Join GitHub today
GitHub is home to over 50 million developers working together to host and review code, manage projects, and build software together.
Sign upDKIM hash mismatch (failure) on sent emails, only when composed as HTML #2004
Comments
|
@ChristopherW You can see: dkim=pass header.i=@test.domain.fr
I use Postfix + Amavisd-new (DKIM). |
|
Salut @Philippe34, merci pour la réponse et le test. I'm still not sure why the DKIM was only failing to validate in certain cases. I'll continue to run my own tests to see if I can figure out what the cause was. |
|
Sorry im late to this party. I dont think its rainloop that would modify your header after it was queued in postfix ( or whatever smtp you use ). I would guess that the mail header is modified by a spam filter after its signed and before it leaves your mail server. |
RainLoop version, browser, OS:
1.14.0 Community on CentOS 6 (connecting to a CentOS 7 mailserver over IMAP). Locally: latest Chrome, Windows 10.
Expected behavior and actual behavior:
Email sent in RainLoop should pass with a valid DKIM header when assessed by the receiving server. Actual behaviour: GMail or any other MX (e.g. Postfix) indicates either an unprotected key or a DKIM failure when the email is received and headers are inspected.
Steps to reproduce the problem:
Send an email with RainLoop using the HTML composer.
Logs or screenshots:
Email sent via RainLoop which fails DKIM on receipt (certain details redacted):
Email sent via IMAP client (Aquamail Pro on Android) which passes DKIM:
To rule out DKIM, DNSSEC or Postfix, I regenerated all DKIM keys, DNSSEC resigned all records and triple-checked everything had propagated. All servers restarted and checked for correct operation.
It seems, reliably, that when an email is sent using RainLoop composed in HTML mode, something is happening which is making the header hash mismatch. Sent in plaintext, DKIM validates perfectly. Emails sent as plaintext or HTML with other clients always pass DKIM. Perhaps it's how RL is encoding HTML special chars?
Happy to do more testing and provide unredacted logs privately.