| Subject: | $request->uri is wrong for paths that start with // |
If the Request-URI starts with two slashes (e.g. GET //foo/bar), the returned request's uri object
is misconfigured (thinking that this is a scheme-relative uri) and the first segment after the
double slash (foo in this case) is interpreted as the host. The attached patch fixes the issue.
| Subject: | abs_path_slashes.patch |
diff -ur old/Parser.pm new/Parser.pm
--- old/Parser.pm 2010-06-15 13:57:43.000000000 -0500
+++ new/Parser.pm 2011-03-01 13:53:02.000000000 -0600
@@ -260,6 +260,12 @@
unless $http and $http =~ /^HTTP\/(\d+)\.(\d+)$/i;
($major,$minor) = ($1,$2);
die 'HTTP requests not allowed' unless $self->{request};
+
+ # If the Request-URI is an abs_path, we need to tell URI that we don't
+ # know the scheme, otherwise it will misinterpret paths that start with
+ # // as being scheme-relative uris, and will interpret the first
+ # component after // as the host (see rfc 2616)
+ $uri = "//$uri" if $uri =~ m(^/);
$obj = $self->{obj} = HTTP::Request->new($method, URI->new($uri));
}
diff -ur old/t/1.t new/t/1.t
--- old/t/1.t 2007-02-24 08:54:48.000000000 -0600
+++ new/t/1.t 2011-03-01 13:34:34.000000000 -0600
@@ -4,7 +4,7 @@
#########################
use strict;
-use Test::More tests => 21;
+use Test::More tests => 22;
# <1>
BEGIN { use_ok('HTTP::Parser') };
@@ -77,3 +77,7 @@
is($res->content, "Some content!\x0d\x0a", 'content is correct');
}
+# <1>
+$parser = HTTP::Parser->new(request => 1);
+$parser->add("GET //foo///bar/baz HTTP/1.1\x0d\x0a\x0d\x0a");
+is $parser->request->uri->path, '//foo///bar/baz';