[LLVMdev] darwin dragon-egg build issues

Duncan Sands baldrick at free.fr
Sun Apr 11 06:25:53 PDT 2010

Hi Jack,

>     Yes, I was trying to use dragon-egg svn with llvm release-2.7 branch (because
> I was in the process of preparing llvm/llvm-gcc42 2.7 fink packages and had
> those handy). I'll try again with llvm svn. FYI, I have posted the --enable-plugin
> fix patch for darwin to the gcc-patches mailing list...
> http://gcc.gnu.org/ml/gcc-patches/2010-04/msg00516.html

yes I saw it - thanks for doing this.  Unfortunately I'm not competent to review

> I think we need one more darwin-specific FSF gcc patch for dragon-egg
> (as gcc 4.5.0 doesn't installed the required darwin-sections.def into
> /sw/lib/gcc4.5/lib/gcc/x86_64-apple-darwin10.3.0/4.5.0/plugin/include/config).
> I am still pondering how to properly do that in the Makefiles.
>    Have any prior dragon-egg gcc patches been committed and do you intend
> to post the i386_static.diff to gcc-patches?

Several dragonegg gcc patches have been previously committed.  I don't plan to
push i386_static.diff to gcc though, because (1) I think I can do without it
(though this will require quite a lot of work), and (2) I'm trying to avoid
pushing patches to gcc if they are only useful to dragonegg.

>                Jack
> ps I did notice the dragon-egg web page said that exception handling was
> broken but I did see some recent commits on that feature. What is the
> status of exception handling in current svn?

It is almost finished, and I expect it to be completely finished next week
sometime.  You will probably find that exception handling already works fine.
Currently things may go wrong in obscure circumstances at high optimization
levels.  The other problem is that if a C++ destructor throws an exception,
then std::terminate may not always be called.



More information about the llvm-dev mailing list