Peter Dimov

If we drop the default value handling in `threadapi-feature.detect`, it would just return `foo` when `` is already `foo`, so there's not much point in having it at all, is...

Addendum: no, it doesn't: ``` error: No best alternative for ../../libs/thread/build/thread_sources next alternative: required properties: win32 multi not matched next alternative: required properties: pthread multi not matched ```

``` pthread windows:win32 ``` works though, except for ``` ************************************************************ Trying to build Boost.Thread with pthread support. If you need pthread you should specify the paths. You can specify them...

I suppose that `win32_pthread_paths` should be somehow using `message` to warn at build time instead of warning at configure time, but I have no idea how this would be expressed....

OK, I'll wait for your patch. At the moment I've been trying ``` rule detect ( properties * ) { local ps = [ property-set.create $(properties) ] ; local api...

I don't quite understand where does the default value of `threading` come from. Under Windows, when I say `b2 libs/system/build`, I get `threading=multi` by default. But on Linux, the default...

Ah so. You're right, it's threading=single for toolset=gcc, and for msvc when runtime-link=static. msvc-8.0 and above no longer has a single-threaded runtime of any kind anyway, so that code is...

But perhaps the correct place to put that default is in default-build of Jamroot?

I can't think of ever wanting a threading=single default build, but there could be scenarios that I'm not aware of. Re top level targets, that's part of another philosophical discussion...

I deduce from the fact that you haven't accepted my PR against type_erasure that you have reservations against this scheme. You are probably right, in a sense. This is not...