r2159: converted samba4 over to UTF-16.

I had previously thought this was unnecessary, as windows doesn't use
standards compliant UTF-16, and for filesystem operations treats bytes
as UCS-2, but Bjoern Jacke has pointed out to me that this means we
don't correctly store extended UTF-16 characters as UTF-8 on
disk. This can be seen with (for example) the gothic characters with
codepoints above 64k.

This commit also adds a LOCAL-ICONV torture test that tests the first
1 million codepoints against the system iconv library, and tests 5
million random UTF-16LE buffers for identical error handling to the
system iconv library.

the lib/iconv.c changes need backporting to samba3
(This used to be commit 756f28ac95feaa84b42402723d5f7286865c78db)

3 files changed, 216 insertions, 84 deletions

diff --git a/source4/lib/charcnv.c b/source4/lib/charcnv.c
index 8204a3bcdd..2109e957d9 100644
--- a/source4/lib/charcnv.c
+++ b/source4/lib/charcnv.c
@@ -47,12 +47,12 @@ static const char *charset_name(charset_t ch)
 {
 	const char *ret = NULL;
 
-	if (ch == CH_UCS2) ret = "UTF-16LE";
+	if (ch == CH_UTF16) ret = "UTF-16LE";
 	else if (ch == CH_UNIX) ret = lp_unix_charset();
 	else if (ch == CH_DOS) ret = lp_dos_charset();
 	else if (ch == CH_DISPLAY) ret = lp_display_charset();
 	else if (ch == CH_UTF8) ret = "UTF8";
-	else if (ch == CH_UCS2BE) ret = "UCS-2BE";
+	else if (ch == CH_UTF16BE) ret = "UTF-16BE";
 
 	if (!ret || !*ret) ret = "ASCII";
 	return ret;
@@ -81,13 +81,13 @@ void init_iconv(void)
 
 	/* so that charset_name() works we need to get the UNIX<->UCS2 going
 	   first */
-	if (!conv_handles[CH_UNIX][CH_UCS2])
-		conv_handles[CH_UNIX][CH_UCS2] = smb_iconv_open(charset_name(CH_UCS2), 
+	if (!conv_handles[CH_UNIX][CH_UTF16])
+		conv_handles[CH_UNIX][CH_UTF16] = smb_iconv_open(charset_name(CH_UTF16), 
 								"ASCII");
 
-	if (!conv_handles[CH_UCS2][CH_UNIX])
-		conv_handles[CH_UCS2][CH_UNIX] = smb_iconv_open("ASCII", 
-								charset_name(CH_UCS2));
+	if (!conv_handles[CH_UTF16][CH_UNIX])
+		conv_handles[CH_UTF16][CH_UNIX] = smb_iconv_open("ASCII", 
+								charset_name(CH_UTF16));
 
 	for (c1=0;c1<NUM_CHARSETS;c1++) {
 		for (c2=0;c2<NUM_CHARSETS;c2++) {
@@ -293,7 +293,7 @@ size_t unix_strupper(const char *src, size_t srclen, char *dest, size_t destlen)
 	size_t size;
 	smb_ucs2_t *buffer;
 	
-	size = convert_string_allocate(CH_UNIX, CH_UCS2, src, srclen,
+	size = convert_string_allocate(CH_UNIX, CH_UTF16, src, srclen,
 				       (void **) &buffer);
 	if (size == -1) {
 		smb_panic("failed to create UCS2 buffer");
@@ -303,7 +303,7 @@ size_t unix_strupper(const char *src, size_t srclen, char *dest, size_t destlen)
 		return srclen;
 	}
 	
-	size = convert_string(CH_UCS2, CH_UNIX, buffer, size, dest, destlen);
+	size = convert_string(CH_UTF16, CH_UNIX, buffer, size, dest, destlen);
 	free(buffer);
 	return size;
 }
@@ -313,7 +313,7 @@ size_t unix_strlower(const char *src, size_t srclen, char *dest, size_t destlen)
 	size_t size;
 	smb_ucs2_t *buffer;
 	
