Re: [jamboost] Re: Re: PCH support

From: Vladimir Prus (ghost_at_[hidden])
Date: 2004-09-09 01:57:48

Thorsten Ottosen wrote:

> | I don't have practical experience, but I believe that all files in my
>
> project
>
> | start to indirectly include BGL headers, the performance on compilers
>
> without
>
> | PCH will be very poor.
>
> yeah, that should not be a problem...it is easy to include the correct
> headers in the non-pch case if you
> can specify bjam -no_pch or something

Ok, agreed. 

> | > I would hope these two headers could be added to boost:
> | >
> | > /boost/std.hpp
>
> attached in beta version.

Could you please send it in some other format than uuencode. My mailer does 
not recognize it as attachment.

I think I'll try playing with this header on boost with gcc 3.4.

> | I think that thanks to another user, I more or less understand 7.1. Help
>
> with
>
> | Comeau will be appreciated. And I can act as the person in charge ;-)
>
> great!
....
> I couldn't quite use that syntax. To create a precompiled header I needed a
> dummy .cpp file:
>
> #include "std.hpp"
>
> int main()
> {
> }
>
> Then I create the file by
>
> como --dinkumware --create_pch=std.pch std.cpp
>
> (notice "=")
>
> Then I use the file as
>
> como --dinkumware --use_pch=std.pch vector_string.cpp

Ok, so this is basically the same as msvc model, except I don't have to 
specify the header file where PCH compilation should stop.

I presume that the 'obj' created when producing PCH should be linked with the 
rest of the object files?

> Having a standard directory for the pch-files is probably a good idea. In
> that case add --pch_dir=<dir>.

What effect does this option have for the case of manual PCH creation?

> hope it helps

Sure. Thanks! 

BTW, are you using Comeau on Windows? If yes, then, supposing I'll write 
Comeau toolset for V2, could you spend a few minutes testing it? If I 
understand correctly, that's the only toolset used now in boost regressions 
which V2 does not have.

- Volodya