Re: [boost] Re:Re:Re:Re:PlatformNeutrality-withoutreinterpret_cast<>andifdef

From: William E. Kempf (williamkempf_at_[hidden])
Date: 2002-08-06 13:40:04

----- Original Message -----
From: "Eric Woodruff" <Eric.Woodruff_at_[hidden]>
To: "boost" <boost_at_[hidden]>
Sent: Tuesday, August 06, 2002 11:18 AM
Subject: [boost]
Re:Re:Re:Re:PlatformNeutrality-withoutreinterpret_cast<>andifdef

> At least consider separate headers for the win32 vs posix headers and put
> the thread_group in the main thread.hpp that includes the implementation
> specific ones. Or... have a _win32_thread_attributes.hpp that contains the
> specific attributes of that platform, so they can be modified externally.
> The check for which platform to include could be done either inside the
> platform specific header, or outside--whichever is decided best.
Personally
> I think it would be cleaner to be done at the location of the
implementation
> to make the thread.hpp cleaner.
>
> Current version:
> private:
> #if defined(BOOST_HAS_WINTHREADS)
>     void* m_thread;
>     unsigned int m_id;
> #elif defined(BOOST_HAS_PTHREADS)
> private:
>     pthread_t m_thread;
> #elif defined(BOOST_HAS_MPTASKS)
>     MPQueueID m_pJoinQueueID;
>     MPTaskID m_pTaskID;
> #endif
>     bool m_joinable;
> };
>
>
> In the clutter it apparently wasn't even noticed that an extra "private:"
> was there.

Actually, there's a missing private there ;).   The extra private
declaration is conforming, and in this case helps to prevent some
maintainance mistakes.  The BOOST_HAS_MPTASKS section wasn't written by
myself, or it would have had a "private:" as well.

Seperating out the above into seperate headers will add a small amount of
comprehension for end users, while vastly complicating maintainance.  It's a
tradeoff.  There's no "right" choice here, and as Beman pointed out no
matter which choice I make there will be someone who dislikes that choice.
Given that, and the fact that the membership has already spoken here, I'll
keep the more maintainable choice.

Bill Kempf