On Wed, 8 Jan 2003, Matthias Andree wrote:
"" should work, "<>" is just a workaround for broken sendmails (up to
and including 8.12.6, maybe even later, haven't checked 8.12.7).
True, but msgblk.return_path[0] is checked to see if there already is a
return path set from the Return-Path: header, and if not, it is set to
resent-sender, or sender, or resent-from, etc. At that time, the
Return-Path: header is no longer available (except for ctl->mda).
So, right now there's no difference between an empty Return-Path: header
and no Return-Path: header at all if you represent an empty return path by
just "".
For that reason, and because sink.c already contains a check for
"<>" (which doesn't even work right now: strcmp(msg->return_path, "<>")
will never be 0), I suggested using "<>". This however breaks
sanitize(from) in open_mda_sink().
If there's no Return-Path: header, and none of the other headers listed to
take it's place in the return_path variable, return_path will be "".
Which will work for mda, maybe not for bsmtp, and for smtp_sink it will
result in <remotename(_at_)server(_dot_)truename>. As a side note: why doesn't
open_bsmtp_sink() include all the return_path related code that's in
open_smtp_sink()?
While -f "<>" is only a workaround for -f "", would it be OK to always use
"<>" instead of "" ? Would it break things?
Regards,
Maarten.