Re: [Boost-users] [multiprecision] noexcept() usage in gmp int

Subject: Re: [Boost-users] [multiprecision] noexcept() usage in gmp int
From: John Maddock (john_at_[hidden])
Date: 2013-06-14 08:41:29

> I was trying out the new multiprecision library (kudos!), and I ran into a
> problem. In order to make certain guarantees in my code regarding 
> exception
> safety, I am mandating that types being used in generic containers are
> marked noexcept() for certain basic operations (namely, destruction and
> move operations).
>
> It seems like this is generally guaranteed in the multiprecision library
> (tried with cpp_int and the mpfr adaptor), with the important exception of
> the gmp integer class. A glance at the headers reveals that (gmp.hpp):
>
> //
> // Within this file, the only functions we mark as noexcept are those that
> manipulate
> // (but don't create) an mpf_t.  All other types may allocate at pretty
> much any time
> // via a user-supplied allocator, and therefore throw.
> //
>
> Additionally, browsing the gmp documentation yields this snippet (
> http://gmplib.org/manual/Custom-Allocation.html):
>
> """
> There's currently no defined way for the allocation functions to recover
> from an error such as out of memory, they must terminate program 
> execution.
> A longjmp or throwing a C++ exception will have undefined results. This 
> may
> change in the future.
> """
>
> Would it be possible then for the mp library authors to consider adding
> basic noexcept() guarantees to the gmp adaptor class in future versions?

Looks like an oversight to me, added BOOST_NOEXCEPT to mpz_int 
destruct/move/swap operations in Trunk.

On a related note, these noexcept specifications seem to be untestable to 
me, am I missing something?

John.