osdir.com


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[Bug 62380] Existing header not overwritten when using the 'always' condition with Header set


https://bz.apache.org/bugzilla/show_bug.cgi?id=62380

--- Comment #8 from Luca Toscano <toscano.luca@xxxxxxxxx> ---
Changing mod_headers might not be as easy as I thought. The following simple
patch solves this specific use case:

Index: modules/metadata/mod_headers.c
===================================================================
--- modules/metadata/mod_headers.c      (revision 1843742)
+++ modules/metadata/mod_headers.c      (working copy)
@@ -791,9 +791,12 @@
             }
             break;
         case hdr_set:
-            if (r->headers_in != headers &&
-                !ap_cstr_casecmp(hdr->header, "Content-Type")) {
-                 ap_set_content_type(r, process_tags(hdr, r));
+            if (r->headers_in != headers) {
+                    if (!ap_cstr_casecmp(hdr->header, "Content-Type")) {
+                        ap_set_content_type(r, process_tags(hdr, r));
+                    } else if (r->err_headers_out == headers) {
+                        apr_table_unset(r->headers_out, hdr->header);
+                    }
             }
             apr_table_setn(headers, hdr->header, process_tags(hdr, r));
             break;

But then how about 'merge', 'setifempty', etc..? The fact that there are two
separate list of headers ('onsuccess' vs 'always') makes it really tricky to
fix all the use cases without introducing regressions.

For example, how is it the right way to handle the following example?

'onsuccess' header list -> X-Foo: bla
'always' header list -> []

Header always merge X-Foo anotherone

In this case the current behavior, i.e. duplication, seems the most "logic" one
given what written above, other solutions would be inconsistent (like moving
X-Foo: bla to the always header list before applying the merge). 

The same example with 'setifempty' is also strange, since not setting X-Foo in
the 'always' list would be incorrect, so duplication seems legitimate.

As far as I can see what it is missing from the documentation is that, to be
completely sure about avoiding header duplication, one should unset an header
in the 'onsuccess' list before doing anything with the 'always' one, like:

Header (onsuccess) unset X-Foo
Header always set X-Foo somevalue

Using the above the mod_proxy_http example works as expected, not duplicating
any header.

Any opinion from other people reading? :)

-- 
You are receiving this mail because:
You are the assignee for the bug.
---------------------------------------------------------------------
To unsubscribe, e-mail: bugs-unsubscribe@xxxxxxxxxxxxxxxx
For additional commands, e-mail: bugs-help@xxxxxxxxxxxxxxxx