-	size = convert_string_allocate(CH_UNIX, CH_UCS2, src, srclen,
+	size = convert_string_allocate(CH_UNIX, CH_UTF16, src, srclen,
 				       (void **) &buffer);
 	if (size == -1) {
 		smb_panic("failed to create UCS2 buffer");
@@ -322,7 +322,7 @@ size_t unix_strlower(const char *src, size_t srclen, char *dest, size_t destlen)
 		free(buffer);
 		return srclen;
 	}
-	size = convert_string(CH_UCS2, CH_UNIX, buffer, size, dest, destlen);
+	size = convert_string(CH_UTF16, CH_UNIX, buffer, size, dest, destlen);
 	free(buffer);
 	return size;
 }
@@ -461,7 +461,7 @@ ssize_t push_ucs2(const void *base_ptr, void *dest, const char *src, size_t dest
 	/* ucs2 is always a multiple of 2 bytes */
 	dest_len &= ~1;
 
-	len += convert_string(CH_UNIX, CH_UCS2, src, src_len, dest, dest_len);
+	len += convert_string(CH_UNIX, CH_UTF16, src, src_len, dest, dest_len);
 	return len;
 }
 
@@ -480,7 +480,7 @@ ssize_t push_ucs2_talloc(TALLOC_CTX *ctx, smb_ucs2_t **dest, const char *src)
 	size_t src_len = strlen(src)+1;
 
 	*dest = NULL;
-	return convert_string_talloc(ctx, CH_UNIX, CH_UCS2, src, src_len, (const void **)dest);
+	return convert_string_talloc(ctx, CH_UNIX, CH_UTF16, src, src_len, (const void **)dest);
 }
 
 
@@ -498,7 +498,7 @@ ssize_t push_ucs2_allocate(smb_ucs2_t **dest, const char *src)
 	size_t src_len = strlen(src)+1;
 
 	*dest = NULL;
-	return convert_string_allocate(CH_UNIX, CH_UCS2, src, src_len, (void **)dest);	
+	return convert_string_allocate(CH_UNIX, CH_UTF16, src, src_len, (void **)dest);	
 }
 
 /**
@@ -603,7 +603,7 @@ size_t pull_ucs2(const void *base_ptr, char *dest, const void *src, size_t dest_
 	if (src_len != (size_t)-1)
 		src_len &= ~1;
 	
-	ret = convert_string(CH_UCS2, CH_UNIX, src, src_len, dest, dest_len);
+	ret = convert_string(CH_UTF16, CH_UNIX, src, src_len, dest, dest_len);
 	if (dest_len)
 		dest[MIN(ret, dest_len-1)] = 0;
 
@@ -627,7 +627,7 @@ ssize_t pull_ucs2_talloc(TALLOC_CTX *ctx, char **dest, const smb_ucs2_t *src)
 {
 	size_t src_len = (strlen_w(src)+1) * sizeof(smb_ucs2_t);
 	*dest = NULL;
-	return convert_string_talloc(ctx, CH_UCS2, CH_UNIX, src, src_len, (const void **)dest);
+	return convert_string_talloc(ctx, CH_UTF16, CH_UNIX, src, src_len, (const void **)dest);
 }
 
 /**
@@ -642,7 +642,7 @@ ssize_t pull_ucs2_allocate(void **dest, const smb_ucs2_t *src)
 {
 	size_t src_len = (strlen_w(src)+1) * sizeof(smb_ucs2_t);
 	*dest = NULL;
-	return convert_string_allocate(CH_UCS2, CH_UNIX, src, src_len, dest);	
+	return convert_string_allocate(CH_UTF16, CH_UNIX, src, src_len, dest);	
 }
 
 /**
diff --git a/source4/lib/iconv.c b/source4/lib/iconv.c
index c869c253cf..9f6526faa5 100644
--- a/source4/lib/iconv.c
+++ b/source4/lib/iconv.c
@@ -39,8 +39,8 @@
  *
  * Note that the only function provided by iconv is conversion between
  * characters.  It doesn't directly support operations like
- * uppercasing or comparison.  We have to convert to UCS-2 and compare
- * there.
+ * uppercasing or comparison.  We have to convert to UTF-16LE and
+ * compare there.
  *
  * @sa Samba Developers Guide
  **/
@@ -55,10 +55,11 @@ static size_t iconv_copy  (void *,const char **, size_t *, char **, size_t *);
 static size_t iconv_swab  (void *,const char **, size_t *, char **, size_t *);
 
 static const struct charset_functions const builtin_functions[] = {
-	/* windows is really neither UCS-2 not UTF-16 */
+	/* windows is closest to UTF-16 */
 	{"UCS-2LE",  iconv_copy, iconv_copy},
 	{"UTF-16LE",  iconv_copy, iconv_copy},
 	{"UCS-2BE",  iconv_swab, iconv_swab},
+	{"UTF-16BE",  iconv_swab, iconv_swab},
 
