Re: rel_ops issues

[Joe Buck <jbuck at racerx dot synopsys dot com>]

Ben writes:
> people can somebody please post
> 
> 1) testsuite additions that clearly cause the regressions that joe is seeing
> 2) patches for the same 

1) is in GNATS, reports 2405 and 2406, though the pages look garbled.

Test case for report 2405: compile

-----------------------------------------------------------------
#include <utility>
using std::rel_ops::operator!=;
#include <vector>
using std::vector;
void add_to(vector<int>& v, int x) { v.push_back(x);}
-----------------------------------------------------------------

Test case for report 2406:
----------------------------------------------------
#include <vector>
using std::vector;

struct foo {
    int a;
};

bool operator==(const foo&, const foo&);

bool veq(const vector<foo>& a, const vector<foo>& b)
{
    return a == b;
}
----------------------------------------------------

Because of legal problems I can't give you patches that have an
assignment, but I can describe one possible patches that I have
tested, it's pretty mechanical.

For 2405, we need to add additional partial specializations for
operator!= (and also <=, >=, and >) for __normal_iterator.
These partial specializations are for the case where both arguments
to operator!= are exactly the same.  This means that there are two,
rather than three, typename arguments.

For 2406, the problem is with std::equal's use of operator!= (the
three-argument version).

There are two possible fixes.  One is to add

	using std::rel_ops::operator!=;

to the body of the function, ignoring Gaby's anguished screams. :-)

The other is to replace the occurrences of != with ! and == .
Actually, only one of the two != operators needs changing, the
one that does not operate on iterators.

I had thought that there would be lots of other cases like 2406.
But it seems that std::equal might be the only one; the other uses
of != that I can find in stl_algo and stl_algobase operate only
on iterators, and for those, __normal_iterator should supply
a definition or else the iterator will be a pointer and have a
valid !=.  So I'm only asking for a fix to std::equal.