[Bug objc/35996] ICE while building simple ObjC code with -fobjc-gc

[daekharel at gmail dot com]

--- Comment #2 from daekharel at gmail dot com  2008-05-16 02:57 ---
I'm not sure that's the case, as I've taken a look at the gcc libobjc code, and
it does seem to have some support for boehm GC integration. For example, a
fragment of code from gcc-4.2.3/libobjc/objects.c runs as follows:

#if OBJC_WITH_GC
  if (CLS_ISCLASS (class))
new = (id) GC_malloc_explicitly_typed (class->instance_size,
   class->gc_object_type);
#else
  if (CLS_ISCLASS (class))
new = (*_objc_object_alloc) (class);
#endif

Of course, perhaps that's just preliminary code for support in some future
version, and the vanilla gcc runtime doesn't support --enable-objc-gc. But if
that's so, there's still the question of why it ICEs on x86-64 but not on x86.
And also of the info page 7.3 regarding garbage collection in the GNU (not
NeXT) runtime.


-- 

daekharel at gmail dot com changed:

   What|Removed |Added

 CC|            |daekharel at gmail dot com


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=35996

[Bug libgcc/56460] _Unwind_Find_FDE is O(n) in the number of frame infos, (and LLVM's JIT will generate many of them)

[daekharel at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=56460

Michael Arntzenius  changed:

   What|Removed |Added

 CC||daekharel at gmail dot com

--- Comment #7 from Michael Arntzenius  ---
Hi, I work on the Pyston project (http://blog.pyston.org/) and we've been
having this issue precisely: we JIT a lot of functions, this makes throwing
exceptions really slow, and _Unwind_Find_FDE seems to be the culprit. Does it
seem likely this issue will be fixed anytime soon? Looks like the last activity
was in 2013.