John Maddock wrote:

I think the regressions for thread and wave libs in the 1.33.1 branch can be removed quite easily if the Jamfiles for these tests stopped using <runtime-link>static.

Anyone know if there was/is any rationale for this? The problem is that some platforms either don't have a static-runtime option, or it may be "optional". For example these tests all fail on my Mandriva Linux box because there are no static runtimes supplied. Other Linux versions may supply them, but as can be seen from the regression test results they sometimes cause more trouble that they are worth.

There is no real rational for this *blush*. The only reason I had to link the libraries statically was driven from the point of view of the MSVC environment, where the only way to link against a non-threaded lib is to use static linking. The aim was to emphasise, that the Wave applet does not use and does not need threads.

BTW, just because the Boost libs are linked statically this does not imply that the runtime must be statically linked (just in case that was causing any confusion).

I'll change it to use dynamic linking ASAP. Regards Hartmut

John Maddock wrote:

I think the regressions for thread and wave libs in the 1.33.1 branch can be removed quite easily if the Jamfiles for these tests stopped using <runtime-link>static.

<nitpicking> FWIW, this doesn't remove the regressions. It just avoids triggering them. </nitpicking> However, since this looks very much like a compiler problem and because this specific problem is *very* annoying (for the need of manual intervention during each test run) this fix is most desirable. Runners of incremental tests will need to remove the Thread/Wave test results and objects/binaries once the suggested fixes have been applied (already happened for Wave). Regards, m Send instant messages to your online friends http://au.messenger.yahoo.com