https://bz.apache.org/bugzilla/show_bug.cgi?id=64775
--- Comment #2 from Sam Horrocks <s...@daemoninc.com> ---
The change in this area was the change made in "Bug 55853 - mod_jk doesn't set
the content length correctly for apache 2.x" which was in r1549202 - that
change was released in 1.2.39 which is when our app broke.
Here's the change made in r1549202 which broke it for us:
--- tomcat/jk/trunk/native/apache-2.0/mod_jk.c 2013/06/26 20:11:12 1497068
+++ tomcat/jk/trunk/native/apache-2.0/mod_jk.c 2013/12/08 20:11:13 1549202
@@ -392,7 +392,7 @@ static int JK_METHOD ws_start_response(j
apr_table_set(r->headers_out, header_names[h], header_values[h]);
}
else if (!strcasecmp(header_names[h], "Content-Length")) {
- apr_table_set(r->headers_out, header_names[h], header_values[h]);
+ ap_set_content_length(r, apr_atoi64(header_values[h]));
}
else if (!strcasecmp(header_names[h], "Transfer-Encoding")) {
apr_table_set(r->headers_out, header_names[h], header_values[h]);
Prior to this change, when header_values[h] was empty, it was passed along as a
string to apr_table_set, which produced a Content-Length header with an empty
string like this:
Content-Length:
After this change, the emptry string was converted to a number (the apr_atoi64
call does that), and then the number is passed to ap_set_content_length, which
then causes a header to be produced like this:
Content-Length: 0
That's when it broke for us - the empty string was being ignored, but the patch
in r1549202 turned it into a definite zero-length even though the
content-length was actually not set (since header_values[h] was empty). The
patch I sent will not set the content-length header at all unless
header_values[h] contains a non-empty string.
--
You are receiving this mail because:
You are the assignee for the bug.
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscr...@tomcat.apache.org
For additional commands, e-mail: dev-h...@tomcat.apache.org
