nullmailer failing with gnutls >= 3.6.3 under Debian
"harald.a-little-linux-box.at via nullmailer" <[email protected]> Fri, 5 Oct 2018 15:22:25 +0200
| Newsgroups | gmane.mail.nullmailer |
|---|---|
| Message-ID | <[email protected]> |
Seems like nullmailer is having a problem with the new 3.6.4 release of gnutls (as well as with 3.6.3 version): 7174 nullmailer-send->gnutls_error_is_fatal(0, 0, 0, 0x560d7cc36460) = 0 7174 nullmailer-send->gnutls_record_send(0x560d7cc36470, 0x560d7cc46430, 36, 0 <unfinished ...> 7174 libgnutls.so.30->gnutls_record_send2(0x560d7cc36470, 0x560d7cc46430, 36, 0 <unfinished ...> 7174 libgnutls.so.30->gnutls_record_get_max_size(0x560d7cc36470, 0x560d7cc46430, 0x7f0ef7d49e60, 0) = 0x4000 7174 libgnutls.so.30->gnutls_record_overhead_size(0x560d7cc36470, 0x560d7cc46430, 0x7f0ef7d49e60, 0 <unfinished ...> 7174 libgnutls.so.30->_gnutls_record_overhead(0x7f0ef7d49e60, 0x7f0ef7d471c0, 0x7f0ef7d49bc0, 1) = 17 7174 <... gnutls_record_overhead_size resumed> ) = 22 7174 libgnutls.so.30->gnutls_aead_cipher_encryptv(0x560d7cc51e20, 0x7ffcecde3f10, 12, 0x7ffcecde3ed0) = 0 7174 <... gnutls_record_send2 resumed> ) = 36 7174 <... gnutls_record_send resumed> ) = 36 7174 nullmailer-send->gnutls_record_recv(0x560d7cc36470, 0x560d7cc58510, 4096, 0x560d7bce839c <unfinished ...> 7174 libgnutls.so.30->gnutls_aead_cipher_decrypt(0x560d7cc51cf8, 0x7ffcecde3de0, 12, 0x7ffcecde3dd0 <unfinished ...> 7174 libgnutls.so.30->gnutls_memcmp(0x560d7cc461fa, 0x7ffcecde3c70, 16, 0xd598af805c08b3d4) = 0 7174 <... gnutls_aead_cipher_decrypt resumed> ) = 0 7174 libgnutls.so.30->gnutls_buffer_append_data(0x560d7cc36798, 0x560d7cc53914, 53, 53) = 0 7174 <... gnutls_record_recv resumed> ) = -28 As you can see the send works but the recv gets -28 which means GNUTLS_E_AGAIN: https://github.com/rsyslog/rsyslog/issues/1840 rsyslog has the same problem https://codesearch.debian.net/search?q=gnutls_record_recv+package%3Aneomutt and this is how neomutt deals with the issue (gnutls/gnutls.h contains GNUTLS_E_AGAIN) Could you please take a look at this issue? Thanks Harald