Changeset 620

Timestamp:

07/05/08 01:15:08 (2 months ago)

Author:

markdoliner

Message:

Fix a potential memory leak when the remote host closes a TCP
connection. This may not actually be a problem in practice. If it
was it seems like we would have noticed it in valgrind before now
and fixed it already.

I'm also adding a note about my change from revision 619 to the
ChangeLog?, and changing the documentation in mio.h a little to hopefully
make it more clear.

Location:

trunk

Files:

• ChangeLog (modified) (1 diff)
• mio/mio.h (modified) (1 diff)
• sx/io.c (modified) (1 diff)

trunk/ChangeLog

@@ +1 @@
+2008-07-04 Mark Doliner <mark@meebo.com>
+        * Fix a potential memory leak when attempting to read from a socket that
+          has been closed.
+        * Improve the documentation for <max_fds> in the configuration files.
+
 2008-05-20 Tomasz Sterna <tomek@xiaoka.com>
         * Ported apr_base64.c from Apache httpd 2.2.8

trunk/mio/mio.h

@@ -80 +80 @@
  * @brief mio - manage i/o
  * 
- * This used to be something large and all inclusive for 1.2/1.4,
- * but for 1.5 and beyond it is the most simple fd wrapper possible.
- * It is also customized per-app and may be limited/extended depending on needs.
+ * This is the most simple fd wrapper possible. It is also customized
+ * per-app and may be limited/extended depending on needs.
+ *
+ * It's basically our own implementation of libevent or libev.
  * 
  * Usage is pretty simple:

trunk/sx/io.c

@@ -205 +205 @@
 
     /* EOF if we got a 0-byte read from the socket */
-    if(read == 0)
+    if(read == 0) {
         /* they went away */
+        _sx_buffer_free(in);
         _sx_state(s, state_CLOSING);
     
-    else {
+    } else {
         _sx_debug(ZONE, "passed %d read bytes", in->len);