Re: C++ exceptions in C [was: Re: CVS commit: src/external/bsd/atf/dist/atf-c++]

[Julio Merino <jmmv%NetBSD.org@localhost>]

On Wed, Jul 7, 2010 at 4:51 PM, Joerg Sonnenberger
<joerg%britannica.bec.de@localhost> wrote:
> On Tue, Jul 06, 2010 at 07:21:27PM +0100, Julio Merino wrote:
>> Summarizing: what should happen when a C++ function calls a C function
>> that, in turn, calls another C++ function that throws exceptions?  Up
>> to where should exceptions propagate?  Intuition tells me that they
>> should never reach C land.  (Yeah, no comments on this bizarre
>> setup...)
>
> For the record I will repeat what I said on IRC. i386 and amd64 differ
> in one important aspect of the default code generation. For amd64
> setting up the frame pointer is optional disabled by default. To make it
> possible to do debugging, amd64 also has the .eh_frame section in that
> case, which makes it possible to unwind the stack. This section is also
> used by the C++ stack unwinding code when looking for an exception
> handler.

OK, thanks.

> Programs and libraries that mix C/C++ and want exceptions to "correctly"
> propagate across C functions have to make sure to compile the C code
> with -fexceptions.

The change I committed prevents from exceptions to propagate into C
altogether, which seems the safest thing to do.

-- 
Julio Merino