diff options
author | Eric S. Raymond <esr@thyrsus.com> | 2000-02-25 02:35:35 +0000 |
---|---|---|
committer | Eric S. Raymond <esr@thyrsus.com> | 2000-02-25 02:35:35 +0000 |
commit | f2fc9429ce077282f1a4ac324f42b29ecaedf4a8 (patch) | |
tree | 91fe891f501221c9042b913a4ba82447b586ac3c | |
parent | 822f4569940ae8ff0a90f04df6ff897f7e40c6f1 (diff) | |
download | fetchmail-f2fc9429ce077282f1a4ac324f42b29ecaedf4a8.tar.gz fetchmail-f2fc9429ce077282f1a4ac324f42b29ecaedf4a8.tar.bz2 fetchmail-f2fc9429ce077282f1a4ac324f42b29ecaedf4a8.zip |
Deal better with absence of a Return-Path.
svn path=/trunk/; revision=2772
-rw-r--r-- | NEWS | 5 | ||||
-rw-r--r-- | sink.c | 22 |
2 files changed, 21 insertions, 6 deletions
@@ -2,6 +2,11 @@ (The `lines' figures total .c, .h, .l, and .y files under version control.) +* Use remotename@hostname for MAIL FROM if we have not been able to deduce + a Return-Path. +* Fix the attempted fix for Joop Susan's ENOTCONN bug. +* Added FAQ material on a Microsoft Exchange bug. + fetchmail-5.3.0 (Tue Feb 22 08:53:31 PST 2000), 18618 lines: * Horst von Brand's improvements to the specfile generator. @@ -716,22 +716,32 @@ int open_sink(struct query *ctl, struct msgblk *msg, /* * Try to get the SMTP listener to take the Return-Path - * address as MAIL FROM . If it won't, fall back on the - * calling-user ID. This won't affect replies, which use the - * header From address anyway. + * address as MAIL FROM. If it won't, fall back on the + * remotename and mailserver host. This won't affect replies, + * which use the header From address anyway; the MAIL FROM + * address is a place for the SMTP listener to send + * bouncemail. The point is to guarantee a FQDN in the MAIL + * FROM line -- some SMTP listeners, like smail, become + * unhappy otherwise. * * RFC 1123 requires that the domain name part of the * MAIL FROM address be "canonicalized", that is a - * FQDN or MX but not a CNAME. We'll assume the From + * FQDN or MX but not a CNAME. We'll assume the Return-Path * header is already in this form here (it certainly * is if rewrite is on). RFC 1123 is silent on whether * a nonexistent hostname part is considered canonical. * * This is a potential problem if the MTAs further upstream * didn't pass canonicalized From/Return-Path lines, *and* the - * local SMTP listener insists on them. + * local SMTP listener insists on them. */ - ap = (msg->return_path[0]) ? msg->return_path : user; + if (msg->return_path[0]) + ap = msg->return_path; + else + { + sprintf(addr, "%s@%s", ctl->remotename, ctl->server.truename); + ap = addr; + } if (SMTP_from(ctl->smtp_socket, ap, options) != SM_OK) return(handle_smtp_report(ctl, msg)); |