Hi, I've downgraded the optimization level from -O2 to -O1 with Clang/LLVM as -O2 seems to generate non-working code, in particular in kernel/gc.hpp:gc_current_pool(). For some reason the memory pool is not set correctly in that function with -O2. I suspect some pointer aliasing issue confusing the optimizer (as if I split that 10-line function into several smaller functions using additional variables for intermediate results then the generated code works) but for now feel safer to go with -O1.
-- Jean-Sebastien On Tue, Sep 6, 2011 at 9:40 AM, Jean-Sebastien Delfino <jsdelf...@apache.org> wrote: > On Mon, Sep 5, 2011 at 10:24 PM, dsh <daniel.hais...@googlemail.com> wrote: >> Have to look at it during the weekend again or so. Didn't have much >> time recently cause I had to focus on other stuff. I think I will >> disable each broken dependency in the FreeBSD port and thus build >> Tuscany only with the proposed set of minimal dependency for now. >> >> Cheers >> Daniel > > Sounds good! > -- > Jean-Sebastien >
