From: Date: December 7 2007 9:57pm
Subject: Re: bk commit into 5.1 tree (kostja:1.2614) BUG#25535
List-Archive: http://lists.mysql.com/commits/39572
Message-Id: <20071207205719.GG6748@bodhi>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii

* Sergei Golubchik <serg@stripped> [07/12/07 23:22]:

> > > > > It basically is a part of the public library API - it's the only
> > > > > way for a client to support /etc/my.cnf configuration, which is
> > > > > supposed to be common to all clients.
> > > > 
> > > > I think we mean different things under words "client library'.
> > > > 
> > > > Client library is a Connector/C. Not "a library to use if you
> > > > build a MySQL command line client".
> > > 
> > > Yes, I mean Connector/C. load_defaults() is de-facto part of
> > > Connector/C because any client (*) that wants to support /etc/my.cnf
> > > have to use load_defaults().
> > 
> > load_defaults is not a documented part of the client library API.
> 
> It's explained in Paul's books.
> 
> > Are you suggesting it is added to the manual?
> 
> Perhaps, yes.
> 
> > > (*) and I mean *any*, not "mysql command line client". Gentoo
> > > portage has 162 ebuilds that depend on virtual/mysql, with examples
> > > like games-strategy/freeciv, mail-filter/dspam, media-sound/amarok.
> > 
> > Sure, they depend on the Connector/C. But I would argue that *none* of
> > them uses load_defaults.
>  
> I didn't check these three, right.
> But a google search reveals quite a few references to load_defaults():
> 
> http://www.jsw4.net/info/listserv_archives/mysql/05-wk02/msg00115.html
> http://www.listsearch.com/MySQL/Message/index.lasso?216520
> http://www.searchy.net/mailinglists/MySQL_General/msg11412.html
> http://www.galileo-press.de/openbook/c_von_a_bis_z/c_027_003.htm
> http://search.cpan.org/~darren/MySQL-Config-1.03/lib/MySQL/Config.pm
> (the last is the pupe-perl emulation, shows that the functionality is
> used indeed)

Serg, the path of least resistance is to assume all these users of
load_defaults were doing the right thing and thus put it back.

There is no single reason load_defaults should belong to connector/C, 
except that it was there already.

The whole point of the removal of this function was to attack the
problem of a header file declaration mess.

You're essentially asking me to withdraw my forces because the
mess is in Paul's book and there are some users.

I don't care as much about the matter as I care to understand your
logic, and it does not look very logical (well, maybe yet).

-- 
-- Konstantin Osipov              Software Developer, Moscow, Russia
-- MySQL AB, www.mysql.com   The best DATABASE COMPANY in the GALAXY