From: Rene Rivera (grafik.list_at_[hidden])
Date: 2005-05-23 10:16:16
Vladimir Prus wrote:
> On Monday 23 May 2005 18:30, Rene Rivera wrote:
>>David Abrahams wrote:
>>>Vladimir Prus <ghost_at_[hidden]> writes:
>>>>On Monday 23 May 2005 16:05, David Abrahams wrote:
>>>>I'd rather not have two files for a single toolset.
>>>Fine with me; I'm inclined the same way.
>>And I'm inclined the other way :-) The reason I gave earlier was that
>>having all that auto detection code in msvc.jam makes for a harder to
>>read and maintain toolset file. For example the detection code in
>>msvc.jam is half of the file, and it's not until about 2/3 down in the
>>file when you get to see the flags, etc.
> Well, we have move the detection code to the bottom of the file ;-)
Don't think my rational matters.. as it's two against one in this case ;-)
>>>>What if no registry keys are available and there are two 'cl.exe' in
>>>>PATH? Then it's not possible to guess the versions of those cl.exe
>>>Corner case. The user uses explicit config for that one:
>>> using msvc : 7.1 : /some/path ;
>>Or we eventually get to the point where can we run "cc --version" (or
>>equivalent thereof) to ask the compiler directly.
> Yea. Another question -- what if the user has just one compiler. Should he
> see "-8.0" in build path or not?
I say yes. Trying to decide if a user *really* has only one compiler is
not worth the effort just to remove that number from the path. And what
would happen when they upgrade, or add another version? Object files
would start to collide without warning.
-- -- Grafik - Don't Assume Anything -- Redshift Software, Inc. - http://redshift-software.com -- rrivera/acm.org - grafik/redshift-software.com -- 102708583/icq - grafikrobot/aim - Grafik/jabber.org
Boost-Build list run by bdawes at acm.org, david.abrahams at rcn.com, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk