I don't see that this should be a problem. Just follow the instructions in the release notes: a) Download spirit 1.61 from SourceForge.net to a local directory. On my machine its c:/spirit161 b) Define (export) environmental variable SPIRIT_ROOT=c:/spirit161. The build/test jamfile set up is sensitive to the SPIRIT_ROOT variable and will put it in the include path for those systems that are known to require it. Right not that is msvc 6.0 and all tested borland compilers. Other compilers just work with the latest version of spirit from the CVS Tree. This works on my system without any other special adjustments other than the above. So I'm asking that the testers set up their environment as above before running the tests. I don't think that's an unreasonable request. In fact, I believe this should be seen as a vindication and demonstration of the power and flexibility of the bjam/jamfile system rather than an extra burden. Robert Ramey Dave Abrahams wrote:
"Robert Ramey" <ramey@rrsd.com> writes:
It seems I've succeeded in getting my serialization project checked in.
(I've held back the Jamfile in the test directory pending a tiny tweak)
And I have a couple of questions regarding testing.
a) in order for libraries to be built for Borland and VC 6.0 which don't support Spirit 1.8 but do support spirit 1.6, the Jamfile checks for the environmental variable SPIRIT_ROOT and if found presumes it contains the root of the directory where Spirit 1.6 is located. This has worked well in my local environment. An environment which tests this library for these compilers will also need to have Spirit 1.6 installed and have the SPIRIT_ROOT variable set accordingly. I don't know who I should inform about this.
That's not going to work well in the Boost regression suite. Tests get run against a current copy of the Boost tree, and we don't have any provision to check out (say) an old branch of Spirit.