| Subject: | Request to remove t/live/jigsaw-chunk.t |
| Date: | Thu, 17 Jul 2008 15:05:20 +0100 |
| To: | bug-libwww-perl [...] rt.cpan.org |
| From: | Andrew Benham <andrew.benham [...] thus.net> |
All (recent ?) versions of libwww-perl include the t/live/jigsaw-chunk.t
test script. This relies on an external http server providing
http://jigsaw.w3.org/HTTP/ChunkedScript
Unfortunately the external server occasionally serves incorrect content,
and the test fails.
I've had a response from the Jigsaw people to my bug report, so I
think the solution is probably to delete the test from libwww-perl.
Show quoted text
-------- Original Message --------
Subject: Occasional bug in http://jigsaw.w3.org/HTTP/ChunkedScript
Date: Wed, 28 May 2008 15:50:14 +0100
From: Andrew Benham <andrew.benham@thus.net>
To: jigsaw@w3.org
Sirs:
Our package build system occasionally suffers test failures when
building the Perl module libwww-perl
(http://search.cpan.org/~gaas/libwww-perl-5.812/)
This Perl code have a test case (attached) which essentially retrieves
the URL http://jigsaw.w3.org/HTTP/ChunkedScript and checks that all
the lines (apart from your first 3 body lines) are the same. This is
test 4 in the attached file.
Now occasionally test 4 fails - we're talking maybe 1 time in 100
invocations. This isn't the end of the world, but nevertheless
it's annoying.
So I repeatedly ran the test case whilst running Wireshark to monitor
the traffic to your web server. And lo, when the test fails Wireshark
shows that the response doesn't contain identical lines :-(
The Wireshark trace file is attached ('bad_response'). If you follow
the TCP stream then you'll see that about three-quarters of the way
through the file your server has sent:
01234567890123456789012345678901234567890123456789012345678901234567890
01234567890123456789012345678901234567890123456789012345678901234567890
01234567890123456789012345678901234567890123456789012345678901234567890
01234567890123456789012345678901234567890123456789012345678901234589012345678901234567890123456789012345678901234567890
01234567890123456789012345678901234567890123456789012345678901234567890
01234567890123456789012345678901234567890123456789012345678901234567890
01234567890123456789012345678901234567890123456789012345678901234567890
01234567890123456767890
01234567890123456789012345678901234567890123456789012345678901234567890
01234567890123456789012345678901234567890123456789012345678901234567890
01234567890123456789012345678901234567890123456789012345678901234567890
It looks like there's a '67' pair of chars which has been missed out of the
long line at around column 67, but this pair of chars appear in the short line
instead. The upshot is that the chunk size hasn't been altered, which
is fortuitous.
So I suspect there's some corner case in your code which can generate
an incorrect response some of the time. It would be good if you could fix
it please - despite your disclaimers about protocol conformance :-)
I'm prepared to help in any way I can.
--
Andrew Benham Demon@Thus andrew.benham@thus.net
Finchley, London N3 2QQ, U.K. Tel: 020 8495 6343 Fax: 020 8495 6037
print "1..5\n";
use strict;
use LWP::UserAgent;
my $ua = LWP::UserAgent->new(keep_alive => 1);
my $req = HTTP::Request->new(GET => "http://jigsaw.w3.org/HTTP/ChunkedScript");
my $res = $ua->request($req);
print "not " unless $res->is_success && $res->content_type eq "text/plain";
print "ok 1\n";
print "not " unless $res->header("Client-Transfer-Encoding") eq "chunked";
print "ok 2\n";
for (${$res->content_ref}) {
s/\015?\012/\n/g;
/Below this line, is 1000 repeated lines of 0-9/ || die;
s/^.*?-----+\n//s;
my @lines = split(/^/);
print "not " if @lines != 1000;
print "ok 3\n";
# check that all lines are the same
my $first = shift(@lines);
my $no_they_are_not;
for (@lines) {
$no_they_are_not++ if ($_ ne $first);
}
print "not " if $no_they_are_not;
print "ok 4\n";
print "not " unless $first =~ /^\d+$/;
print "ok 5\n";
}
Message body not shown because it is not plain text.