Refactoring, stable&functional intermediate state

1 files changed, 56 insertions, 24 deletions

diff --git a/http_parser.c b/http_parser.c
index 05cd094..30e74c8 100644
--- a/http_parser.c
+++ b/http_parser.c
@@ -37,7 +37,47 @@
 
 #include "main.h"
 
+static inline char * handle_colon(char ** start, char ** end, char ** search) {
+        // HTTP headers and content are separated by an empty line (which is two
+        // newlines), but before that there are pairs of names and values, separated
+        // by a colon, e.g. Content-Type: text/html and this function is called, when
+        // *end reaches a colon.
+
+        *end[0] = '\0'; // remember header 'names' and search for the value
+        (*end)++;       // jump over the colon
+
+        if (0 == strcasecmp("Content-Type", *start)) {
+                *search = "\r\n;"; // (more unlikely) also search for a semicolon in Content-Type: [...]; boundary=[...]
+        } else {
+                *search = "\r\n";  // (likely) search for some kind of newline
+        }
+        return *start;  // remember, where name starts, will be important in the newline case
+}
+
+static inline void handle_semicolon(char ** start, char ** end, char **  search) {
+        // find the form-data boundary in the main header
+        *start += strspn(*start, "; "); // remove spaces and semicolons (boundary check implicit; also stops at '\0')
+
+        const char s_multipart_form_data[] = "boundary=";
+        if(0 < strcasecmp(*start, s_multipart_form_data))
+        {
+                *end = *end + sizeof(s_multipart_form_data) + 1;
+                *end += strspn(*end, "-");
+                DEBUG("> Boundary found, now looking where it ends...\n");
+        }
+
+        *search = "\r\n"; // do not search further semicolons
+}
+
 void parse_http(size_t new_socket, char * request, size_t request_length) {
+        // this http parser modifies the input buffer and replaces single characters
+        // with \0-chars to terminate them, but it does not copy strings, because
+        // that would require tedious length checks and would also allow the user
+        // to submit forms with unreasonable long strings in strange places or to
+        // make it shorter: it would make boundary checks necessary for every single
+        // heaer name, value, sub-value and content. This implementation is more
+        // generic and only requires content length bounary checks.
+
         char * start = request;
         char * end = NULL;
         char * search = "\r\n";
@@ -50,30 +90,12 @@ void parse_http(size_t new_socket, char * request, size_t request_length) {
                 size_t matchlen = strspn(end, search);
                 switch(end[0]) {
                         case ':':
-                                end[0] = '\0'; // {{{ remember header 'names' and search for the value
-                                end++;         // jump over the colon
-
-                                name = start;  // remember, where name starts, will be important in the newline case
-
-                                if (0 == strcasecmp("Content-Type", start)) {
-                                        search = "\r\n;"; // (more unlikely) also search for a semicolon in Content-Type: [...]; boundary=[...]
-                                } else {
-                                        search = "\r\n";  // (likely) search for some kind of newline
-                                } // }}}
-                                break;
+                                handle_colon(&start, &end, &search);
+                                name = start;
+                        break;
                         case ';':
-                                // {{{ find the form-data boundary in the main header
-                                start += strspn(start, "; "); // remove spaces and semicolons (boundary check implicit; also stops at '\0')
-
-                                const char s_multipart_form_data[] = "boundary=";
-                                if(NULL == http_header.boundary && 0 < strcasecmp(start, s_multipart_form_data))
-                                {
-                                        http_header.boundary = end + sizeof(s_multipart_form_data) + 1;
-                                        http_header.boundary += strspn(http_header.boundary, "-");
-                                        DEBUG("> Boundary found, now looking where it ends...\n");
-                                        search = "\r\n";
-                                        continue;
-                                } /// }}}
+                                handle_semicolon(&start, &end, &search);
+                                http_header.boundary = end;
                                 break;
                         case '\r': // fallthrough
                         case '\n':
@@ -165,8 +187,18 @@ void parse_http(size_t new_socket, char * request, size_t request_length) {
                         start = end + matchlen;
         }
 
+        // failed to find URL? Use a default to avoid NULL pointer exceptions later
+        if(!http_header.url || http_header.url[0] == '\0' || http_header.url[1] == '\0') {
+                http_header.url = "/index.html";
+                DEBUG("Warning: Request had no URL and is probably invalid: %d", http_header.url[0]);
+        }
+
         DEBUG("> sending answer...\n");
-        send_answer(&http_header, new_socket);
+        FILE * f = fdopen((size_t) new_socket, "w");
+        send_answer(&http_header, f);
+
+        fflush(f);
+        fclose(f);
         DEBUG("> answer sent.\n");
 }