mantisbt-dev

Re: [mantisbt-dev] Withdraw version 1.2.5 of MantisBT ?

[Roland B. <ro...@at...>]

Betreff: Re: [mantisbt-dev] Withdraw version 1.2.5 of MantisBT ?
Gesendet: Do, 07. Apr 2011
Von: John Reese<jo...@no...>

> On 4/6/2011 5:27 PM, Roland Becker wrote:
> > the new installation process is broken. (at least when using MySQL)
> > This is caused by a commit to fix
> > http://www.mantisbt.org/bugs/view.php?id=12675 (changed wrong line
> > with $g_db->Connect)
> 
> Is there any information on why that change breaks the upgrade? 
> Especially seeing that it seems to work just fine on my own public 
> installation.

The change does not break the upgrade.
If i understand right the change was to fix a bug when _upgrading_ MantisBT and using _PGSQL_
AFAIK there was no problem when upgrading and using _MySQL_
But seems that the wrong line of the code has been changed. The changed code is executed during a _new_ installation from scratch.
You get an error that wrong version of MySQl is used,. The database is not created, the installation process gets interrupted.

Re: [mantisbt-dev] Withdraw version 1.2.5 of MantisBT ?

[David H. <d...@hx...>]

On Thu, 2011-04-07 at 07:17 +0200, Roland Becker wrote:
> The change does not break the upgrade.
> If i understand right the change was to fix a bug when _upgrading_ MantisBT and using _PGSQL_
> AFAIK there was no problem when upgrading and using _MySQL_
> But seems that the wrong line of the code has been changed. The changed code is executed during a _new_ installation from scratch.
> You get an error that wrong version of MySQl is used,. The database is not created, the installation process gets interrupted.

Thanks Roland. When connecting to PostgreSQL servers a database name is
required. The install script has a feature whereby the username/password
for a database server is verified prior to verification of the database
being created on the server.

It appears that this functionality only works on certain database
backends. I think with PostgreSQL you can only connect to an existing
database (and not to the 'server' as a whole).

For the time being I have rolled back the changes from #12675.

I suspect we'll need to remove the database existence check entirely and
simply report OK/FAIL to the entire connection process (passing database
errors directly to the user for debugging purposes).

The new database API cannot come sooner. I plan on rewriting the install
and upgrade scripts too so they're more reliable, user friendly, modern,
etc.

Regards,

David