Re: [boost] [locale] review part 2.2: source

Subject: Re: [boost] [locale] review part 2.2: source
From: Gevorg Voskanyan (v_gevorg_at_[hidden])
Date: 2011-04-15 10:46:42

Artyom wrote:
> > line 211: I would tend to   use
> >   icu-sources = [ glob icu/*.cpp ] ;
> >   result  +=  <source>$(icu-sources)
> >   I see you're using  something  like
> >   this later on.
> > 
> 
> The problem  comes when you want to put something 
> temporary there out of  project.
> 
> Never liked glob

Liking or not glob may be a matter of taste, but here Steven was also suggesting 
to drop the 'for' loop and take advantage of variable expansion semantics 
instead. The code can be simplified to:
local s = boundary codecvt ... ;
result += <source>$(s).cpp ;

> > line 278: Boost.Thread ought to set  this  flag
> >   automatically.  I'm assuming from the  fact
> >   that  you have it, that it doesn't.
> >   Looks
> >   like a bug in the thread  Jamfile.

I'd say rather a bug in boost Jamroot, should have <define>BOOST_ALL_NO_LIB=1 in 
project's usage requirements.

> > (Similarly,
> >   you should add   <define>BOOST_LOCALE_NO_LIB
> >   in the usage requirements  of  boost_locale)
> 
> I explain, I use boost.thread's mutexes but I  don't
> want to link with this library. So in many cases
> I use header-only  boost.thread so I prefer specify
> explicitly that I don't need  Boost.Thread
> 
> (Boost.Thread is linked only for ICU backend where TLS is  needed)

Provided Boost.Locale lives in the boost tree, there is no need for 
<define>BOOST_THREAD_NO_LIB=1 as boost's Jamroot has <define>BOOST_ALL_NO_LIB=1 
in project requirements.

> > 
> > test/Jamfile.v2: You don't actually  need
> >    test-suite.  test-suite is exactly  the
> >   same as alias.
> > 
> 
> What do you mean?

alias boost_locale_test
   :
   ...
;

Gevorg