Re: [nail-devel] nail glitch in sendmail command line

[Gunnar R. <Gun...@pl...>]

"REIBER, CHRISTIAN" <CHRISTIAN.REIBER@ZEPPELIN.COM> wrote:

> Testcase (with nail-11.20):
>    nail     command: echo test | mail -s test user1 -r user1
>    sendmail command: /usr/sbin/sendmail -i -r user1           # NOT OK
>
>    nail     command: echo test | mail -s test -r user1 user1
>    sendmail command: /usr/sbin/sendmail -i -r user1 user1     # OK
>
>    nail     command: echo test | mail -s test user1 -r user2
>    sendmail command: /usr/sbin/sendmail -i -r user2 user1     # OK
>
> This is a somewhat exotic situation as the command line is not
> posixly correct,

This is exactly the point here. In your first example, "-r" is not
understood as an option, but as a recipient's address. Nail then
eliminates the duplicate "user1" and reorders the addresses for
sending.

In your third example, it works by chance; "-r" is understood as
an address by nail in this case too.

Your second example is the only one that is syntactically correct.

> but it might point to a problem which may have other consequences too.

The behavior was introduced consciously in version 11.0 for alignment
with System V and POSIX mailx, and was documented in the ChangeLog. I
don't see why the first example actually seemed to work before, but in
any case, that was an accident. It has never been my intent to accept
options after operands.

The only thing I might change here is adding a "--" to tell sendmail
when "-r" is an address. Your examples would then look like

  sendmail command: /usr/sbin/sendmail -i -- -r user1
  sendmail command: /usr/sbin/sendmail -i -r user1 -- user1
  sendmail command: /usr/sbin/sendmail -i -- -r user2 user1

(The "--" is superfluous in the second example, so I would not actually
pass it then. It is here just to highlight the concept.)

	Gunnar