> I'm porting the regression test system to Boost.Build V2 and
config_link_test
> causes some problems. The test does not run, because it can find shared
> library. This happens on Linux.
>
> It works with V1 because it set LD_LIBRARY_PATH before running the test.
> In V2, there's a choice. First alternative would be to use the same logic
as
> V1 uses. But on gcc, it's possible to use new <hardcode-dll-paths>
feature,
> which will append necessary -rpath options to compiler. In that case, the
> binary would run ok even if no environment variable is set.
>
> The question is: will adding -rpath options be ok? I actually don't
understand
> what the test does so can't judge myself.

The test simply verifies that dll building and linking work OK, it doesn't
really matter how the dll is found at runtime.

John.