More optimisations

git-svn: trunk@3213

ChangeLog

@@ -1,3 +1,7 @@
+Wed Sep 12 13:36:37 BST 2007 (njh)
+----------------------------------
+  * libclamav:	More optimisations
+
 Tue Sep 11 10:33:21 BST 2007 (njh)
 ----------------------------------
   * libclamav:	Various code clean ups and optimisations

libclamav/line.c

@@ -69,53 +69,6 @@ static	char	const	rcsid[] = "$Id: line.c,v 1.11 2007/02/12 20:46:08 njh Exp $";
 #include "line.h"
 #include "others.h"
 
-#ifdef	OLD
-line_t *
-lineCreate(const char *data)
-{
-	line_t *ret = (line_t *)cli_malloc(sizeof(struct line));
-
-	if(ret == NULL)
-		return NULL;
-
-	ret->l_str = strdup(data);
-	if(ret->l_str == NULL) {
-		free(ret);
-		return NULL;
-	}
-	ret->l_refs = 1;
-
-	return ret;
-}
-
-line_t *
-lineLink(line_t *line)
-{
-	line->l_refs++;
-	return line;
-}
-
-line_t *
-lineUnlink(line_t *line)
-{
-	/*printf("%d:\n\t'%s'\n", line->l_refs, line->l_str);*/
-
-	if(--line->l_refs == 0) {
-		free(line->l_str);
-		free(line);
-		return NULL;
-	}
-	return line;
-}
-
-const char *
-lineGetData(const line_t *line)
-{
-	return line ? line->l_str : NULL;
-}
-
-#else
-
 line_t *
 lineCreate(const char *data)
 {
@@ -163,10 +116,3 @@ lineGetData(const line_t *line)
 {
 	return line ? &line[1] : NULL;
 }
-
-unsigned char
-lineGetRefCount(const line_t *line)
-{
-	return (unsigned char)line[0];
-}
-#endif

libclamav/line.h

@@ -37,21 +37,13 @@
 #ifndef __LINE_H
 #define __LINE_H
 
-#ifdef	OLD
-/* easier to read, but slower */
-
-typedef struct line {
-	char	*l_str;	/* the line's contents */
-	unsigned int	l_refs;	/* the number of references to the data */
-} line_t;
-#else
 typedef	char	line_t;	/* first byte is the ref count */
-#endif
 
 line_t	*lineCreate(const char *data);
 line_t	*lineLink(line_t *line);
 line_t	*lineUnlink(line_t *line);
 const	char	*lineGetData(const line_t *line);
-unsigned	char	lineGetRefCount(const line_t *line);
+
+#define	lineGetRefCount(line)	((unsigned char)line[0])
 
 #endif

libclamav/mbox.c

@@ -1705,8 +1705,6 @@ parseEmailFile(FILE *fin, const table_t *rfc821, const char *firstLine, const ch
 		return NULL;
 	}
 
-	messageClean(ret);
-
 	cli_dbgmsg("parseEmailFile: return\n");
 
 	return ret;
@@ -1715,12 +1713,8 @@ parseEmailFile(FILE *fin, const table_t *rfc821, const char *firstLine, const ch
 /*
  * The given message contains a raw e-mail.
  *
- * Returns the message's body with the correct arguments set
+ * Returns the message's body with the correct arguments set, empties the
- *
+ * given message's contents (note that it isn't destroyed)
- * The downside of this approach is that for a short time we have two copies
- * of the message in memory, the upside is that it makes for easier parsing
- * of encapsulated messages, and in the long run uses less memory in those
- * scenarios
  *
  * TODO: remove the duplication with parseEmailFile
  */
@@ -1729,7 +1723,7 @@ parseEmailHeaders(message *m, const table_t *rfc821)
 {
 	bool inHeader = TRUE;
 	bool bodyIsEmpty = TRUE;
-	const text *t;
+	text *t;
 	message *ret;
 	bool anyHeadersFound = FALSE;
 	int commandNumber = -1;
@@ -1814,13 +1808,15 @@ parseEmailHeaders(message *m, const table_t *rfc821)
 					fullline = ptr;
 					strcat(fullline, line);
 				}
-
 				assert(fullline != NULL);
 
 				if(next_is_folded_header(t))
 					/* Add arguments to this line */
 					continue;
 
