Hi Steven, all,

I have just noticed an issue in Faber, and I'm wondering how you solved this in b2:

A target (bound) filename is generated using the target properties / features (I never quite figured out the nomenclature you use), at least the non-incidental ones. So, part of a library's filename will be a subdirectory path encoding whether the library is shared or static. Fine so far. Now consider two object files, `lib.o`, being an intermediate file that's being linked (or archived) into the library. The other, `main.o`, being an intermediate file being compiled into an executable, which is *linked* with the library.

How are the bound filenames of those object files computed, and in particular, what properties / features are participating, to make the filenames unambiguous ?

While I can see how the (essential) properties of a library percolate to the constituent object files, and thus, generating the proper `.../shared/lib.o` and `.../static/lib.o` filenames. But I'm unsure about how to do that properly with `main.o`: What if it links to two libraries, one static, the other shared ? It seems quite hard to come up with a clear rule as to how to generate the filename in that case. How do you do that in b2 ?

Thanks,

-- 

      ...ich hab' noch einen Koffer in Berlin...