}