7beta3 has naughty parser bug with ALTER USER .. PASSWORD - Mailing list pgsql-bugs

From Mike S.
Subject 7beta3 has naughty parser bug with ALTER USER .. PASSWORD
Date
Msg-id 20000407151140.13005.qmail@web3806.mail.yahoo.com
Whole thread Raw
Responses Re: 7beta3 has naughty parser bug with ALTER USER .. PASSWORD  (Tom Lane <tgl@sss.pgh.pa.us>)
List pgsql-bugs
=======================================================================
                        POSTGRESQL BUG REPORT TEMPLATE
=======================================================================


Your name        :Mike Shell
Your email address    :mikes1987@yahoo.com


System Configuration
---------------------
  Architecture (example: Intel Pentium)      : AMD K6-2 400 (ix86)

  Operating System (example: Linux 2.0.26 ELF)     : Linux 2.2.14 ELF

  PostgreSQL version (example: PostgreSQL-6.5.1): PostgreSQL-7.0beta3

  Compiler used (example:  gcc 2.8.0)        : gcc 2.91.66


Please enter a FULL description of your problem:
------------------------------------------------

I am unable to create users with passwords or give passwords to
existing users - under pgaccess or the SQL interface!
Attempting to give any user a password results in a nice parse error.




Please describe a way to repeat the problem.   Please try to provide a
concise reproducible example, if at all possible:
----------------------------------------------------------------------


Example for pgaccess: there is a user 'tester' for the 'root' database
which has a nice table (only 4 rows) 'addresses'. I then highlight
user 'tester' and click [design]. A password (mypass) is entered for
the
user. The result is a nice parser error:

ALTER user "tester" WITH PASSWORD "mypass" NOCREATEDB NOCREATEUSER

PostgreSQL error message: ERROR: parser: parse error at or near """

Entering the ALTER command manually (without the quotes) via
psql gives me the same result and this should work given the examples
in the user.ps docs.

Futhermore, when a user finally is given a password via the fix below,
and you edit this user in pgaccess, pgaccess shows this user as having
no password. ( ******* should appear in the password and verify
password
areas when editing users with existing passwords or at least a little
note to indicate the password has to be reentered).

Forgive me if this is corrected in beta4, but I found no mention of
this
problem in the bugs archives or mbox. This in itself surprises me.
Everyone IS using passwords on their DB, right?

Also, there are several links in mailing list archives on the
postgresql.org site which aren't pointing to any valid web pages -
usually the latest month. It would also be nice to put IBM's DB2 in the

DB comparison chart - show 'em Postgresql is aiming high!




If you know how this problem might be fixed, list the solution below:
---------------------------------------------------------------------


Playing around with pgsql, I was able to determine the nature of the
problem:

1.<FAILS>: ALTER USER tester with PASSWORD bigpass
2.<FAILS>: ALTER "USER" tester with PASSWORD "bigpass"
3.<FAILS>: ALTER 'USER' tester with PASSWORD 'bigpass'
4.<WORKS>: ALTER USER tester with PASSWORD 'bigpass'
5.<WORKS>: ALTER USER "tester" with PASSWORD 'bigpass'
6.<WORKS>: ALTER USER tester with PASSWORD 'bigpass'

So, we seem to have some inconsistancy in what delimiter
to use. user.ps indicates that the ALTER command, the USER
has no quotes and neither does the PASSWORD. However, VALID UNTIL
requires 'single quote' delimiters because of the white space between
the elements of the date.

I am VERY new to Postgresql, SQL and DB's in general, so I do not feel
very secure in making judgements here IN "What is right' - pun
intended. ;)

However, a brief look at Don Chamberlin's DB2 Universal Database book
tends to suggest that we should be able to do with either case 1 or
case 3 above or any combination, except when the password contains
white space in which case we need single quotes around it.
(Yes, ALTER USER ... PASSWORD is not standard SQL, but borrowing from
similar constructs..)
pgaccess seems to be using "" double quotes, which is OK by me but
not the parser.



You poor folks sure are advancing PostgreSQL at a blistering pace.
Ease back on the throttle a bit before something breaks (sanity ;)

I would love to help more, but I have my hands full with creating the
site which is gonna use the datebase!

Thank you all for this wonderful, powerful database.

 M. Shell



__________________________________________________
Do You Yahoo!?
Talk to your friends online with Yahoo! Messenger.
http://im.yahoo.com

pgsql-bugs by date:

Previous
From: Tom Lane
Date:
Subject: Re: Select from view crashes backend
Next
From: Tom Lane
Date:
Subject: Re: 7beta3 has naughty parser bug with ALTER USER .. PASSWORD