Subject: Re: Safe Integer Types (P0228R0)
From: John McFarlane (john_at_[hidden])
Date: 2019-08-03 06:28:24
P0554 mentions an run-time-only overflow handler, `overflow_integer`
(similar to P0228's type) and a compile-time overflow avoider,
`elastic_integer`. They ought to find their way into a numeric TS, the
beginnings of which -- fingers-crossed -- will be in the post-Cologne
Christopher: the unique_function paper is a typo, that's actually P0288.
Use https://wg21.link/p0228r0 to get Robert's paper.
On Sat, 3 Aug 2019 at 05:49, Federico Kircheis via Std-Discussion <
> On 02.08.19 20:11, David Svoboda via Std-Discussion wrote:
> > Hello, I am curious as to the status of the document P0228R0 in C++?
> > http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2016/p0228r0.pdf
> > It was written in Feb 2016, but I can't seem to access any WG21 meeting
> > minutes newer than that.
> > More to the point, I am wondering if there are any proposals for safe
> > integer operations, and if so, what is their status? Did they get
> > adopted into a TS? They didn't seem to make it into C++17 or 20.
> I've chatted with Robert Ramey a couple of years ago about it, this is
> what he wrote me (in case he is not in the mailing list):
> I made a presentation to the SG-6 committee via skype in February 2016.
> The response was not negative, but not enthusiastic either. I concluded
> that getting something like safe numerics into the standard would be a
> lot of work, require attendance at many meetings, be extremely tedious
> and take several years. I've also concluded that for very complex
> libraries such as this one, the standards track doesn't work well. I
> concluded that it wasn't worth it. When I was recently contacted by the
> committee about the proposal we discussed the above and they decided to
> put the proposal on "hold" which was fine by me.
> I redirected my efforts to getting safe numerics accepted as a boost
> library. It got 5 reviews - all positive - and so was accepted -
> subject to a long list of conditions. This list was basically what was
> needed to "finish" the library: fix bugs, refine concepts, add more
> tests, etc. etc. I've been working on this since the review - it's a
> lot of work. I thought the library was in very good shape. The review
> made it clear that I was wrong about this. On the good side, the
> changes are working out well and I'm optimistic that the library will be
> of high quality when merged into boost. I'm hoping this will occur
> before CPPCon in september.
> Then the real work begins. Most programmers do not believe that
> erroneous arithmetic results are a serious problem and that addressing
> them is not worth the hit on performance. I'm convinced otherwise of
> course so I will have to spend time evangelizing the usage of the safe
> numerics library.
> So, unless something has changed (discussion was in 2017) there is no
> further work, and that's why you could only find the minutes from Feb 2016.
> Std-Discussion mailing list
STD-DISCUSSION list run by herb.sutter at gmail.com
Older Archives on Google Groups