s3-charcnv: Move convert_string() et al to lib/util/charset

This is the first step to this being the common convert_string
implementation.

Andrew Bartlett

Signed-off-by: Andrew Tridgell <tridge@samba.org>

4 files changed, 4 insertions, 447 deletions

diff --git a/source3/Makefile.in b/source3/Makefile.in
index 2c74ff8cf4..c20323c078 100644
--- a/source3/Makefile.in
+++ b/source3/Makefile.in
@@ -442,7 +442,7 @@ LIB_OBJ = $(LIBSAMBAUTIL_OBJ) $(UTIL_OBJ) $(CRYPTO_OBJ) \
 	  lib/util_transfer_file.o ../lib/async_req/async_sock.o \
 	  lib/addrchange.o \
 	  $(TDB_LIB_OBJ) \
-	  $(VERSION_OBJ) lib/charcnv.o lib/fstring.o ../lib/util/debug.o ../lib/util/debug_s3.o ../lib/util/fault.o \
+	  $(VERSION_OBJ) lib/charcnv.o ../lib/util/charset/convert_string.o lib/fstring.o ../lib/util/debug.o ../lib/util/debug_s3.o ../lib/util/fault.o \
 	  lib/interface.o lib/pidfile.o lib/dumpcore.o \
 	  lib/system.o lib/sendfile.o lib/recvfile.o lib/time.o \
 	  lib/username.o \
diff --git a/source3/include/proto.h b/source3/include/proto.h
index 3f44b949f2..5b96bdfbeb 100644
--- a/source3/include/proto.h
+++ b/source3/include/proto.h
@@ -72,6 +72,7 @@ int bitmap_find(struct bitmap *bm, unsigned ofs);
 
 /* The following definitions come from lib/charcnv.c  */
 
+void lazy_initialize_conv(void);
 void gfree_charcnv(void);
 void init_iconv(void);
 bool convert_string(charset_t from, charset_t to,
diff --git a/source3/lib/charcnv.c b/source3/lib/charcnv.c
index a8719070c3..5c01052d21 100644
--- a/source3/lib/charcnv.c
+++ b/source3/lib/charcnv.c
@@ -22,25 +22,9 @@
 */
 #include "includes.h"
 
-/**
- * @file
- *
- * @brief Character-set conversion routines built on our iconv.
- *
- * @note Samba's internal character set (at least in the 3.0 series)
- * is always the same as the one for the Unix filesystem.  It is
- * <b>not</b> necessarily UTF-8 and may be different on machines that
- * need i18n filenames to be compatible with Unix software.  It does
- * have to be a superset of ASCII.  All multibyte sequences must start
- * with a byte with the high bit set.
- *
- * @sa lib/iconv.c
- */
-
-
 static bool initialized;
 
-static void lazy_initialize_conv(void)
+void lazy_initialize_conv(void)
 {
 	if (!initialized) {
 		load_case_tables_library();
@@ -72,434 +56,6 @@ void init_iconv(void)
 								true, global_iconv_handle);
 }
 
