Questions, powerpc/powerpc64 clang 3.8.0/libc++ buildworld context: which _Unwind_RaiseException implementation(s)?

[Mark Millard]

Anyone know for clang 3.8.0 or devel/powerpc64-gcc based and WITH_LIBCPLUSPLUS= (use of libcxxrt and libc++ for the system) which _Unwind_RaiseException implementation is supposed to be used for TARGET_ARCH=powerpc? TARGET_ARCH=powerpc64? Alternatives around include:

/usr/src/contrib/llvm/projects/libunwind/src/Unwind-EHABI.cpp having:

> _Unwind_RaiseException(_Unwind_Exception *exception_object) {

/usr/src/contrib/llvm/projects/libunwind/src/UnwindLevel1.c having:

> _Unwind_RaiseException(_Unwind_Exception *exception_object) {

/usr/src/contrib/gcc/unwind.inc having:

> _Unwind_RaiseException(struct _Unwind_Exception *exc)

Should contrib/libcxxrt/exception.cc in a clang 3.8.0/libc++ or devel/powerpc64-gcc based buildworld context ever use contrib/gcc/unwind.inc's _Unwind_RaiseException implementation?

For both TARGET_ARCH=powerpc (clang 3.8.0 based buildworld) and TARGET_ARCH=powerpc64 (devel/powerpc64-gccb based buildworld) I currently get the contrib/gcc/unwind.inc binding.



But c++ exceptions fail to work for both contexts:

TARGET_ARCH=powerpc gets a SEGV in _Unwind_GetGR
(contrib/gcc/unwind-dw2.c:177).

TARGET_ARCH=powerpc64 loops without bound in _Unwind_RaiseException
(in contrib/gcc/unwind.inc).



For TARGET_ARCH=powerpc using a different compiler than clang 3.8.0:

> # g++5 -I /usr/include/c++/v1/ -L /usr/lib/ -g -Wall -pedantic exception_test.cpp
> or
> # g++5 -g -Wall -pedantic exception_test.cpp
> or
> # g++49 -g -Wall -pedantic exception_test.cpp
> or
> # g++49 -I /usr/include/c++/v1/ -L /usr/lib/ -g -Wall -pedantic exception_test.cpp

(Note the lack of -Wl,-rpath=/usr/local/lib/gcc49 or -Wl,-rpath=/usr/local/lib/gcc5 use.)

I get the same SEGV at the same place (return statement in _unwind_GetGR) despite the use of _Unwind_Resume_or_Rethrow being the caller of _Unwind_RaiseException (backtrace omitted here):

> # ./a.out
> terminate called after throwing an instance of 'std::exception'
> Segmentation fault (core dumped)


By contrast the following do not fail:

> # g++49 -Wl,-rpath=/usr/local/lib/gcc49 -g -Wall -pedantic exception_test.cpp
> # ./a.out
> # g++5 -Wl,-rpath=/usr/local/lib/gcc5 -g -Wall -pedantic exception_test.cpp
> # ./a.out
> # 



For TARGET_ARCH=powerpc64 (devel/powerpc64-gcc is already gcc5 based but has no /usr/local/lib/gcc5/ so just using lang/gcc49):

> # g++49 -I /usr/include/c++/v1/ -L /usr/lib/ -g -Wall -pedantic exception_test.cpp

I get the same "loops without bound in _Unwind_RaiseException" as with clang++ and devel/powerpc64-gcc.

But for:

> # g++49 -Wl,-rpath=/usr/local/lib/gcc49 -g -Wall -pedantic exception_test.cpp


the ./a.out works fine for the powerpc64 context.



All this suggests problems in contrib/gcc/ for stack unwinding for powerpc and for powerpc64 for each of the g++49, g++5, and clang 3.8.0 compilers.

Some sort of exception-ABI difference(s) with gcc 4.2.1 for each of the 2 TARGET_ARCH's and more modern compilers?

Note that for powerpc64 buildworld was done by devel/powerpc64-gcc, not by clang 3.8.0, vastly limiting clang's involvement. Yet a problem still exists.


The following 8 line program is enough to show the behavior:

> #include <exception>
> 
> int main(void)
> {
>     try { throw std::exception(); }
>     catch (std::exception& e) {} // same result without &
>     return 0;
> }

But the original problem was found and isolated in kyua and other programs before generating the simple example.

===
Mark Millard
markmi at dsl-only.net