Lars Viklund
By "tested", do you mean attempted to build, or actually ran tests? I'm going to assume the former, as the latter is near-impossible for the layman.
It's a bit disheartening to see so few attempts to run integrity tests, but I guess it's an artifact of it being underdocumented and hard to run tests and get output from them without being a proper regression test runner.
By "tested", I mean that in addition to building Boost, I have compiled some of our own projects and run their regression tests, which make heavy use of many Boost libraries. I do sometimes run the Boost regression test suite and have done so for a previous 1.56 build, but there are many expected failures that haven't been marked as such for newer MSVC versions as well as compiler crashes, so I often don't know which failures are critical.