Message-Id: <19990113163756.14770.qmail@hyperreal.org>
Date: 13 Jan 1999 16:37:56 -0000
From: Phil Dietz <pedietz@west.com>
Reply-To: pedietz@west.com
To: apbugs@hyperreal.org
Subject: apache-api/3657: apache returning status 304 on images that are
 updated
Sender: apache-bugdb-owner@apache.org
Precedence: bulk


>Number:         3657
>Category:       apache-api
>Synopsis:       apache returning status 304 on images that are updated
>Confidential:   no
>Severity:       serious
>Priority:       medium
>Responsible:    apache
>State:          open
>Class:          sw-bug
>Submitter-Id:   apache
>Arrival-Date:   Wed Jan 13 08:40:01 PST 1999
>Last-Modified:
>Originator:     pedietz@west.com
>Organization:
apache
>Release:        1.3.3 1.3.4
>Environment:
AIX 4.3.2
both cc and gcc
browsers IE4.1, IE5.0 beta, (Netscape4.5 generally works right)

>Description:
A CGI script updates an image and displays it every so often.
He ported his script from Netscape server to apache 1.3.3 yesterday.
His graphics quit updating.  The graphic displayed is the initial graphic when he first opened his browser.  None of the updates come across.

After snooping with a proxy and later apache source code, it appears the ap_find_token() command is not working correctly.

Internet Explorer 4.1 and 5.0beta browser sends:
                If-None-Match: W/"2f30-f20-369bbd06"
apache returns  Etag: W/"2f30-11b3-369bd28d"

The two are completely different which jives since the image was updated.

Inside of ap_find_token() (which was called by ap_meets_conditions() in the If-None-Match" section), I put in some debugging code:

/* find http tokens, see the definition of token from RFC2068 */
API_EXPORT(int) ap_find_token(pool *p, const char *line, const char *tok)
{
    const unsigned char *start_token;
    const unsigned char *s;

    if (!line)
        return 0;

    s = (const unsigned char *)line;
    for (;;) {
        /* find start of token, skip all stop characters, note NUL
         * isn't a token stop, so we don't need to test for it
         */
        while (TEST_CHAR(*s, T_HTTP_TOKEN_STOP)) {
            ++s;
        }
        if (!*s) {
            return 0;
        }
        start_token = s;
        /* find end of the token */
        while (*s && !TEST_CHAR(*s, T_HTTP_TOKEN_STOP)) {
            ++s;
        }

        DEBUG(123456, "debug: %s, %s, %d\n", start_token, tok, s - start_token);

        if (!strncasecmp((const char *)start_token, (const char *)tok, s - start_token)) {
            return 1;
        }
        if (!*s) {
            return 0;
        }
    }
}

A result is:
debug: W/"2f2d-1299-369cbbd9", W/"2f2d-12a5-369cbf5e", 1

Notice the strncasecmp length of 1.  That means only the leading W's are being compared, which do match.

Shouldn't the entire string be compared ?

Because the leading W's match, ap_meets_conditions() is sending back a HTTP_NOT_MODIFIED status for my images -- when they have updated.

Any help ?
>How-To-Repeat:
I tested this on both apache1.3.3 and apache1.3.4.

>Fix:
Should the entire ETag: be compared to If-None-Match: or just the leading char ?
>Audit-Trail:
>Unformatted:
[In order for any reply to be added to the PR database, ]
[you need to include <apbugs@Apache.Org> in the Cc line ]
[and leave the subject line UNCHANGED.  This is not done]
[automatically because of the potential for mail loops. ]
[If you do not include this Cc, your reply may be ig-   ]
[nored unless you are responding to an explicit request ]
[from a developer.                                      ]
[Reply only with text; DO NOT SEND ATTACHMENTS!         ]