From a848231ae0f346dc7cc000973fbeb65b0894ee92 Mon Sep 17 00:00:00 2001 From: Daniel Baumann Date: Wed, 10 Apr 2024 21:59:03 +0200 Subject: Adding upstream version 3.8.5. Signed-off-by: Daniel Baumann --- implementation-notes/ENHANCED_STATUS_CODES | 58 ++++++++++++++++++++++++++++++ 1 file changed, 58 insertions(+) create mode 100644 implementation-notes/ENHANCED_STATUS_CODES (limited to 'implementation-notes/ENHANCED_STATUS_CODES') diff --git a/implementation-notes/ENHANCED_STATUS_CODES b/implementation-notes/ENHANCED_STATUS_CODES new file mode 100644 index 0000000..2b4c2af --- /dev/null +++ b/implementation-notes/ENHANCED_STATUS_CODES @@ -0,0 +1,58 @@ +Postfix enhanced status code implementation notes +================================================= + +RFC 3463 Enhanced status code support is implemented in stages. In +the first stage, the goal is to minimize code changes (it's several +hundred pages of context diffs already). For this reason, the +pre-existing status variables (success, defer, etc.) are still +updated separately from the diagnostic text and the RFC 3463 enhanced +status code. All this means that one has to be careful when updating +the code, to keep things in sync. + +Specific issues that one should be aware of: + +- In the SMTP client, update the enhanced status code and text +whenever smtp_errno or resp->code are updated, or place an explicit +comment that says no update is needed. + +- In the SMTP client, don't worry about the initial enhanced status +digit when reporting failure to look up or connect to a host. For +convenience, the SMTP client top-level code automatically changes +the initial digit into '4' or '5' as appropriate. + +- In the SMTP server, don't worry about the initial enhanced status +code digit when an smtpd_mumble_restriction rejects access. For +convenience, the smtpd_check_reject() routine automatically changes +the initial digit into '4' or '5' as appropriate. + +- Some low-level support routines update the diagnostic text but +not the enhanced status code. To identify these, search for functions +that are called with why->vstring as output parameter, and make +sure that the caller updates the enhanced status code in all +appropriate cases. + +- By design, the pipe, local and virtual delivery agent code never +update the diagnostic text separately from the enhanced status code. + +- Don't rely on the system errno value after calling a routine that +performs or prepares for mail delivery. Instead, have that routine +update the enhanced status code (and text) when the error happens. +This was an issue with mailbox, maildir and file delivery. Currently +there remains one exception to this errno usage rule; the maildir +delivery routines log a helpful warning when delivery fails with +EACCES. The latter happens to work because mbox_open() does not +need to unlock the output file, so it won't clobber the errno value. + +- Avoid passing around strings that combine enhanced status code +and diagnostic text. Instead, use separate variables for status +code and text, so that the compiler can enforce that everything has +a status code. Currently there are two exceptions to this rule: +the cleanup server status reply, and the delivery agent status +reply. Once these protocols are updated we can remove the dns_prepend() +routine. The third exception, enhanced status codes in external +command output, is a feature. + +- The bounce/defer/sent library modules will catch the cases where +an enhanced status code does not match the reject/defer/success +status. They log a warning message, and replace the incorrect +enhanced status code by a generic one. -- cgit v1.2.3