This is the mail archive of the libstdc++@gcc.gnu.org mailing list for the libstdc++ project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: RFC: Allow moved-from strings to be non-empty


On 12/21/18 9:42 PM, Jonathan Wakely wrote:
On 30/10/18 07:48 +0100, François Dumont wrote:
On 10/27/18 12:33 AM, Ville Voutilainen wrote:
On Sat, 27 Oct 2018 at 01:27, Joe Buck <joe.buck@synopsys.com> wrote:
The reason move constructors were introduced was to speed up code in cases where an object Is copied and the copy is no longer needed.  It is unfortunate that there may now be code out there that relies on accidental properties of library implementations.  It would be best if the Implementation is not constrained.  Unless the standard mandates that, after a string is moved, the string is empty, the user should only be able to assume that it is in some consistent but
unspecified state.  Otherwise we pay a performance penalty forever.

If the standard explicitly states that the argument to the move constructor is defined to be
empty after the call, we're stuck.
It certainly doesn't specify so for an SSO string, so we're not stuck.
On the other hand, we already get
a speed-up, it's just not as much of a speed-up as it can be. What I
really loathe is the potential implementation
divergence; it's all good for the priesthood to refer to the standard
and say "you shouldn't have done that", but
that's, as a good friend of mine provided as a phrasing on a different
manner, spectacularly unhelpful.

+1 to allow moved-from strings to be non-empty

In many cases move will take place on temporary instances which future is just to be destroyed.

It's already too bad that in such situation we sometimes do some allocation because current implementation do not support none (std::deque)

Note that on vector we already have this undefined behavior within libstdc++ itself. On std::vector move constructor empty the moved instance but allocator extended one with equal allocator instance don't, it just swap content.

But the content of the new object being constructed is empty, because
it's a new object being constructed. It doesn't have any elements. So
if you swap the content, the moved instance becomes empty. Or have I
misunderstood what you're saying?

I am even misunderstanding myself ! Sorry, wrong analogy indeed.


I'll add to my TODO to detect this kind of invalid usage in Debug pendantic mode.

I'm not sure we want to do that; code relying on it might be
non-portable, but it's not undefined.

Yes, pedantic mode is not meant for that. We would need an option to detect portability issue for behaviors that are not described by the Standard but if it is limited to this situation it might not worth it. This is why I consider using pedantic mode for it in the first place I think.

François



Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]