Make shared hash keys to be \0-terminated:

one possible resolution for
"UTF-8, weird \w behaviour after HASH-KEY-ification"
http://www.xray.mpe.mpg.de/mailing-lists/perl5-porters/2002-01/msg01327.html
The hash keys were shared (the SvLEN(sv) = 0 was the giveaway).
The hash keys weren't \0-terminated.  This meant that the EOL ($)
in regmatch() got the nextchr beyond the last character.  Since
the keys were UTF-8, the nextchr was \1, not the usual
string-terminating \0.  Wham, no match.

I think another possible resolution could be to stop the nextchr
computation in regmatch() from peeking beyond the last character
of the string:

    nextchr = locinput < PL_regeol ? UCHARAT(locinput) : 0;

p4raw-id: //depot/perl@14908

diff --git a/hv.c b/hv.c
index e4cc6c9..7efa086 100644 (file)
--- a/hv.c
+++ b/hv.c
@@ -85,9 +85,10 @@ S_save_hek(pTHX_ const char *str, I32 len, U32 hash)
       is_utf8 = TRUE;
     }
 
-    New(54, k, HEK_BASESIZE + len + 1, char);
+    New(54, k, HEK_BASESIZE + len + 2, char);
     hek = (HEK*)k;
     Copy(str, HEK_KEY(hek), len, char);
+    HEK_KEY(hek)[len] = 0;
     HEK_LEN(hek) = len;
     HEK_HASH(hek) = hash;
     HEK_UTF8(hek) = (char)is_utf8;

diff --git a/hv.h b/hv.h
index 688663a..369bf3c 100644 (file)
--- a/hv.h
+++ b/hv.h
@@ -23,6 +23,8 @@ struct hek {
     U32                hek_hash;       /* hash of key */
     I32                hek_len;        /* length of hash key */
     char       hek_key[1];     /* variable-length hash key */
+    /* the hash-key is \0-terminated */
+    /* after the \0 there is a byte telling whether the key is UTF8 */
 };
 
 /* hash structure: */
@@ -211,7 +213,7 @@ C<SV*>.
 #define HEK_HASH(hek)          (hek)->hek_hash
 #define HEK_LEN(hek)           (hek)->hek_len
 #define HEK_KEY(hek)           (hek)->hek_key
-#define HEK_UTF8(hek)  (*(HEK_KEY(hek)+HEK_LEN(hek)))
+#define HEK_UTF8(hek)          (*(HEK_KEY(hek)+HEK_LEN(hek)+1))
 
 /* calculate HV array allocation */
 #if defined(STRANGE_MALLOC) || defined(MYMALLOC)