While scanning the [email protected] mailing list, I've
learned time(2) may return the wrong value in the first 1 to 2.5
ms of every second.  While I'm not sure if the Date: response
header matters to anyone, returning the correct time seems
prudent.

Link: https://lore.kernel.org/git/[email protected]/
Link: 
https://inbox.sourceware.org/libc-alpha/[email protected]/T/
Link: https://sourceware.org/bugzilla/show_bug.cgi?id=30200
---
 ext/unicorn_http/httpdate.c | 21 ++++++++++++++++-----
 1 file changed, 16 insertions(+), 5 deletions(-)

diff --git a/ext/unicorn_http/httpdate.c b/ext/unicorn_http/httpdate.c
index 3f512dd..27a8f51 100644
--- a/ext/unicorn_http/httpdate.c
+++ b/ext/unicorn_http/httpdate.c
@@ -1,5 +1,5 @@
 #include <ruby.h>
-#include <time.h>
+#include <sys/time.h>
 #include <stdio.h>
 
 static const size_t buf_capa = sizeof("Thu, 01 Jan 1970 00:00:00 GMT");
@@ -43,13 +43,24 @@ static struct tm * my_gmtime_r(time_t *now, struct tm *tm)
 static VALUE httpdate(VALUE self)
 {
        static time_t last;
-       time_t now = time(NULL); /* not a syscall on modern 64-bit systems */
+       struct timeval now;
        struct tm tm;
 
-       if (last == now)
+       /*
+        * Favor gettimeofday(2) over time(2), as the latter can return the
+        * wrong value in the first 1 .. 2.5 ms of every second(!)
+        *
+        * 
https://lore.kernel.org/git/[email protected]/
+        * 
https://inbox.sourceware.org/libc-alpha/[email protected]/T/
+        * https://sourceware.org/bugzilla/show_bug.cgi?id=30200
+        */
+       if (gettimeofday(&now, NULL))
+               rb_sys_fail("gettimeofday");
+
+       if (last == now.tv_sec)
                return buf;
-       last = now;
-       gmtime_r(&now, &tm);
+       last = now.tv_sec;
+       gmtime_r(&now.tv_sec, &tm);
 
        /* we can make this thread-safe later if our Ruby loses the GVL */
        snprintf(buf_ptr, buf_capa,

Reply via email to