llvm dev - May 2014 - [LLVMdev] Do the LLVM SJLJ intrinsics interact with C++ exception handling and execute C++ cleanup actions?

Do the LLVM Setjmp/Longjmp intrinsics interact with C++ exception handling and
execute C++ cleanup actions?

I compiled a small example using setjmp/longjmp using clang and they did not
execute C++ cleanup actions and calling longjmp bypassed calling destructors on
objects that were allocated between the call to setjmp and longjmp.

I’m wondering if the LLVM intrinsics behave differently from setjmp/longjmp.

Best,

.Chris.

Christian Schafmeister
Professor, Chemistry Department
Temple University

The intrinsics are completely distinct from the standard library functions. The
names are historical legacy from GCC. They’re used for 32-bit iOS exception
handling. I’m not aware of them being tested or supported in any other context.

-Jim

On May 27, 2014, at 1:26 PM, Christian Schafmeister <chris.schaf at
verizon.net> wrote:
> 
> 
> Do the LLVM Setjmp/Longjmp intrinsics interact with C++ exception handling
and execute C++ cleanup actions?
> 
> I compiled a small example using setjmp/longjmp using clang and they did
not execute C++ cleanup actions and calling longjmp bypassed calling destructors
on objects that were allocated between the call to setjmp and longjmp.
> 
> I’m wondering if the LLVM intrinsics behave differently from
setjmp/longjmp.
> 
> Best,
> 
> .Chris.
> 
> Christian Schafmeister
> Professor, Chemistry Department
> Temple University
> 
> 
> _______________________________________________
> LLVM Developers mailing list
> LLVMdev at cs.uiuc.edu         http://llvm.cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev

Thank you!

I was envisioning all sorts of wonderful uses for the SJLJ intrinsics to
implement non-local exits for the Common Lisp CATCH/THROW, BLOCK/RETURN-FROM and
TAGBODY/GO special operators that would work in a way that was compatible with
C++.    Currently I’m using C++ exception handling and it works fine but I have
been under the misapprehension from reading
http://llvm.org/docs/ExceptionHandling.html#setjmp-longjmp-exception-handling
that SJLJ intrinsics were a parallel mechanism to C++ exception handling that
offered different trade-offs in run-time cost/overhead.

I see now that SJLJ is a more primitive exception handling mechanism for a
subset 32-bit iOS.

Best,

.Chris.



On May 27, 2014, at 4:51 PM, Jim Grosbach <grosbach at apple.com> wrote:
> The intrinsics are completely distinct from the standard library functions.
The names are historical legacy from GCC. They’re used for 32-bit iOS exception
handling. I’m not aware of them being tested or supported in any other context.
> 
> -Jim
> 
> On May 27, 2014, at 1:26 PM, Christian Schafmeister <chris.schaf at
verizon.net> wrote:
> 
>> 
>> 
>> Do the LLVM Setjmp/Longjmp intrinsics interact with C++ exception
handling and execute C++ cleanup actions?
>> 
>> I compiled a small example using setjmp/longjmp using clang and they
did not execute C++ cleanup actions and calling longjmp bypassed calling
destructors on objects that were allocated between the call to setjmp and
longjmp.
>> 
>> I’m wondering if the LLVM intrinsics behave differently from
setjmp/longjmp.
>> 
>> Best,
>> 
>> .Chris.
>> 
>> Christian Schafmeister
>> Professor, Chemistry Department
>> Temple University
>> 
>> 
>> _______________________________________________
>> LLVM Developers mailing list
>> LLVMdev at cs.uiuc.edu         http://llvm.cs.uiuc.edu
>> http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev
>