Re: [boost] [bbv2] install target. Differences between bbv1 and bbv2

From: Rene Rivera (grafikrobot_at_[hidden])
Date: 2006-10-21 11:23:04

Roland Schwarz wrote:
> I just tried a install of bbv1 and bbv2 on win2k
> using the mingw toolset.

Thanks Roland for doing this testing... I have a few new TODOs now :-)

> (For bbv2 I put in user.config.jam:
>   using gcc : mgw3.4.2 : C:\Mingw\bin\g++.exe ; )

This is the cause of not getting the version number on the generated 
libs. To get the toolset version into the tag the version number has to 
be first. For example, for STLport+MinGW I have:

using gcc : 3.4.5~mingw~stlport51 : C:/MinGW/bin/g++.exe ;

> I also tried to use the lib files of bbv2.
> They indeed work without having an import
> lib available. (No import lib might be a
> problem when trying to link from MSVC.
> But then, since the name mangling is
> different this might not be an issue,
> since one wouldn't need to do this anyways.)

Yea, the MinGW linker is smart enough to auto generate the import lib 
directly from the DLL :-) Oh, how I wish for the day when Microsoft 
tools are that user friendly.

> Static libs are only generated for:
> prg_exec_monitor, test_exec_monitor and
> unit_test_framework. Also they end in *.a
> instead of the expected *.lib. (But this might
> still be corrct.)
> e.g:
> libboost_unit_test_framework-gcc-mt-1_34.a

Hm, if I remember correctly from when I researched what libs work with 
MinGW, it can handle the *.a variants. It searches for a large variety 
of library names when the "-lsomelib" is specified to try and find it. 
Although I chose *.lib names in BBv1 because that was consistent to how 
the other Windows toolsets work.

> I'd like to raise the question if library tagging
> and set of generated libraries indeed should be
> allowed to be so different between bbv1 and bbv2.

I'd say no they should not differ as much as possible. The larger the 
difference, the more pain we inflict on users who have to change their 
builds to accommodate new names.

> I also doubt that gcc name is sufficient without
> a version and platform (cygwin / mingw).

I don't doubt it, I'm certain it's insufficient :-)

-- 
-- Grafik - Don't Assume Anything
-- Redshift Software, Inc. - http://redshift-software.com
-- rrivera/acm.org - grafik/redshift-software.com
-- 102708583/icq - grafikrobot/aim - grafikrobot/yahoo