Re: improving user.c error messages - Mailing list pgsql-hackers

From Nathan Bossart
Subject Re: improving user.c error messages
Date
Msg-id 20230127231507.GA2221340@nathanxps13
Whole thread Raw
In response to Re: improving user.c error messages  (Alvaro Herrera <alvherre@alvh.no-ip.org>)
Responses Re: improving user.c error messages
List pgsql-hackers
On Fri, Jan 27, 2023 at 07:31:19PM +0100, Alvaro Herrera wrote:
> On 2023-Jan-26, Nathan Bossart wrote:
>>              ereport(ERROR,
>>                      (errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
>> -                     errmsg("permission denied: bootstrap user must be superuser")));
>> +                     errmsg("permission denied to alter role"),
>> +                     errdetail("The bootstrap user must be superuser.")));
> 
> I think this one isn't using the right errcode; this is not a case of
> insufficient privileges.  There's no priv you can acquire that lets you
> do it.  So I'd change it to unsupported operation.

І fixed this in v4.  I've also attached a second patch in which I've
adjusted the messages that Peter mentioned upthread.

One thing that feels a bit odd is how some of the DETAILs mention the
operation being attempted while others do not.  For example, we have

    ERROR:  permission denied to drop role
    DETAIL: You must have SUPERUSER privilege to drop roles with SUPERUSER.

In this case, the DETAIL explains the action that is prohibited.  In other
cases, we have something like

    ERROR:  permission denied to alter role
    DETAIL: You must have CREATEROLE privilege and ADMIN OPTION on role "myrole".

which does not.  I think this is okay because adding "to alter the role" to
the end of the DETAIL seems kind of awkward.  But in other cases, such as

    ERROR:  permission denied to use replication slots
    DETAIL:  You must have REPLICATION privilege.

adding the operation to the end seems less awkward (i.e., "You must have
REPLICATION privilege to use replication slots.").  I don't think there's
any information lost by omitting the action in the DETAIL, so perhaps this
is just a stylistic choice.  I think I'm inclined to add the action to the
DETAIL whenever it doesn't make the message lengthy and awkward, and leave
it out otherwise.  Thoughts?

-- 
Nathan Bossart
Amazon Web Services: https://aws.amazon.com

Attachment

pgsql-hackers by date:

Previous
From: Andres Freund
Date:
Subject: Re: Optimizing PostgreSQL with LLVM's PGO+LTO
Next
From: Tom Lane
Date:
Subject: Re: heapgettup() with NoMovementScanDirection unused in core?