[Second sending attempt after 6 days]

On 2011-03-04 10:15, Al Grant wrote:
>
> Standard C [7.13.2.1] is explicit that floating-point status flags
> must be preserved by longjmp() and not reverted to what they
> were at the time of setjmp().  Is there any equivalent guarantee
> in C++ about the status flags when an exception is thrown?
>
> (I realise<fenv.h>  wasn't around in time to be incorporated by
> reference in C++98, but access to the FP environment is pretty
> fundamental to numeric programming and I would have hoped
> that some time in the last decade someone would have said
> whatever needed to be said to do it predictably in C++.)

There is no such guarantee, but there is a guarantee that the
floating-point environment is hold per thread, see [cfenv.syn]
p. 3+4:

3 The floating-point environment has thread storage duration (3.7.2).
The initial state for a thread   s floatingpoint environment is the
state of the floating-point environment of the thread that constructs
the corresponding std::thread object (30.3.1) at the time it
constructed the object. [ Note: That is, the child thread gets the
floating-point state of the parent thread at the time of the child   s
creation.    end note ]

4 A separate floating-point environment shall be maintained for each
thread. Each function accesses the environment corresponding to its
calling thread.

I don't see any reason why exceptions should have a special rule as
you refer to. I don't think that the *jmp functions can be directly
compared to exceptions in this regard. Could you elaborate which
problems you are envisioning if exceptions miss to do that? A code
example would be helpful.

Thanks & Greetings from Bremen,

Daniel Kr   gler


--
[ comp.std.c++ is moderated.  To submit articles, try posting with your ]
[ newsreader.  If that fails, use mailto:std-cpp-submit@vandevoorde.com ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://www.comeaucomputing.com/csc/faq.html                      ]

Standard C [7.13.2.1] is explicit that floating-point status flags
must be preserved by longjmp() and not reverted to what they
were at the time of setjmp().  Is there any equivalent guarantee
in C++ about the status flags when an exception is thrown?

(I realise <fenv.h> wasn't around in time to be incorporated by
reference in C++98, but access to the FP environment is pretty
fundamental to numeric programming and I would have hoped
that some time in the last decade someone would have said
whatever needed to be said to do it predictably in C++.)


--
[ comp.std.c++ is moderated.  To submit articles, try posting with your ]
[ newsreader.  If that fails, use mailto:std-cpp-submit@vandevoorde.com ]
[              --- Please see the FAQ before posting. ---               ]
[ FAQ: http://www.comeaucomputing.com/csc/faq.html                      ]