Re: Initializing static reference (non-POD) member variables

From:

"Alf P. Steinbach" <alfps@start.no>

Newsgroups:

comp.lang.c++

Date:

Tue, 24 Jul 2007 20:07:37 +0200

Message-ID:

<13acfvc5kkp7f2d@corp.supernews.com>

* Victor Bazarov:

Alf P. Steinbach wrote:

* Grey Alien:

[..]
The problem being:

1. Class A contains a static reference to Class B
2. Class B has no default ctor(s)
3. Class B's ctor takes a reference as one of its non-default
arguments 4. The reference parameter required to construct B is provided
via
A's ctor Is there a way to do this ?

Yes

Yes?

Yes, because the spec is so vague. You just need to think outside the
box. The box you seem to imagine is not present in the spec, which is,
for all practical purposes, useless anyway.

, but (1) that isn't what your code exemplifies, and (2)
initialization of non-local statics generally happens before main() is
called, and at that point you probably don't have any database
connection yet.

It's not impossible to establish a connection before 'main', so that
should not really be an issue.

In practice it's an issue. In theory you can do a lot via
initialization of non-local statics. In practice you run into
initialization order fiasco, error handling impossibilities, etc.

Why don't you explain what you're trying to achieve by using that
"static"?

Sharing the instance of 'B' between all instances of 'A', perhaps?

Whatever, we don't know.

--
A: Because it messes up the order in which people normally read text.
Q: Why is it such a bad thing?
A: Top-posting.
Q: What is the most annoying thing on usenet and in e-mail?