-/**
- * Convert string from one encoding to another, making error checking etc
- * Slow path version - uses (slow) iconv.
- *
- * @param src pointer to source string (multibyte or singlebyte)
- * @param srclen length of the source string in bytes
- * @param dest pointer to destination string (multibyte or singlebyte)
- * @param destlen maximal length allowed for string
- * @param converted size is the number of bytes occupied in the destination
- *
- * @returns false and sets errno on fail, true on success.
- *
- * Ensure the srclen contains the terminating zero.
- *
- **/
-
-static bool convert_string_internal(charset_t from, charset_t to,
-		      void const *src, size_t srclen, 
-		      void *dest, size_t destlen, size_t *converted_size)
-{
-	size_t i_len, o_len;
-	size_t retval;
-	const char* inbuf = (const char*)src;
-	char* outbuf = (char*)dest;
-	smb_iconv_t descriptor;
-	struct smb_iconv_handle *ic;
-
-	lazy_initialize_conv();
-	ic = get_iconv_handle();
-	descriptor = get_conv_handle(ic, from, to);
-
-	if (srclen == (size_t)-1) {
-		if (from == CH_UTF16LE || from == CH_UTF16BE) {
-			srclen = (strlen_w((const smb_ucs2_t *)src)+1) * 2;
-		} else {
-			srclen = strlen((const char *)src)+1;
-		}
-	}
-
-
-	if (descriptor == (smb_iconv_t)-1 || descriptor == (smb_iconv_t)0) {
-		errno = EINVAL;
-		return false;
-	}
-
-	i_len=srclen;
-	o_len=destlen;
-
-	retval = smb_iconv(descriptor, &inbuf, &i_len, &outbuf, &o_len);
-	if (retval == (size_t)-1) {
-		return false;
-	}
-	*converted_size = destlen-o_len;
-	return true;
-}
-
-/**
- * Convert string from one encoding to another, making error checking etc
- * Fast path version - handles ASCII first.
- *
- * @param src pointer to source string (multibyte or singlebyte)
- * @param srclen length of the source string in bytes, or -1 for nul terminated.
- * @param dest pointer to destination string (multibyte or singlebyte)
- * @param destlen maximal length allowed for string - *NEVER* -1.
- * @param converted size is the number of bytes occupied in the destination
- *
- * @returns false and sets errno on fail, true on success.
- *
- * Ensure the srclen contains the terminating zero.
- *
- * This function has been hand-tuned to provide a fast path.
- * Don't change unless you really know what you are doing. JRA.
- **/
-
-bool convert_string_error(charset_t from, charset_t to,
-			    void const *src, size_t srclen,
-			    void *dest, size_t destlen,
-			    size_t *converted_size)
-{
-	/*
-	 * NB. We deliberately don't do a strlen here if srclen == -1.
-	 * This is very expensive over millions of calls and is taken
-	 * care of in the slow path in convert_string_internal. JRA.
-	 */
-
-#ifdef DEVELOPER
-	SMB_ASSERT(destlen != (size_t)-1);
-#endif
-
-	if (srclen == 0) {
-		*converted_size = 0;
-		return true;
-	}
-
-	if (from != CH_UTF16LE && from != CH_UTF16BE && to != CH_UTF16LE && to != CH_UTF16BE) {
-		const unsigned char *p = (const unsigned char *)src;
-		unsigned char *q = (unsigned char *)dest;
-		size_t slen = srclen;
-		size_t dlen = destlen;
-		unsigned char lastp = '\0';
-		size_t retval = 0;
-
-		/* If all characters are ascii, fast path here. */
-		while (slen && dlen) {
-			if ((lastp = *p) <= 0x7f) {
-				*q++ = *p++;
-				if (slen != (size_t)-1) {
-					slen--;
-				}
-				dlen--;
-				retval++;
-				if (!lastp)
-					break;
-			} else {
-#ifdef BROKEN_UNICODE_COMPOSE_CHARACTERS
-				goto general_case;
-#else
-				bool ret = convert_string_internal(from, to, p, slen, q, dlen, converted_size);
-				*converted_size += retval;
-				return ret;
-#endif
-			}
-		}
-
-		*converted_size = retval;
-
-		if (!dlen) {
-			/* Even if we fast path we should note if we ran out of room. */
-			if (((slen != (size_t)-1) && slen) ||
-					((slen == (size_t)-1) && lastp)) {
-				errno = E2BIG;
-				return false;
-			}
-		}
-		return true;
-	} else if (from == CH_UTF16LE && to != CH_UTF16LE) {
-		const unsigned char *p = (const unsigned char *)src;
-		unsigned char *q = (unsigned char *)dest;
-		size_t retval = 0;
-		size_t slen = srclen;
-		size_t dlen = destlen;
-		unsigned char lastp = '\0';
