Changeset 570

Timestamp:

Jul 2, 2006 9:29:09 AM (18 years ago)

Author:

bennylp

Message:

Fixed bugs in scanner: (1) pj_cis_match() takes int argument, so when ASCII character above 127 is given, it will access the array with negative index, and (2) pj_scan_get_newline() may incorrectly eat two newlines when the second newline is a header continuation

Location:

pjproject/trunk/pjlib-util

Files:

• build (modified) (1 prop)
• include/pjlib-util/scanner.h (modified) (1 diff)
• src/pjlib-util/scanner.c (modified) (1 diff)

pjproject/trunk/pjlib-util/build

@@ -7 +7 @@
 *vcproj*
 *.sln
+*.plg

@@ -7 +7 @@
 *vcproj*
 *.sln
+*.plg

pjproject/trunk/pjlib-util/include/pjlib-util/scanner.h

@@ -157 +157 @@
  * @return          Non-zero if match (not necessarily one).
  */
-PJ_INLINE(int) pj_cis_match( const pj_cis_t *cis, int c )
+PJ_INLINE(int) pj_cis_match( const pj_cis_t *cis, pj_uint8_t c )
 {
     return PJ_CIS_ISSET(cis, c);

pjproject/trunk/pjlib-util/src/pjlib-util/scanner.c

@@ -402 +402 @@
     scanner->start_line = scanner->curptr;
 
-    if (PJ_SCAN_IS_PROBABLY_SPACE(*scanner->curptr) && scanner->skip_ws) {
-        pj_scan_skip_whitespace(scanner);
-    }
+    /**
+     * This probably is a bug, see PROTOS test #2480.
+     * This would cause scanner to incorrectly eat two new lines, e.g.
+     * when parsing:
+     *   
+     *  Content-Length: 120\r\n
+     *  \r\n
+     *  <space><space><space>...
+     *
+     * When pj_scan_get_newline() is called to parse the first newline
+     * in the Content-Length header, it will eat the second newline
+     * too because it thinks that it's a header continuation.
+     *
+     * if (PJ_SCAN_IS_PROBABLY_SPACE(*scanner->curptr) && scanner->skip_ws) {
+     *    pj_scan_skip_whitespace(scanner);
+     * }
+     */
 }
 

@@ -7 +7 @@
 *vcproj*
 *.sln
+*.plg