Re: Ticket 298: bug on pg_hba.conf editor

On Saturday, July 16, 2011, Guillaume Lelarge <guillaume@lelarge.info> wrote:
> On Sat, 2011-07-16 at 21:11 +0100, Dave Page wrote:
>> On Sat, Jul 16, 2011 at 1:12 PM, Guillaume Lelarge
>> <guillaume@lelarge.info> wrote:
>> > Hi,
>> >
>> > I worked a bit this morning on this bug. The editor was made in a way
>> > that invalid configuration lines are not displayed which is wrong
>> > because you can't fix a line if you stored it wrong once.
>> >
>> > So I did the change to allow the change of an invalid configuration
>> > line, and that works well.
>> >
>> > But I now have many other lines that aren't supposed to appear:
>> >
>> > # local      DATABASE  USER  METHOD  [OPTIONS]
>> > # host       DATABASE  USER  ADDRESS  METHOD  [OPTIONS]
>> > # hostssl    DATABASE  USER  ADDRESS  METHOD  [OPTIONS]
>> > # hostnossl  DATABASE  USER  ADDRESS  METHOD  [OPTIONS]
>> > # host name, or it is
>> >
>> > All are considered comments, and all have a valid first column, so all
>> > are displayed. Which is a bit disturbing because they are part of the
>> > comments in pg_hba.conf, they are not supposed to be "actual" lines.
>> >
>> > So, they match our process of identifiying lines, and so they are
>> > displayed. Do you have any idea how we could not display these? I mean,
>> > I can simply add a check on the line string to see if they are equal to
>> > the one of the five strings above, but it seems quite a ugly hack.
>>
>> Why don't we just ignore anything that starts with a # ?
>>
>
> Because we need to guess which comment is an actual comment and which
> comment is a disabled configuration. That allows us to hide actual
> comments, and show disabled configuration. Problem is that our guess is
> wrong sometimes.

Sounds like you're trying to be too clever. We don't normally care
about commented lines in configuration files. If you really want to do
so tough, check if a token is wrapped in [ ] - that never happen in a
valid configuration I don't believe.

>> > Or do we simply choose to not care? we prefer to have the bugfix even if
>> > it means to show some not "actual" config lines?
>>
>> Not those.
>>
>
> I don't get it, sorry :)
>
> What do you mean by "not those"?

We don't want to show those lines.

>> > Another related question: peer, radius are not available in the method.
>> > As we are in beta, I won't add them to 1.14 branch, will I?
>>
>> I would consider their omission to be a bug.
>>
>
> Hmmm, OK. Will fix then.

Thanks.

--
Dave Page
Blog: http://pgsnake.blogspot.com
Twitter: @pgsnake

EnterpriseDB UK: http://www.enterprisedb.com
The Enterprise PostgreSQL Company