Greg Ames wrote:
>
>
> core dump #12 is another newer one.
...and so are core dump #14 and 15. Both have signs of header
truncation (less headers in r->headers_in than the apparent input
buffer).
In #14, there's 4 Accept: headers in what looks like the input buffer,
containing huge numbers of acceptable mime types. Input buffer Accept
header #4 looks fubar toward the end. The Accept: header in
r->header_in does not have all of the same mime headers you see in what
I'm guessing was the input buffer.
#15 is simpler. r->headers_in is empty, but the request says it's
HTTP/1.0 . bzzzzt - wrong. Searching the core dump with vi for the
URI, I find:
GET /jetspeed/site/images/logo.gif HTTP/1.0^M
Host: jakarta.apache.org^M
ache.org/^M
Accept-Language: en-us^M
Accept-Encoding: gzip, deflate^M
User-Agent: Mozilla/4.0 (compatible; MSIE 5.5; Windows 98; Win 9x
4.90)^M
Host: www.apache.org^M
Connection: Keep-Alive^M
^M
The first Host: header looks right - that file does exist on
jakarta.apache.org. After that, who knows. The second Host: header is
probably residual.
Jeff suggested earlier that we stash away the input buffer somewhere if
we're built in debug mode. I think that's a great idea.
Greg
