Mike Crowe sends this patch:

Change various uses of time(NULL) to new MHD_monotonic_time() function to
make timeouts immune to the system real time clock changing. Leave one call
to time(2) which continues to use the real time for the HTTP Date: header.

If clock_gettime does not support CLOCK_MONOTONIC then fall through to
calling time(NULL) instead. This will be problematic if clock_gettime
sometimes fails and sometimes succeeds but that shouldn't happen.

The autoconf magic was lifted from
http://lists.gnu.org/archive/html/autoconf/2010-08/msg00035.html

1 files changed, 3 insertions, 0 deletions

diff --git a/configure.ac b/configure.ac
index 41d54722..7b472784 100644
--- a/configure.ac
+++ b/configure.ac
@@ -190,6 +190,9 @@ AC_CHECK_HEADERS([plibc.h],our_private_plibc_h=0,our_private_plibc_h=1)
 AM_CONDITIONAL(USE_PRIVATE_PLIBC_H, test x$our_private_plibc_h = x1)    
 
 AC_CHECK_FUNCS(memmem)
+AC_SEARCH_LIBS([clock_gettime], [rt], [
+                                AC_DEFINE(HAVE_CLOCK_GETTIME, 1, [Have clock_gettime])
+])
 
 # IPv6
 AC_MSG_CHECKING(for IPv6)