Re: slices ; Stroustrup & libstdc++

[<llewelly at dbritsch dot dsl dot xmission dot com>]

On Sat, 17 Jun 2000 llewelly@dbritsch.dsl.xmission.com wrote:

> On Sat, 17 Jun 2000, brent verner wrote:
> 
> > on Fri, Jun 16, 2000 at 12:59:00PM +0200, Gerhard Wesp typed aloud:
> > |   hi all,
> > | 
> > |   i just tried a slice_array example from stroustrup, 3rd ed., p. 671
> > | which fails with the following error:
> > | 
> > | grisu ~/c++% cat slice.cc 
> > | #include <valarray>
> > | 
> > | void f(std::valarray<double>& d)
> > | {
> > |   std::slice_array<double>& v_even = d[std::slice(0,d.size()/2,2)];

A bit more info: 
26.3.2.4 has:

  slice_array<T> operator[](slice);

and since temporaries cannot be assigned to non-const references, this
  code seems to be wrong.

> > | }
> > 
> > I'm seeing the same error. I've spent the evening digging around
> > valarray & friends. I can't even decide if the code above _should_
> > work or not -- have you found out anything more? Would someone 
> > enlighten us? I've compared the code to the standard, and everything
> > _looks_ OK (by the std).
> > 
> > FWIW, the following code _does_ work.
> > 
> > #include <valarray>
> > #include <iostream>
> > 
> > void 
> > f(std::valarray<double>& d)
> > {
> >   std::slice oh(0,d.size()/2,2);
> >   d[oh] = 100;
> >   // d[oh] += d; // uncomment this to see its effect... i'm not
> >                  // sure it is correct. :\ .
> > }
> > 
> > int
> > main()
> > {
> >   std::valarray<double> a(10);
> >   f(a);
> >   for(unsigned i = 0; i < a.size(); ++i)
> >     std::cout << "a[" << i << "] = " << a[i] << std::endl;
> >   return 0;
> > }
> > 
> > | grisu ~/c++% g++ -fhonor-std -c slice.cc
> > | slice.cc: In function `void f(std::valarray<double> &)':
> > | slice.cc:5: initialization of non-const reference type `class std::slice_array<double> &'
> > | slice.cc:5: from rvalue of type `std::slice_array<double>'
> > | grisu ~/c++% g++ -v
> > | Reading specs from /usr/local/lib/gcc-lib/i586-pc-linux-gnu/2.95.2/specs
> > | gcc version 2.95.2 19991024 (release)
> > | 
> > |   is stroustrup wrong, or is libstdc++ (i'm using version
> 
> It might be worthwhile to check the errata for 3rd edition, at
>   www.research.att.com/~bs/3rd_errata.html ; there are several errors
>   mentioned on pgs 670-671.
> 
> unfortunately, I cannot find my copy of 3rd edition, and the errata are
>   hard to read without one; I cannot tell what might be relevant.
> 
> > | libstdc++.so.3.0.0) wrong, or am i missing something?
> > | 
> > | greetings,
> > | -gerhard
> > 
> > 
> 
>