
On Mon, Jun 27, 2011 at 10:45:16AM +0100, Daniel James wrote:
On 27 June 2011 09:57, Lars Viklund
wrote: The test suite for the library fails miserably on VisualAge C++ 11.1, .. The test suite fails similarly on SunStudio 12.1 on Solaris 10, ..
It's unrealistic to expect a library to support such platforms before it has been added to the regression tests. Most of us don't have direct access to them.
It sounds like the library submission process is somewhat flawed in that regard. In my opinion, a library shouldn't be up for review until it's sufficiently Boostified. That would imply that it should be possible to graft it into a subset of test runners. This would provide some early warnings when finishing up the library, so that when it comes to review time, there's a remote chance of awesome compilers being supported, particularly as we have some vendor presence on the list. Sure, doing so would increase the load on our voluntary testers, but it might be worth the cost, at least for libraries that are in the process of getting into review. Looking at the set of supported compilers in the documentation for Assign V2, it only lists a few toolchains out of the many that Boost pretends to target. Even if it's not possible to have automated tests, at least there could've been some call for volunteers to spin off a couple of manual tests on slightly more non-mainstream compilers. Even if it fails miserably, it would provide valuable data points for both compiler vendors and developers. -- Lars Viklund | zao@acc.umu.se