 	/* we include the UTF-8 alias to cope with differing locale settings */
 	{"UTF8",   utf8_pull,  utf8_push},
@@ -217,23 +218,23 @@ smb_iconv_t smb_iconv_open(const char *tocode, const char *fromcode)
 #endif
 
 	/* check for conversion to/from ucs2 */
-	if (strcasecmp(fromcode, "UCS-2LE") == 0 && to) {
+	if (strcasecmp(fromcode, "UTF-16LE") == 0 && to) {
 		ret->direct = to->push;
 		return ret;
 	}
-	if (strcasecmp(tocode, "UCS-2LE") == 0 && from) {
+	if (strcasecmp(tocode, "UTF-16LE") == 0 && from) {
 		ret->direct = from->pull;
 		return ret;
 	}
 
 #ifdef HAVE_NATIVE_ICONV
-	if (strcasecmp(fromcode, "UCS-2LE") == 0) {
+	if (strcasecmp(fromcode, "UTF-16LE") == 0) {
 		ret->direct = sys_iconv;
 		ret->cd_direct = ret->cd_push;
 		ret->cd_push = NULL;
 		return ret;
 	}
-	if (strcasecmp(tocode, "UCS-2LE") == 0) {
+	if (strcasecmp(tocode, "UTF-16LE") == 0) {
 		ret->direct = sys_iconv;
 		ret->cd_direct = ret->cd_pull;
 		ret->cd_pull = NULL;
@@ -460,100 +461,231 @@ static size_t iconv_copy(void *cd, const char **inbuf, size_t *inbytesleft,
 static size_t utf8_pull(void *cd, const char **inbuf, size_t *inbytesleft,
 			 char **outbuf, size_t *outbytesleft)
 {
-	while (*inbytesleft >= 1 && *outbytesleft >= 2) {
-		const uint8_t *c = (const uint8_t *)*inbuf;
-		uint8_t *uc = (uint8_t *)*outbuf;
-		int len = 1;
+	size_t in_left=*inbytesleft, out_left=*outbytesleft;
+	const uint8_t *c = (const uint8_t *)*inbuf;
+	uint8_t *uc = (uint8_t *)*outbuf;
 
