Boost Testing :
From: Reece Dunn (msclrhd_at_[hidden])
Date: 2006-06-24 18:30:46
There are currently some issues when running:
bjam --v2 install msvc-7.1 # 
bjam --v2 install msvc-7.1 link=static,shared # 
These were identified in the "[Boost-build] boost.build v2 and toolset="
discussion on the boost.build mailing list.
Running  highlights a problem with Boost.Wave. Since wave
adds <rtti>off for msvc-7.1, this is causing problems when it
is pulling in Boost.FileSystem. The <tag> rule is using the <rtti>on
version of filesystem, not the <rtti>off version, so BBv2 ends up
trying to create two versions of boost_filesystem-vc71-mt-gd-1_35.lib.
The workaround is to open the following files:
and comment out the line:
<toolset-msvc:version>7.1:<rtti>off # workaround for compiler bug
in each, however this is not an ideal solution :(!
Running  gives an error for boost_date_time-vc71-mt-gd-1_35.lib:
error: Duplicate name of actual target:
error: added properties: <define>BOOST_DATE_TIME_STATIC_LINK <link>static
error: removed properties: <define>BOOST_ALL_DYN_LINK=1 <link>shared
This appears to be because the different <link> values aren't being
Both of these appear to be problems that arise due to when <tag> is
lib foo : bar.cpp : : <tag>foo.bar ;
exe bar : foo.cpp : : <rtti>off ;
the <tag> rule appears to be applied before <rtti>off updates the property
set for foo. This results in the boost_filesystem problem (when inheriting the
<rtti>off property from boost_wave) and I suspect (although I have not
confirmed this) that it is the cause of the boost_date_time problem as well.
As running bjam install is a common thing to do as a user, why isn't this
a part of the regression tests? At the moment, it isn't being picked up
by any of the regression tests, so it is only by users trying to run install
that these are being picked up.
Try Live.com: where your online world comes together - with news, sports, weather, and much more.