performance improvements: replace memcmp

 - looping forward while moving a pointer and doing memcmp isn't fast. replace it by strstr
 - \r\n can be done in a pointer dereferencing comparison.

Note: maybe we need to fallback to the memcmp way if there is binary content?
In that case we should do a little smarter and try to use contentlength-headers etc.

diff --git a/libcitadel/lib/libcitadel.h b/libcitadel/lib/libcitadel.h
index 557280edd7b38b1f8cdf6057cc4ab5eaa0020547..2070c14fc23b5b69e4ad7d0aadac31118f6b90c1 100644 (file)
--- a/libcitadel/lib/libcitadel.h
+++ b/libcitadel/lib/libcitadel.h
@@ -167,6 +167,11 @@ void mime_decode(char *partnum,
                 MimeParserCallBackType PostMultiPartCallBack,
                 void *userdata,
                 int dont_decode);
+int mime_decode_now (char *part_start, 
+                    size_t length,
+                    char *encoding,
+                    char **decoded,
+                    size_t *bytes_decoded);
 void the_mime_parser(char *partnum,
                     char *content_start, char *content_end,
                     MimeParserCallBackType CallBack,

diff --git a/libcitadel/lib/mime_parser.c b/libcitadel/lib/mime_parser.c
index a9089ccc730170ca051312ad147145e56c136570..50b51497152bca582016d8e3ea4691adc2ff19cc 100644 (file)
--- a/libcitadel/lib/mime_parser.c
+++ b/libcitadel/lib/mime_parser.c
@@ -213,9 +213,17 @@ void mime_decode(char *partnum,
        /* If this part is not encoded, send as-is */
        if ( (strlen(encoding) == 0) || (dont_decode)) {
                if (CallBack != NULL) {
-                       CallBack(name, filename, fixed_partnum(partnum),
-                               disposition, part_start,
-                               content_type, charset, length, encoding, id, userdata);
+                       CallBack(name, 
+                                filename, 
+                                fixed_partnum(partnum),
+                                disposition, 
+                                part_start,
+                                content_type, 
+                                charset, 
+                                length, 
+                                encoding, 
+                                id,
+                                userdata);
                        }
                return;
        }
@@ -253,6 +261,65 @@ void mime_decode(char *partnum,
        free(decoded);
 }
 
+/*
+ * this is the extract of mime_decode which can be called if 'dont_decode' was set; 
+ * to save the cpu intense process of decoding to the time when it realy wants the content. 
+ * returns: 
+ *   - > 0 we decoded something, its on *decoded, you need to free it.
+ *   - = 0 no need to decode stuff. *decoded will be NULL.
+ *   - < 0 an error occured, either an unknown encoding, or alloc failed. no need to free.
+ */
+int mime_decode_now (char *part_start, 
+                    size_t length,
+                    char *encoding,
+                    char **decoded,
+                    size_t *bytes_decoded)
+{
+       *bytes_decoded = 0;
+       *decoded = NULL;
+       /* Some encodings aren't really encodings */
+       if (!strcasecmp(encoding, "7bit"))
+               strcpy(encoding, "");
+       if (!strcasecmp(encoding, "8bit"))
+               strcpy(encoding, "");
+       if (!strcasecmp(encoding, "binary"))
+               strcpy(encoding, "");
+
+       /* If this part is not encoded, send as-is */
+       if (strlen(encoding) == 0) {
+               return 0;
+       }
+       
+
+       /* Fail if we hit an unknown encoding. */
+       if ((strcasecmp(encoding, "base64"))
+           && (strcasecmp(encoding, "quoted-printable"))) {
+               return -1;
+       }
+
+       /*
+        * Allocate a buffer for the decoded data.  The output buffer is slightly
+        * larger than the input buffer; this assumes that the decoded data
+        * will never be significantly larger than the encoded data.  This is a
+        * safe assumption with base64, uuencode, and quoted-printable.
+        */
+       *decoded = malloc(length + 32768);
+       if (decoded == NULL) {
+               return -1;
+       }
+
+       if (!strcasecmp(encoding, "base64")) {
+               *bytes_decoded = CtdlDecodeBase64(*decoded, part_start, length);
+               return 1;
+       }
+       else if (!strcasecmp(encoding, "quoted-printable")) {
+               *bytes_decoded = CtdlDecodeQuotedPrintable(*decoded, part_start, length);
+               return 1;
+       }
+       return -1;
+}
+
+
 /*
  * Break out the components of a multipart message
  * (This function expects to be fed HEADERS + CONTENT)
@@ -269,7 +336,6 @@ void the_mime_parser(char *partnum,
 {
 
        char *ptr;
-       char *srch = NULL;
        char *part_start, *part_end = NULL;
        char buf[SIZ];
        char *header;
@@ -435,13 +501,13 @@ void the_mime_parser(char *partnum,
 
                part_start = NULL;
                do {
-                       next_boundary = NULL;
-                       for (srch=ptr; srch<content_end; ++srch) {
-                               if (!memcmp(srch, startary, startary_len)) {
-                                       next_boundary = srch;
-                                       srch = content_end;
-                               }
-                       }
+                       char tmp;
+
+                       tmp = *content_end;
+                       *content_end = '\0';
+                       
+                       next_boundary = strstr(ptr, startary);
+                       *content_end = tmp;
 
                        if ( (part_start != NULL) && (next_boundary != NULL) ) {
                                part_end = next_boundary;
@@ -483,7 +549,9 @@ void the_mime_parser(char *partnum,
                                        /* Determine whether newlines are LF or CRLF */
                                        evaluate_crlf_ptr = part_start;
                                        --evaluate_crlf_ptr;
-                                       if (!memcmp(evaluate_crlf_ptr, "\r\n", 2)) {
+                                       if ((*evaluate_crlf_ptr == '\r') && 
+                                           (*(evaluate_crlf_ptr + 1) == '\n'))
+                                       {
                                                crlf_in_use = 1;
                                        }
                                        else {