Re: pg_hba.conf editor is buggy

Le 06/01/2011 22:33, damien clochard a écrit :
> hi guys,
>
> Last month during a PG training, i found a very bad behaviour from
> pgadmin. Here's the step to follow :
>
> 0- backup your pg_hba.conf
> 1- launch pgadmin
> 2- edit the pg_hba.conf
> 3- Add a new line with an error inside the CIDR address.
> For example : host all all 10.224.24.26 trust
> 4- save
> 5- close pgadmin
> 6- now if you edit pg_hba.conf with a generic text editor
> the line you've added will appear.
> 7- launch pgadmin again
> 8- edit pg_hba.conf
> 9- your line doesn't appear
> 10- you need to modify pg_ba.conf with a text editor or restore your backup
>
> As you may have noticed, between step 4 and step 10 you'll be unable to
> start PostgreSQL because the pg_hba.conf has a syntax error.
>
> I can't describe how much this is confusing for newbies : they write a
> pg_hba rule that seems correct (they just wrote an IP Address instead of
> a CIDR ) and then they restart PG to test it... Now the server won't
> start and if they restart pgAdmin, then the faulty line is hidden !
>
> What's even more confusing is the dialog window that adds a new line....
> There's a field for each part of the pg_hba line. But when it comes to
> the CIDR Address the field label is "IP Address" instead of "CIDR
> Address" or "IP Address + Netmask"
>
> At the very least, i think the behaviour should coherent. Either pgadmin
> accepts faulty lines then it must be able to display them, or
> Pgadmin refuses to write a faulty line in the pg_hba.conf and in that
> case it's ok hide them...
>
> I think the best behaviour would be :
>
> a/ pgadmin refuses to write a faulty line
> b/ pgadmin display faulty lines **in red** (if any)
>
> ...which is basically the opposite to what pgAdmin is doing now... :)
>
>
> Here's more detail :
>
> * Platform : Windows XP and Linux
> * Language : French
> * Distribution : EDB installer / Ubuntu package / Source
> * Version : 1.10.5 and 1.12.2
> * nothing in pgadmin.log
>

I added a ticket (http://code.pgadmin.org/trac/ticket/298). It's not a
part of the code I know much, but I'll try to see what I can do about it.

--
Guillaume
http://www.postgresql.fr
http://dalibo.com