httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Stefan Fritsch ...@sfritsch.de>
Subject Re: Is this a test framework bug?
Date Mon, 18 Apr 2011 18:46:15 GMT
On Monday 18 April 2011, Torsten Förtsch wrote:
> On Monday, April 18, 2011 10:36:13 Joe Orton wrote:
> > If you change the CGI script to send a 100 rather than 102, does
> > it work?  LWP should treat all 1xx as interim responses so I'd
> > say it is an LWP bug.
> 
> It is certainly triggered by the LWP version upgrade. I also agree
> that it's a bug in LWP. But my question was also what do we test
> here? If the purpose of the test is to see if the 102 response is
> passed through (as suggested by the comment) then I'd think the
> test is wrong too.

IMHO the purpose is to test if the real (200) response is passed 
through.

For me, a tcpdump gives:
================
GET /reverse/modules/cgi/nph-102.pl HTTP/1.0
Host: localhost:8541
User-Agent: libwww-perl/6.01

HTTP/1.1 200 OK
Date: Mon, 18 Apr 2011 18:33:25 GMT
Server: Apache/2.3.12-dev (Unix) mod_ssl/2.3.12-dev OpenSSL/1.0.0d 
DAV/2
Content-Type: text/plain
Connection: close

this is nph-stdout
================

Note that the test requests HTTP/1.0. RFC 2616 says "servers MUST NOT 
send a 1xx response to an HTTP/1.0 client", so mod_proxy is correct 
not to forward the 102 response to the client. Maybe the newer LWP 
sends an HTTP/1.1 request? Can you confirm this with tcpdump? Which 
version of LWP do you use?


For HTTP/1.1 clients, RFC 2616 says " A client MUST be prepared to 
accept one or more 1xx status responses  prior to a regular response, 
even if the client does not expect a 100 (Continue) status message. 
Unexpected 1xx status responses MAY be ignored by a user agent." I am 
not sure what is meant with the last sentence, but the first sentence 
seems to indicate that LWP returning status 102 is a bug.

Cheers,
Stefan

Mime
View raw message