Home » Mail » Mail » Bug #2931

[2004-12-09 02:19 UTC] john at curioussymbols dot com

Description:
------------
Sample code below.

The address string

$address_string = '"Joe Doe \(from Somewhere\)" <doe@example.com>, postmaster@example.com, root';

doesn't parse correctly, and the following script gives 'Something is wrong'

Script is taken from PHP manual for imap RFC822 parse function, which works correctly with the same code; I have modified it to use the PEAR equivalent.

I've just downloaded the latest version of PEAR and Mail and tested with those.

Reproduce code:
---------------
<?php
require_once("Mail/RFC822.php");
$address_string = '"Joe Doe \(from Somewhere\)" <doe@example.com>, postmaster@example.com, root';
//$address_string = "Joe Doe from Somewhere <doe@example.com>, postmaster@example.com, root";
echo($address_string."\n");
$address_array = Mail_RFC822::parseAddressList($address_string, "example.com");

if (!is_array($address_array) || count($address_array) < 1) {
die("something is wrong\n");
}
foreach ($address_array as $val) {
echo "mailbox : " . $val->mailbox . "<br />\n";
echo "host : " . $val->host . "<br />\n";
echo "personal: " . $val->personal . "<br />\n";
}
print_r($address_array);
?>

Expected result:
----------------
Something like this (tested it on another server which had IMAP extensions):

X-Powered-By: PHP/4.0.6
Content-type: text/html

mailbox : doe<br />
host : example.com<br />
personal: Joe Doe (from Somewhere)<br />
adl : <br />
mailbox : postmaster<br />
host : example.com<br />
personal: <br />
adl : <br />
mailbox : root<br />
host : example.com<br />
personal: <br />
adl : <br />

Actual result:
--------------
john@cruncher2 ~
$ php test.php
"Joe Doe \(from Somewhere\)" <doe@example.com>, postmaster@example.com, root
something is wrong

[2005-02-07 16:13 UTC] john at curioussymbols dot com

I wouldn't put the quotes there myself, but I'm using this function to parse incoming emails, and the incoming emails sometimes have these. It's an Outlook or Outlook Express thing, I believe.

I think the intention of the backslashes is to avoid the 'from somewhere' being interpreted as a 'comment' (see RFC 2822, section 3.2.3,

http://www.faqs.org/rfcs/rfc2822.html

I think that basically you're supposed to have quotes around a name if it has more than one word. So,

John <john@doe.com>
"Doe, John" <john@doe.com>

I'm not sure if the brackets in the following count as a comment or not. I think maybe they do, and that would be why MS escapes them in my original example.

"Doe (John)" <john@doe.com>

Note section 4.4 and A.6.1 in RFC 2822. These say that you can also (obsoletely) have

John Doe <john@doe.com>

That probably confuses things - but I reckon it probably still happens a bit sometimes from dodgy webmail apps etc.

JP

[2005-02-07 16:15 UTC] john at curioussymbols dot com

Change that:
"I wouldn't put the quotes there myself"
to
"I wouldn't put the backslashes there myself"

but even then, I'm not so sure. I can't see anything in RFC2822 that says you're allowed to have a comment inside a quoted-string.

[2005-05-12 21:29 UTC] mic at uts dot cc dot utexas dot edu

We are seeing the same problem at our IMP installation. We have changed the test to

return !preg_match('/[\x0D\\\\"]/', preg_replace('/\\\\./', '', $qstring));

This is a two-stage process expressed as a one liner:

1) account for valid escaped characters (by the expedient of removing them from consideration using preg_replace())
2) fail on any remaining illegal (i.e. unescaped) characters, i.e. carriage returns, backslashes, quotes.

This change has worked so far with our test scripts and with IMP. I hope it is comprehensive enough to represent a worthwhile fix for this bug.