I believe what happened here was that the person who ran the test did not have the compiler specified in the user-config.jam, so the compiler was never called. A sample way to fix this problem, would be to have the first test be for compiling and running a "hello world" type of program. Perhaps test 1) would be compile and run "hello world". test 2) would be compile and run a "hello world" where cout << "hello world" is in a shared library. and test 3) could be compile and run "hello world" where cout << "hello world" is in a staticly linked library. If any of these test fail, then there is a misconfiguruation of the system, and all test can be expected to fail. Therefore, an error message should be printed to the screen, and regression.py, should halt execution, without reporting results. I discussed this with Valodia the other day in #boost. I don't know if he has done anything with the idea yet, so I wanted to put the idea out in the mailing list. Douglas Gregor <dgregor@cs.indiana.edu> wrote: Boost Regression test failures Report time: 2006-12-07T21:54:52Z This report lists all regression test failures on release platforms. Detailed report: http://engineering.meta-comm.com/boost-regression/CVS-RC_1_34_0/developer/is... The following platforms have a large number of failures: borland-5.8.2 1127 failures in 53 libraries (30 are from non-broken platforms) a --------------------------------- Any questions? Get answers on any topic at Yahoo! Answers. Try it now.