Philip A. Naecker wrote:
After you built jam, were you able to build the entire boost library using jam?
I did not try. As you pointed out in your previous mail, VMS-specific changes should be done not only to build.jam, but to other jam files as well. Until this is done, there is no point in even trying to use jam on VMS, I believe. In 2004, I was trying to build and test boost on VMS (this task is currently not on my list, but it can reappear at some point). Instead of using jam, I was converting jam files to DCL scripts. As far as I remember, most of the times it was pretty straightforward. The easiest part was header-only libraries, of course, but even a library with a run-time binary component like date_time library was easy enough. I was running boost tests for a number of libraries and the results were on par with what you get on a supported platform. All this was under DCL. There is no reason VMS cannot be a supported platform for boost. Recently, Tru64 was certified as supported platform. AlphaVMS has the same C++ compiler and std. library that Tru64. VMS/IA64 has better std. library and the compiler which is as least as good as that on AlphaVMS. VMS C++ compilers are using the EDG front end and the C++ std. library is Rogue Wave library (v2.0 on AlphaVMS and v3.0 on VMS/IA64). The main problem on VMS is tools, like jam. And a toolset for VMS, of course. If you are interested in more details like compilation options or environmental settings I was using, please, let me know. If you are interested in porting automated boost regression test environment to VMS, please, let me know. We can arrange a machine for you and provide all necessary assistance. Thanks, Boris