+				lineUnlink(t->t_line);
+				t->t_line = NULL;
+
 				if(count_quotes(fullline) & 1)
 					continue;
 
@@ -1853,9 +1849,9 @@ parseEmailHeaders(message *m, const table_t *rfc821)
 			}
 			/*if(t->t_line && isuuencodebegin(t->t_line))
 				puts("FIXME: add fast visa here");*/
-			/*cli_dbgmsg("Add line to body '%s'\n", line);*/
+			cli_dbgmsg("parseEmailHeaders: inished with headers, moving body\n");
-			if(messageAddLine(ret, t->t_line) < 0)
+			messageMoveText(ret, t, m);
-				break;
+			break;
 		}
 	}
 
@@ -1878,8 +1874,6 @@ parseEmailHeaders(message *m, const table_t *rfc821)
 		return NULL;
 	}
 
-	messageClean(ret);
-
 	cli_dbgmsg("parseEmailHeaders: return\n");
 
 	return ret;
@@ -2444,8 +2438,6 @@ parseEmailBody(message *messageIn, text *textIn, mbox_ctx *mctx, unsigned int re
 						if(rc == VIRUS)
 							infected = TRUE;
 						break;
-					default:
-						messageClean(aMessage);
 				}
 			}
 
@@ -2543,8 +2535,7 @@ parseEmailBody(message *messageIn, text *textIn, mbox_ctx *mctx, unsigned int re
 						assert(aMessage == messages[htmltextPart]);
 						messageDestroy(aMessage);
 						messages[htmltextPart] = NULL;
-					}
+					} else if(rc == VIRUS) {
-					else if(rc == VIRUS) {
 						infected = TRUE;
 						break;
 					}
@@ -3312,7 +3303,6 @@ parseMimeHeader(message *m, const char *cmd, const table_t *rfc821Table, const c
 				cli_dbgmsg("Invalid content-type '%s' received, no subtype specified, assuming text/plain; charset=us-ascii\n", ptr);
 			else {
 				int i;
-				char *mimeArgs;	/* RHS of the ; */
 
 				buf = cli_malloc(strlen(ptr) + 1);
 				if(buf == NULL) {
@@ -3406,10 +3396,10 @@ parseMimeHeader(message *m, const char *cmd, const table_t *rfc821Table, const c
 				 * we find the boundary argument set it
 				 */
 				i = 1;
-				while((mimeArgs = cli_strtokbuf(ptr, i++, ";", buf)) != NULL) {
+				while(cli_strtokbuf(ptr, i++, ";", buf) != NULL) {
-					cli_dbgmsg("mimeArgs = '%s'\n", mimeArgs);
+					cli_dbgmsg("mimeArgs = '%s'\n", buf);
 
-					messageAddArguments(m, mimeArgs);
+					messageAddArguments(m, buf);
 				}
 			}
 			break;

libclamav/message.c

@@ -682,7 +682,7 @@ messageGetFilename(const message *m)
 	if(filename)
 		return filename;
 
-	return (char *)messageFindArgument(m, "name"); 
+	return (char *)messageFindArgument(m, "name");
 }
 
 void
@@ -963,6 +963,83 @@ messageAddStrAtTop(message *m, const char *data)
 	return 1;
 }
 
