Lighttpd

Here is a clearer example:

Date set to exactly 1 hour before 'last-modified' date:

curl -I -s -H "Host: example.com" -H "If-Modified-Since: Wed, 04 Aug 2009 23:47:19 GMT" http://example.com/example.css
HTTP/1.1 304 Not Modified
Content-Type: text/css
Accept-Ranges: bytes
ETag: "1397551174"
Last-Modified: Wed, 05 Aug 2009 00:47:19 GMT
Date: Wed, 05 Aug 2009 00:47:52 GMT
Server: lighttpd

Date set to 1 hour and 1 second before last-modified date:

curl -I -s -H "Host: example.com" -H "If-Modified-Since: Wed, 04 Aug 2009 23:47:18 GMT" http://example.com/example.css
HTTP/1.1 200 OK
Content-Type: text/css
Accept-Ranges: bytes
ETag: "1397551174"
Last-Modified: Wed, 05 Aug 2009 00:47:19 GMT
Content-Length: 25
Date: Wed, 05 Aug 2009 00:47:58 GMT
Server: lighttpd

I think your strptime overrides tm_isdst (or mktime is broken); you could try setting it to 0 again after strptime (src/http-header-glue.c line 334), before mktime.

I added `tm.tm_isdst = 0;' just before t_header = mktime(&tm);, recompiled and same issue as before.

mktime changes tm_isdst; that behaviour isn't documented in the manpage, and wrong imho:
(posix: http://www.opengroup.org/onlinepubs/009695399/functions/mktime.html)

Upon successful completion, the values of the tm_wday and tm_yday components of the structure shall be set appropriately, and the other components are set to represent the specified time since the Epoch, but with their values forced to the ranges indicated in the <time.h> entry; the final value of tm_mday shall not be set until tm_mon and tm_year are determined.

As the tm_isdst field is "within the range" (no range specified, but all values are "valid" as their interpretation is specified), i don't see a reason to change it.

In eglibc (used in debian), strptime sets tm_isdst to zero, so you don't see the problem there.

The solution is to add tm.tm_isdst = 0 before every call to mktime.

That solution does indeed work. Thanks!