-
-		/* If all characters are ascii, fast path here. */
-		while (((slen == (size_t)-1) || (slen >= 2)) && dlen) {
-			if (((lastp = *p) <= 0x7f) && (p[1] == 0)) {
-				*q++ = *p;
-				if (slen != (size_t)-1) {
-					slen -= 2;
-				}
-				p += 2;
-				dlen--;
-				retval++;
-				if (!lastp)
-					break;
-			} else {
-#ifdef BROKEN_UNICODE_COMPOSE_CHARACTERS
-				goto general_case;
-#else
-				bool ret = convert_string_internal(from, to, p, slen, q, dlen, converted_size);
-				*converted_size += retval;
-				return ret;
-#endif
-			}
-		}
-
-		*converted_size = retval;
-
-		if (!dlen) {
-			/* Even if we fast path we should note if we ran out of room. */
-			if (((slen != (size_t)-1) && slen) ||
-					((slen == (size_t)-1) && lastp)) {
-				errno = E2BIG;
-				return false;
-			}
-		}
-		return true;
-	} else if (from != CH_UTF16LE && from != CH_UTF16BE && to == CH_UTF16LE) {
-		const unsigned char *p = (const unsigned char *)src;
-		unsigned char *q = (unsigned char *)dest;
-		size_t retval = 0;
-		size_t slen = srclen;
-		size_t dlen = destlen;
-		unsigned char lastp = '\0';
-
-		/* If all characters are ascii, fast path here. */
-		while (slen && (dlen >= 2)) {
-			if ((lastp = *p) <= 0x7F) {
-				*q++ = *p++;
-				*q++ = '\0';
-				if (slen != (size_t)-1) {
-					slen--;
-				}
-				dlen -= 2;
-				retval += 2;
-				if (!lastp)
-					break;
-			} else {
-#ifdef BROKEN_UNICODE_COMPOSE_CHARACTERS
-				goto general_case;
-#else
-				bool ret = convert_string_internal(from, to, p, slen, q, dlen, converted_size);
-				*converted_size += retval;
-				return ret;
-#endif
-			}
-		}
-
-		*converted_size = retval;
-
-		if (!dlen) {
-			/* Even if we fast path we should note if we ran out of room. */
-			if (((slen != (size_t)-1) && slen) ||
-					((slen == (size_t)-1) && lastp)) {
-				errno = E2BIG;
-				return false;
-			}
-		}
-		return true;
-	}
-
-#ifdef BROKEN_UNICODE_COMPOSE_CHARACTERS
-  general_case:
-#endif
-	return convert_string_internal(from, to, src, srclen, dest, destlen, converted_size);
-}
-
-bool convert_string(charset_t from, charset_t to,
-		      void const *src, size_t srclen,
-		      void *dest, size_t destlen,
-		      size_t *converted_size)
-{
-	bool ret = convert_string_error(from, to, src, srclen, dest, destlen, converted_size);
-
-	if(ret==false) {
-		const char *reason="unknown error";
-		switch(errno) {
-			case EINVAL:
-				reason="Incomplete multibyte sequence";
-				DEBUG(3,("convert_string_internal: Conversion error: %s(%s)\n",
-					 reason, (const char *)src));
-				break;
-			case E2BIG:
-			{
-				struct smb_iconv_handle *ic;
-				lazy_initialize_conv();
-				ic = get_iconv_handle();
-
-				reason="No more room";
-				if (from == CH_UNIX) {
-					DEBUG(3,("E2BIG: convert_string(%s,%s): srclen=%u destlen=%u - '%s'\n",
-						 charset_name(ic, from), charset_name(ic, to),
-						 (unsigned int)srclen, (unsigned int)destlen, (const char *)src));
-				} else {
-					DEBUG(3,("E2BIG: convert_string(%s,%s): srclen=%u destlen=%u\n",
-						 charset_name(ic, from), charset_name(ic, to),
-						 (unsigned int)srclen, (unsigned int)destlen));
-				}
-				break;
-			}
-			case EILSEQ:
-				reason="Illegal multibyte sequence";
-				DEBUG(3,("convert_string_internal: Conversion error: %s(%s)\n",
-					 reason, (const char *)src));
-				break;
-			default:
-				DEBUG(0,("convert_string_internal: Conversion error: %s(%s)\n",
-					 reason, (const char *)src));
-				break;
-		}
-		/* smb_panic(reason); */
-	}
-	return ret;
-}
-
-
-/**
- * Convert between character sets, allocating a new buffer using talloc for the result.
- *
- * @param srclen length of source buffer.
- * @param dest always set at least to NULL
- * @parm converted_size set to the number of bytes occupied by the string in
- * the destination on success.
- * @note -1 is not accepted for srclen.
- *
- * @return true if new buffer was correctly allocated, and string was
- * converted.
- *
- * Ensure the srclen contains the terminating zero.
- *
- * I hate the goto's in this function. It's embarressing.....
- * There has to be a cleaner way to do this. JRA.
