Re: [boost] relative includes

From: Beman Dawes (bdawes_at_[hidden])
Date: 2002-02-04 11:13:13

At 03:00 AM 2/4/2002, Darin Adler wrote:
 >On 2/3/02 4:42 PM, "Beman Dawes" <bdawes_at_[hidden]> wrote:
 >
 >> The only thing that jumped out at me was the failure of the smart 
pointer
 >> test for Metrowerks.  The fix is easy; file 
boost/detail/atomic_count.hpp
 >> line 94: #include "atomic_count_win32.hpp" resulted in "file
 >> 'atomic_count_win32.hpp' cannot be opened."  There are a couple of
 >similar
 >> usages on the lines that follow.  Peter?
 >
 >I'm thinking this is not something that should be different in different
 >libraries. If we need to use <boost/...> includes exclusively in the
 ><boost/...> directories, we should make that consistent throughout.
 >
 >The following are the other files in the <boost/...> directory that have
 >relative includes in them:
 >
 >    boost/thread/condition.hpp
 >    boost/thread/mutex.hpp
 >    boost/thread/recursive_mutex.hpp
 >
 >We also have a mix of <boost/...> and "boost/..." to this day. I think we
 >should fix that with a global replace.

I think you are right about the global replace.  We should really move to 
the more correct "boost/..." form, and then document that as the "boost 
way".  But not right before a release!  Also, before committing the change 
to CVS, I'd like to make it to my working copy, and then try the Win32 
regression test to make sure it doesn't break with some compiler.  (My 
memory was the reason we started with the <boost/...> form was some broken 
Win32 compiler, but I don't know which one.)

--Beman

--Beman