"No valid recipients" with address containing UTF-8 chars in email domain and no display name

[Player701]

It seems that if the recipient list contains an address that has UTF-8 characters in the domain part (after "@") and no display name, Roundcube silently ignores it. This results in a "no valid recipients" error when there are no other addresses in the list.

Example contents of the "To" field in the message form and behavior:

• "test@тест.рф": errors out with "no valid recipients"
• "Тест <test@тест.рф>": OK and message gets sent

Note that the problem is with Roundcube, not the SMTP server. Enabling SMTP debugging reveals that in the first case Roundcube does not send any RCPT TO commands, which (correctly) results in the "no valid recipients" error being issued by the SMTP server.

Tested with Roundcube 1.5.2 and PHP 8.0.17.

[alecpl]

I do not support Russian here.

[Player701]

I am not going to argue politics with you, but this is still a bug, and it is not solely restricted to anything "Russian" (regardless of what exactly you may have meant by that).

This bug affects all addresses where the TLD part ends in a non-Latin letter. There are many such TLDs, and most of them don't have anything to do with Russia. Please see here for the complete list of IDN TLDs.

I've managed to trace the exact cause down to the fix for issue #7899, namely, this one part:

roundcubemail/program/include/rcmail_sendmail.php

Lines 762 to 768 in ab91416

	// address without brackets and without name (add brackets)
	else if (preg_match('/^'.$email_regexp.'$/', $item)) {
	// Remove trailing non-letter characters (#7899)
	$item = preg_replace('/[^a-zA-Z]$/', '', $item);
	$item = rcube_utils::idn_to_ascii($item);
	$result[] = $item;
	}

Note that the regexp itself is wrong, since it strips any non-Latin letter, and preg_replace is also called before idn_to_ascii - I'm not a PHP expert, but this may also cause problems if preg_replace expects the input to be ASCII-encoded.

[Player701]

UPD: If preg_replace supports Unicode property escapes then this should probably work:
$item = preg_replace('/[^\p{L}]$/u', '', $item);
The general intent is still the same, but it should not erroneously affect letters from non-Latin alphabets anymore.