Re: lda fails in parse_angle_addr if sieve is enabled
Hi,
This code path is only executed when there is a return-path header. Given the fact that this fails consistently, can you provide an example of a message successfully delivered using an older version of Dovecot; i.e., as it is stored in the mailbox? There should be a return-path header there.
Alternatively, you could temporarily wrap dovecot-lda in a shell script that copies the message to some file you can read after.
I just need one example to confirm my suspicions.
Regards,
Stephan.
Op 12-4-2018 om 11:54 schreef Leo Gaspard:
(sorry for the absence of In-Reply-To, I could neither find it on the ML interface nor figure out a way to customize it in Thunderbird, and am not subscribed)
Just wanted to chime in to say I've hit the same issue with the upgrade from 2.3.0.1 to 2.3.1. My setup is an opensmtpd sending mail to rspamd, that then delivers to dovecot LDA.
In the logs, I see (slightly reformated and censored):
Panic: file message-address.c: line 147 (parse_angle_addr): assertion failed: (*ctx->parser.data == '<') Error: Raw backtrace: /nix/store/[…]-dovecot-2.3.1/lib/dovecot/libdovecot.so.0(+[…]) […] -> /nix/store/[…]-dovecot-2.3.1/lib/dovecot/libdovecot.so.0(+[…]) […] -> /nix/store/[…]-dovecot-2.3.1/lib/dovecot/libdovecot.so.0(i_fatal+0) […] -> /nix/store/[…]-dovecot-2.3.1/lib/dovecot/libdovecot.so.0(+[…]) […] -> /nix/store/[…]-dovecot-2.3.1/lib/dovecot/libdovecot.so.0(+[…]) […] -> /nix/store/[…]-dovecot-2.3.1/lib/dovecot/libdovecot.so.0(message_address_parse_path+[…]) […] -> /nix/store/[…]-dovecot-2.3.1/lib/dovecot/libdovecot-lda.so.0(mail_deliver_get_return_address+[…]) […] -> /etc/dovecot/modules/lib90_sieve_plugin.so(+[…]) […] -> /nix/store/[…]-dovecot-2.3.1/lib/dovecot/libdovecot-lda.so.0(mail_deliver+[…]) [0x7f04e2d08e20] -> /nix/store/[…]-dovecot-2.3.1/libexec/dovecot/deliver(main+[…]) […] -> /nix/store/[…]-glibc-2.26-131/lib/libc.so.6(__libc_start_main+[…]) […] -> /nix/store/[…]-dovecot-2.3.1/libexec/dovecot/deliver(_start+[…]) […]
For some reason there are not many symbols, but hope that helps!
participants (1)
-
Stephan Bosch