diff options
author | Nick Mathewson <nickm@torproject.org> | 2005-03-28 01:07:14 +0000 |
---|---|---|
committer | Nick Mathewson <nickm@torproject.org> | 2005-03-28 01:07:14 +0000 |
commit | 4246f6f8bcb0d4c9f134b1fba011c4ca3e1f17ef (patch) | |
tree | dd456ed228515459d7b58818cc9cdb9f12007474 | |
parent | 96c331e4ed0608357fed5830a4d9651aba374c26 (diff) | |
download | tor-4246f6f8bcb0d4c9f134b1fba011c4ca3e1f17ef.tar.gz tor-4246f6f8bcb0d4c9f134b1fba011c4ca3e1f17ef.zip |
Clarify that we don't, and why we don't, need to redefine FD_SETSIZE.
svn:r3904
-rw-r--r-- | src/or/or.h | 17 |
1 files changed, 8 insertions, 9 deletions
diff --git a/src/or/or.h b/src/or/or.h index 5ceeffff4c..6f74a4e3ea 100644 --- a/src/or/or.h +++ b/src/or/or.h @@ -102,15 +102,14 @@ #define MAXCONNECTIONS 15000 #ifdef MS_WINDOWS -/* This trick makes winsock resize fd_set, which defaults to the - * insanely low 64. */ -#define FD_SETSIZE MAXCONNECTIONS -/* XXXX But Windows FD_SET and FD_CLR are tremendously ugly, and linear in - * the total number of sockets set! Perhaps we should eventually use - * WSAEventSelect and WSAWaitForMultipleEvents instead of select? -NM - * I'm told these funcs have an unchangeable 64 conn limit on 95/98, - * so maybe not. -RD */ - +/* No, we don't need to redefine FD_SETSIZE before including winsock: + * we use libevent now, and libevent handles the select() stuff. Yes, + * some documents imply that we need to redefine anyway if we're using + * select() anywhere in our application or in anything it links to: these + * documents are either the holy texts of a cargo cult of network + * programmers, or more likely a simplification of what's going on for + * people who haven't read winsock[2].c for themselves. + */ #if (_MSC_VER <= 1300) #include <winsock.h> #else |