+	while (in_left >= 1 && out_left >= 2) {
 		if ((c[0] & 0x80) == 0) {
 			uc[0] = c[0];
 			uc[1] = 0;
-		} else if ((c[0] & 0xf0) == 0xe0) {
-			if (*inbytesleft < 3) {
-				DEBUG(0,("short utf8 char\n"));
-				goto badseq;
+			c  += 1;
+			in_left  -= 1;
+			out_left -= 2;
+			uc += 2;
+			continue;
+		}
+
+		if ((c[0] & 0xe0) == 0xc0) {
+			if (in_left < 2 ||
+			    (c[1] & 0xc0) != 0x80) {
+				errno = EILSEQ;
+				goto error;
+			}
+			uc[1] = (c[0]>>2) & 0x7;
+			uc[0] = (c[0]<<6) | (c[1]&0x3f);
+			c  += 2;
+			in_left  -= 2;
+			out_left -= 2;
+			uc += 2;
+			continue;
+		}
+
+		if ((c[0] & 0xf0) == 0xe0) {
+			if (in_left < 3 ||
+			    (c[1] & 0xc0) != 0x80 || 
+			    (c[2] & 0xc0) != 0x80) {
+				errno = EILSEQ;
+				goto error;
 			}
 			uc[1] = ((c[0]&0xF)<<4) | ((c[1]>>2)&0xF);
 			uc[0] = (c[1]<<6) | (c[2]&0x3f);
-			len = 3;
-		} else if ((c[0] & 0xe0) == 0xc0) {
-			if (*inbytesleft < 2) {
-				DEBUG(0,("short utf8 char\n"));
-				goto badseq;
+			c  += 3;
+			in_left  -= 3;
+			out_left -= 2;
+			uc += 2;
+			continue;
+		}
+
+		if ((c[0] & 0xf8) == 0xf0) {
+			unsigned int codepoint;
+			if (in_left < 4 ||
+			    (c[1] & 0xc0) != 0x80 || 
+			    (c[2] & 0xc0) != 0x80 ||
+			    (c[3] & 0xc0) != 0x80) {
+				errno = EILSEQ;
+				goto error;
 			}
-			uc[1] = (c[0]>>2) & 0x7;
-			uc[0] = (c[0]<<6) | (c[1]&0x3f);
-			len = 2;
+			codepoint = 
+				(c[3]&0x3f) | 
+				((c[2]&0x3f)<<6) | 
+				((c[1]&0x3f)<<12) |
+				((c[0]&0x7)<<18);
+			if (codepoint < 0x10000) {
+				/* accept UTF-8 characters that are not
+				   minimally packed, but pack the result */
+				uc[0] = (codepoint & 0xFF);
+				uc[1] = (codepoint >> 8);
+				c += 4;
+				in_left -= 4;
+				out_left -= 2;
+				uc += 2;
+				continue;
+			}
+
+			codepoint -= 0x10000;
+
+			if (out_left < 4) {
+				errno = E2BIG;
+				goto error;
+			}
+
+			uc[0] = (codepoint>>10) & 0xFF;
+			uc[1] = (codepoint>>18) | 0xd8;
+			uc[2] = codepoint & 0xFF;
+			uc[3] = ((codepoint>>8) & 0x3) | 0xdc;
+			c  += 4;
+			in_left  -= 4;
+			out_left -= 4;
+			uc += 4;
+			continue;
 		}
 
-		(*inbuf)  += len;
-		(*inbytesleft)  -= len;
-		(*outbytesleft) -= 2;
-		(*outbuf) += 2;
+		/* we don't handle 5 byte sequences */
+		errno = EINVAL;
+		goto error;
 	}
 
-	if (*inbytesleft > 0) {
+	if (in_left > 0) {
 		errno = E2BIG;
-		return -1;
+		goto error;
 	}
-	
+
+	*inbytesleft = in_left;
+	*outbytesleft = out_left;
+	*inbuf = c;
+	*outbuf = uc;	
 	return 0;
 
-badseq:
-	errno = EINVAL;
+error:
+	*inbytesleft = in_left;
+	*outbytesleft = out_left;
+	*inbuf = c;
+	*outbuf = uc;
 	return -1;
 }
 
 static size_t utf8_push(void *cd, const char **inbuf, size_t *inbytesleft,
-			 char **outbuf, size_t *outbytesleft)
+			char **outbuf, size_t *outbytesleft)
 {
-	while (*inbytesleft >= 2 && *outbytesleft >= 1) {
-		uint8_t *c = (uint8_t *)*outbuf;
-		const uint8_t *uc = (const uint8_t *)*inbuf;
-		int len=1;
-
-		if (uc[1] & 0xf8) {
-			if (*outbytesleft < 3) {
-				DEBUG(0,("short utf8 write\n"));
-				goto toobig;
+	size_t in_left=*inbytesleft, out_left=*outbytesleft;
+	uint8_t *c = (uint8_t *)*outbuf;
+	const uint8_t *uc = (const uint8_t *)*inbuf;
+
+	while (in_left >= 2 && out_left >= 1) {
+		unsigned int codepoint;
+
+		if (uc[1] == 0 && !(uc[0] & 0x80)) {
+			/* simplest case */
+			c[0] = uc[0];
+			in_left  -= 2;
+			out_left -= 1;
+			uc += 2;
+			c  += 1;
+			continue;
+		}
+
+		if ((uc[1]&0xf8) == 0) {
+			/* next simplest case */
+			if (out_left < 2) {
+				errno = E2BIG;
+				goto error;
 			}
-			c[0] = 0xe0 | (uc[1]>>4);
-			c[1] = 0x80 | ((uc[1]&0xF)<<2) | (uc[0]>>6);
-			c[2] = 0x80 | (uc[0]&0x3f);
-			len = 3;
-		} else if (uc[1] | (uc[0] & 0x80)) {
-			if (*outbytesleft < 2) {
-				DEBUG(0,("short utf8 write\n"));
-				goto toobig;
+			c[0] = 0xc0 | (uc[0]>>6) | (uc[1]<<2);
+			c[1] = 0x80 | (uc[0] & 0x3f);
+			in_left  -= 2;
+			out_left -= 2;
+			uc += 2;
+			c  += 2;
+			continue;
+		}
+
+		if ((uc[1] & 0xfc) == 0xdc) {
+			/* its the second part of a 4 byte sequence. Illegal */
+			if (in_left < 4) {
+				errno = EINVAL;
+			} else {
+				errno = EILSEQ;
 			}
-			c[0] = 0xc0 | (uc[1]<<2) | (uc[0]>>6);
-			c[1] = 0x80 | (uc[0]&0x3f);
-			len = 2;
-		} else {
-			c[0] = uc[0];
+			goto error;
 		}
 
