Re: [boost] [guid] - review results? - inline missing

From: Martin Bonner (Martin.Bonner_at_[hidden])
Date: 2007-06-01 04:16:29

From: Martin Bonner
> 

> From: Peter Dimov
> Sent: Thu 31/05/2007 18:04

>> Hartmut Kaiser wrote:
>>> You need to add an explicit 'inline' specifier to any function
>>> defined out of class but in a header, regardless if it's a
templatized
>>> function or not.

>> I don't think that this is true.

> Why don't you think this is true?
 
> If you have the same function defined in two translation units 
> (because it is dragged into both via #include), it's a violation 
> of the One Definition Rule unless the definitions have inline.
 
> Apologies for the crap quoting - blame Exchange Webmail

OK.  I've fixed the crap quoting.  Now it's time to apologize for the
crap content.  Specifically, section 3.2 [one definition rule] paragraph
5 says:

"There can be more than one definition of a ... non-static function
template ... in a program provided that each definition appears in a
different translation unit, and provided the definitions satisfy the
following requirements."

(The requirements are met by a definition dragged into two translation
units via #include.)

In other words Peter is right; you don't need to declare a function
template as "inline". If you do, the only effect may be to encourage the
compiler to inline it.
-- 
Martin Bonner 
Project Leader 
PI SHURLOK LTD
Telephone: [phone] / 203894 (direct)
Fax: [phone]
Email: martin.bonner_at_[hidden]
<mailto:martin.bonner_at_[hidden]>  
www.pi-shurlok.com <http://www.pi-shurlok.com/>