- */
-bool convert_string_talloc(TALLOC_CTX *ctx, charset_t from, charset_t to,
-			   void const *src, size_t srclen, void *dst,
-			   size_t *converted_size)
-
-{
-	size_t i_len, o_len, destlen = (srclen * 3) / 2;
-	size_t retval;
-	const char *inbuf = (const char *)src;
-	char *outbuf = NULL, *ob = NULL;
-	smb_iconv_t descriptor;
-	void **dest = (void **)dst;
-	struct smb_iconv_handle *ic;
-
-	*dest = NULL;
-
-	if (src == NULL || srclen == (size_t)-1) {
-		errno = EINVAL;
-		return false;
-	}
-
-	if (srclen == 0) {
-		/* We really should treat this as an error, but
-		   there are too many callers that need this to
-		   return a NULL terminated string in the correct
-		   character set. */
-		if (to == CH_UTF16LE|| to == CH_UTF16BE || to == CH_UTF16MUNGED) {
-			destlen = 2;
-		} else {
-			destlen = 1;
-		}
-		ob = talloc_zero_array(ctx, char, destlen);
-		if (ob == NULL) {
-			errno = ENOMEM;
-			return false;
-		}
-		*converted_size = destlen;
-		*dest = ob;
-		return true;
-	}
-
-	lazy_initialize_conv();
-	ic = get_iconv_handle();
-	descriptor = get_conv_handle(ic, from, to);
-
-	if (descriptor == (smb_iconv_t)-1 || descriptor == (smb_iconv_t)0) {
-		DEBUG(0,("convert_string_talloc: Conversion not supported.\n"));
-		errno = EOPNOTSUPP;
-		return false;
-	}
-
-  convert:
-
-	/* +2 is for ucs2 null termination. */
-	if ((destlen*2)+2 < destlen) {
-		/* wrapped ! abort. */
-		DEBUG(0, ("convert_string_talloc: destlen wrapped !\n"));
-		TALLOC_FREE(outbuf);
-		errno = EOPNOTSUPP;
-		return false;
-	} else {
-		destlen = destlen * 2;
-	}
-
-	/* +2 is for ucs2 null termination. */
-	ob = (char *)TALLOC_REALLOC(ctx, ob, destlen + 2);
-
-	if (!ob) {
-		DEBUG(0, ("convert_string_talloc: realloc failed!\n"));
-		errno = ENOMEM;
-		return false;
-	}
-	outbuf = ob;
-	i_len = srclen;
-	o_len = destlen;
-
-	retval = smb_iconv(descriptor,
-			   &inbuf, &i_len,
-			   &outbuf, &o_len);
-	if(retval == (size_t)-1) 		{
-	    	const char *reason="unknown error";
-		switch(errno) {
-			case EINVAL:
-				reason="Incomplete multibyte sequence";
-				DEBUG(3,("convert_string_talloc: Conversion error: %s(%s)\n",reason,inbuf));
-				break;
-			case E2BIG:
-				goto convert;
-			case EILSEQ:
-				reason="Illegal multibyte sequence";
-				DEBUG(3,("convert_string_talloc: Conversion error: %s(%s)\n",reason,inbuf));
-				break;
-		}
-		DEBUG(0,("Conversion error: %s(%s)\n",reason,inbuf));
-		/* smb_panic(reason); */
-		TALLOC_FREE(ob);
-		return false;
-	}
-
-	destlen = destlen - o_len;
-	/* Don't shrink unless we're reclaiming a lot of
-	 * space. This is in the hot codepath and these
-	 * reallocs *cost*. JRA.
-	 */
-	if (o_len > 1024) {
-		/* We're shrinking here so we know the +2 is safe from wrap. */
-		ob = (char *)TALLOC_REALLOC(ctx,ob,destlen + 2);
-	}
-
-	if (destlen && !ob) {
-		DEBUG(0, ("convert_string_talloc: out of memory!\n"));
-		errno = ENOMEM;
-		return false;
-	}
-
-	*dest = ob;
-
-	/* Must ucs2 null terminate in the extra space we allocated. */
-	ob[destlen] = '\0';
-	ob[destlen+1] = '\0';
-
-	/* Ensure we can never return a *converted_size of zero. */
-	if (destlen == 0) {
-		/* As we're now returning false on a bad smb_iconv call,
-		   this should never happen. But be safe anyway. */
-		if (to == CH_UTF16LE|| to == CH_UTF16BE || to == CH_UTF16MUNGED) {
-			destlen = 2;
-		} else {
-			destlen = 1;
-		}
-	}
-
-	*converted_size = destlen;
-	return true;
-}
-
 bool unix_strupper(const char *src, size_t srclen, char *dest, size_t destlen)
 {
 	size_t size;
diff --git a/source3/wscript_build b/source3/wscript_build
index 71d91388dd..6eb0c4c464 100755
--- a/source3/wscript_build
+++ b/source3/wscript_build
@@ -963,7 +963,7 @@ bld.SAMBA3_SUBSYSTEM('tdb-wrap3',
                     vars=locals())
 
 bld.SAMBA3_SUBSYSTEM('CHARSET3',
-                    source='''lib/util_str.c lib/charcnv.c lib/fstring.c''',
+                    source='''lib/util_str.c lib/charcnv.c  ../lib/util/charset/convert_string.c lib/fstring.c''',
                     public_deps='ICONV_WRAPPER CODEPOINTS',
                     deps='DYNCONFIG')