Sorry but on a second thought, this is wrong for several reasons.
The main one is that aliases transforms recipient addresses BEFORE they are
matched
to a local user, so imagine you have something like:
root: gilles
@: gilles
If I send mail to e...@poolp.org and there's a local user eric, what should it
do ?
Either you consider that local user eric should get it but this contradicts
aliases
configuration as it now explicitly states that *@poolp.org should resolve to
gilles
and, if that's what I really wanted, it would be the only way to express it.
Or you consider that gilles is supposed to get it but then it means that a
catchall
will get mail for any recipient that is not listed in the aliases
configuration. So
to receive mail for eric, you'd need to list him as such:
root: gilles
eric: eric
@: gilles
... and it turns domain aliases into virtual domains aliases as the main
difference
between both is precisely that aliases are allowed to not resolve to a user
whereas
virtual requires resolution to succeed.
You want a mechanism that has enough knowledge of the recipients to determine
which
ones do not exist and should be handled by the catchall but aliases aren't
meant to
do that: they are expectedly allowed to have missing keys for bypass.
If I were you, I'd try to fit my use-case in virtual rather than aliases as
this is
the mechanism intended for what you're trying to do. I had a look at Postfix
and if
I'm not mistaken they also have catchall part of virtual rather than aliases,
which
would likely be for the same reasons.
Gilles
April 15, 2022 1:10 AM, gil...@poolp.org wrote:
> Forget about the whole fallback idea, it can't work because an envelope can
> only
> successfully match a single rule for _very_ good reasons that are not related
> to
> the implementation itself but to how SMTP works more generally. The ruleset
> does
> a first-match evaluation and once it found a rule that matched, that is the
> only
> rule it can consider for that envelope otherwise TONS of unfixable issues
> arise.
> I can explain if there's interest, but it's very technical and unrelated to
> your
> issue.
>
> The only reason you are trying to fallback is because aliases_get() did not
> have
> the same wildcard logic as virtual_get(), otherwise with a configuration like:
>
> match from any for domain "domain.com" action "local_mail"
>
> and an alias record of "@: username" (or an inlined { "@" = username }), the
> use
> case you have is solved without having to think about a fallback rule.
>
> Care to try the following diff ?
>
> Index: aliases.c
> ===================================================================
> RCS file: /cvs/src/usr.sbin/smtpd/aliases.c,v
> retrieving revision 1.78
> diff -u -p -r1.78 aliases.c
> --- aliases.c 28 Apr 2020 21:46:43 -0000 1.78
> +++ aliases.c 14 Apr 2022 22:40:24 -0000
> @@ -66,8 +66,15 @@ aliases_get(struct expand *expand, const
>
> /* no user-part tag, try looking up user */
> ret = table_lookup(mapping, K_ALIAS, buf, &lk);
> + if (ret < 0)
> + return (-1);
> + if (ret)
> + goto expand;
> +
> + /* Failed ? We lookup for a *global* catch all */
> + ret = table_lookup(mapping, K_ALIAS, "@", &lk);
> if (ret <= 0)
> - return ret;
> + return (ret);
>
> expand:
> /* foreach node in table_alias expandtree, we merge */
>
> April 14, 2022 9:00 PM, "Beau Ford" <bf...@0x.co> wrote:
>
>> On Thu, 14 Apr 2022, Beau Ford wrote:
>
> 1) Am I still using my aliases that are defined in:
>
> table aliases file:/usr/local/etc/mail/aliases
> action "local_mail" mbox alias <aliases>
>
> ... or are those superceded now by the catchall ?
>> It turns out I am *not* using my aliases anymore and that my rules, as
>> they are written, are mutually exclusive.
>>
>> If I order them like this:
>>
>> match from any for domain "domain.com" action "local_mail"
>> match from any for domain "domain.com" action "catchall"
>>
>> ... then my aliases table is processed, like normal, and the catchall is
>> ignored. This makes sense, because the rule matches and it completes -
>> there is no reason to move on to the catchall address.
>>
>> Nonexistent addresses (that I hoped to get with the catchall) bomb out
>> with 550 Invalid recipient.
>>
>> HOWEVER, if I reverse the order:
>>
>> match from any for domain "domain.com" action "catchall"
>> match from any for domain "domain.com" action "local_mail"
>>
>> ... then the catch-all works and I lose all of my aliases. My aliases
>> table is not used. Again, makes sense - the first rule matches and
>> completes.
>>
>> -----
>>
>> So, how can I say (pseudocode):
>>
>> match from any person actually a user or in my aliases table action
>> "local_mail"
>>
>> ... which would fail for nonexistent addresses, which is GOOD, and then
>> successfully move down to the catchall match ?
