Users discussed a bug where rejected email bodies in the /admin/email-logs/rejected section appeared as unreadable Base64 gibberish instead of the original UTF-8 text. lilydjwg reported that while the database stored the data in a damaged state, the original emails from providers like Tencent and Netease were intact. lilydjwg identified the root cause: the Mail library’s part object had correct charset information, but the body was incorrectly treated as ASCII-8BIT by the email cleaner module. supermathie noted that a fix had recently been merged in PR #37125 and requested a sample email to verify the issue. lilydjwg provided a sanitized test case, but upon supermathie’s inquiry, lilydjwg confirmed that the problem was already resolved in the updated version.
The correct mail body starts with 谢谢 in UTF-8 (the full body is much longer), and the damaged text is m6Mtyx7x7j+rjuVupyY= in base64.
The mails in the database are already in this damaged state. Comparing that one with the one from the mail server, I find that the mail was rewritten, and I suspect that the email cleaner module scrambled the mail. (But I don’t have enough Ruby knowledge to understand its unique string encoding feature.)
No. They come from Tencent and Netease. The original mail (I kept it on the mail server to investigate this issue) has no issue with either mutt or Ruby’s Mail lib.
Is this email (assuming it’s one that did not get mangled by Discourse) something you can pass to me so that we can investigate further and possibly use it as a test case?
