Re: hba.c:3160:18: warning: comparison of unsigned enum expression

Magnus Hagander <magnus@hagander.net>

From: Magnus Hagander <magnus@hagander.net>
To: Erik Rijkers <er@xs4all.nl>, Jacob Champion <pchampion@vmware.com>, Michael Paquier <michael@paquier.xyz>
Cc: "pgsql-hackers@lists.postgresql.org" <pgsql-hackers@lists.postgresql.org>
Date: 2021-04-07T11:24:01Z
Lists: pgsql-hackers
On Wed, Apr 7, 2021 at 1:01 PM Erik Rijkers <er@xs4all.nl> wrote:
>
> Recently (last day or so), I get this warning from gcc 10.2:
>
> -----
> hba.c:3160:18: warning: comparison of unsigned enum expression < 0 is always false [-Wtautological-compare]
>         if (auth_method < 0 || USER_AUTH_LAST < auth_method)
>             ~~~~~~~~~~~ ^ ~
> 1 warning generated.
> -----

This one is from 9afffcb833d3c5e59a328a2af674fac7e7334fc1 (adding
Jacob and Michael to cc)

And it makes sense to give warning on that. AuthMethod is an enum. It
can by definition not have a value that's not in the enum. That check
simply seems wrong/unnecessary.

The only other use fo USER_AUTH_LAST is in fill_hba_line() which also
gets the name of the auth. That one uses :
        StaticAssertStmt(lengthof(UserAuthName) == USER_AUTH_LAST + 1,
                         "UserAuthName[] must match the UserAuth enum");

Which seems like a more reasonable check.

But that also highlights -- shouldn't that function then also be made
to use hba_authname(), and the assert moved into the function? That
seems like the cleanest way?

-- 
 Magnus Hagander
 Me: https://www.hagander.net/
 Work: https://www.redpill-linpro.com/



Commits

  1. Refactor hba_authname

  2. Add some information about authenticated identity via log_connections