+		if ((uc[1] & 0xfc) != 0xd8) {
+			codepoint = uc[0] | (uc[1]<<8);
+			if (out_left < 3) {
+				errno = E2BIG;
+				goto error;
+			}
+			c[0] = 0xe0 | (codepoint >> 12);
+			c[1] = 0x80 | ((codepoint >> 6) & 0x3f);
+			c[2] = 0x80 | (codepoint & 0x3f);
+			
+			in_left  -= 2;
+			out_left -= 3;
+			uc  += 2;
+			c   += 3;
+			continue;
+		}
 
-		(*inbytesleft)  -= 2;
-		(*outbytesleft) -= len;
-		(*inbuf)  += 2;
-		(*outbuf) += len;
+		/* its the first part of a 4 byte sequence */
+		if (in_left < 4) {
+			errno = EINVAL;
+			goto error;
+		}
+		if ((uc[3] & 0xfc) != 0xdc) {
+			errno = EILSEQ;
+			goto error;
+		}
+		codepoint = 0x10000 + (uc[2] | ((uc[3] & 0x3)<<8) | 
+				       (uc[0]<<10) | ((uc[1] & 0x3)<<18));
+		
+		if (out_left < 4) {
+			errno = E2BIG;
+			goto error;
+		}
+		c[0] = 0xf0 | (codepoint >> 18);
+		c[1] = 0x80 | ((codepoint >> 12) & 0x3f);
+		c[2] = 0x80 | ((codepoint >> 6) & 0x3f);
+		c[3] = 0x80 | (codepoint & 0x3f);
+		
+		in_left  -= 4;
+		out_left -= 4;
+		uc       += 4;
+		c        += 4;
 	}
 
-	if (*inbytesleft == 1) {
+	if (in_left == 1) {
 		errno = EINVAL;
-		return -1;
+		goto error;
 	}
 
-	if (*inbytesleft > 1) {
+	if (in_left > 1) {
 		errno = E2BIG;
-		return -1;
+		goto error;
 	}
+
+	*inbytesleft = in_left;
+	*outbytesleft = out_left;
+	*inbuf  = uc;
+	*outbuf = c;
 	
 	return 0;
 
-toobig:
-	errno = E2BIG;
+error:
+	*inbytesleft = in_left;
+	*outbytesleft = out_left;
+	*inbuf  = uc;
+	*outbuf = c;
 	return -1;
 }
 
diff --git a/source4/lib/util_unistr.c b/source4/lib/util_unistr.c
index 2bd990836e..63d68fa12e 100644
--- a/source4/lib/util_unistr.c
+++ b/source4/lib/util_unistr.c
@@ -96,9 +96,9 @@ static int check_dos_char(smb_ucs2_t c)
 	char buf[10];
 	smb_ucs2_t c2 = 0;
 	int len1, len2;
-	len1 = convert_string(CH_UCS2, CH_DOS, &c, 2, buf, sizeof(buf));
+	len1 = convert_string(CH_UTF16, CH_DOS, &c, 2, buf, sizeof(buf));
 	if (len1 == 0) return 0;
-	len2 = convert_string(CH_DOS, CH_UCS2, buf, len1, &c2, 2);
+	len2 = convert_string(CH_DOS, CH_UTF16, buf, len1, &c2, 2);
 	if (len2 != 2) return 0;
 	return (c == c2);
 }