+/*
+ * Put the contents of the given text at the end of the current object.
+ * Can be used either to move a text object into a message, or to move a
+ * message's text into another message only moving from a given offset.
+ * The given text emptied; it can be used again if needed, though be warned that
+ * it will have an empty line at the start.
+ * Returns 0 for failure, 1 for success
+ */
+int
+messageMoveText(message *m, text *t, message *old_message)
+{
+	int rc;
+
+	if(m->body_first == NULL) {
+		if(old_message) {
+			text *u;
+			/*
+			 * t is within old_message which is about to be
+			 * destroyed
+			 */
+			assert(old_message->body_first != NULL);
+
+			m->body_first = t;
+			for(u = old_message->body_first; u != t;) {
+				text *next;
+
+				if(u->t_line)
+					lineUnlink(u->t_line);
+				next = u->t_next;
+
+				free(u);
+				u = next;
+
+				if(u == NULL) {
+					cli_errmsg("messageMoveText sanity check: t not within old_message\n");
+					return -1;
+				}
+			}
+			assert(m->body_last->t_next == NULL);
+
+			m->body_last = old_message->body_last;
+			old_message->body_first = old_message->body_last = NULL;
+
+			/* Do any pointers need to be reset? */
+			if((old_message->bounce == NULL) &&
+			   (old_message->encoding == NULL) &&
+			   (old_message->binhex == NULL) &&
+			   (old_message->yenc == NULL))
+				return 0;
+
+			m->body_last = m->body_first;
+			rc = 0;
+		} else {
+			m->body_last = m->body_first = textMove(NULL, t);
+			if(m->body_first == NULL)
+				rc = -1;
+			else
+				rc = 0;
+		}
+	} else {
+		m->body_last = textMove(m->body_last, t);
+		if(m->body_last == NULL) {
+			rc = -1;
+			m->body_last = m->body_first;
+		} else
+			rc = 0;
+	}
+
+	while(m->body_last->t_next) {
+		m->body_last = m->body_last->t_next;
+		if(m->body_last->t_line)
+			messageIsEncoding(m);
+	}
+
+	return rc;
+}
+
 /*
  * See if the last line marks the start of a non MIME inclusion that
  * will need to be scanned
@@ -974,11 +1051,6 @@ messageIsEncoding(message *m)
 	static const char binhex[] = "(This file must be converted with BinHex 4.0)";
 	const char *line = lineGetData(m->body_last->t_line);
 
-	/* not enough matches to warrant this test */
-	/*if(lineGetRefCount(m->body_last->t_line) > 1) {
-		return;
-	}*/
-
 	if((m->encoding == NULL) &&
 	   (strncasecmp(line, encoding, sizeof(encoding) - 1) == 0) &&
 	   (strstr(line, "7bit") == NULL))
@@ -1014,18 +1086,6 @@ messageGetBody(message *m)
 	return m->body_first;
 }
 
-/*
- * Clean up the message by removing trailing spaces and blank lines
- */
-void
-messageClean(message *m)
-{
-	text *newEnd = textClean(m->body_first);
-
-	if(newEnd)
-		m->body_last = newEnd;
-}
-
 /*
  * Export a message using the given export routines
  *

libclamav/message.h

@@ -65,8 +65,8 @@ encoding_type	messageGetEncoding(const message *m);
 int	messageAddLine(message *m, line_t *line);
 int	messageAddStr(message *m, const char *data);
 int	messageAddStrAtTop(message *m, const char *data);
+int	messageMoveText(message *m, text *t, message *old_message);
 text	*messageGetBody(message *m);
-void	messageClean(message *m);
 unsigned	char	*base64Flush(message *m, unsigned char *buf);
 fileblob	*messageToFileblob(message *m, const char *dir, int destroy);
 blob	*messageToBlob(message *m, int destroy);

libclamav/text.c

@@ -134,18 +134,6 @@ textDestroy(text *t_head)
 	}
 }
 
-/*
- * Remove trailing spaces from the lines and trailing blank lines
- * This could be used to remove trailing blank lines, empty lines etc.,
- *	but it probably isn't worth the time taken given that it won't reclaim
- *	much memory
- */
-text *
-textClean(text *t_head)
-{
-	return t_head;
-}
-
 /* Clone the current object */
 static text *
 textCopy(const text *t_head)
@@ -424,8 +412,7 @@ addToFileblob(const line_t *line, void *arg)
 	if(line) {
 		const char *l = lineGetData(line);
 
-		if(l)
+		fileblobAddData(fb, (const unsigned char *)l, strlen(l));
-			fileblobAddData(fb, (const unsigned char *)l, strlen(l));
 	}
 	fileblobAddData(fb, (const unsigned char *)"\n", 1);
 }

libclamav/text.h

@@ -49,7 +49,6 @@ typedef struct text {
 #include "message.h"
 
 void	textDestroy(text *t_head);
-text	*textClean(text *t_head);
 text	*textAddMessage(text *aText, message *aMessage);
 text	*textMove(text *t_head, text *t);
 blob	*textToBlob(text *t, blob *b, int destroy);