Re: [boost] [Report] 2 regressions on RC_1_34_0 (2007-04-14)

From: Fernando Cacciola (fernando_cacciola_at_[hidden])
Date: 2007-04-14 21:23:43

"Vladimir Prus" <ghost_at_[hidden]> wrote in message
news:E1Hcko5-0001yn-4W_at_zigzag.lvk.cs.msu.su...
> Douglas Gregor wrote:
>
> > |utility|
> >   none_test:  borland-5.8.2
>
> This appears to be releated to none.hpp changes, see
> http://tinyurl.com/yqta5d. What do we do about this?
>
This is the NEW test for "boost::none" that I added in the last commit
(which btw passes all other platforms)

The purpose of the test is to make sure that the expression "none" doesn't
convert to an integral type.
And this failure does not come from such a conversion. What happens is that
"none" also refuses to convert to optional here (or so, the compiler output
isn't in English so I had to guess from the code fragments in the message).

However, the testsuite for Boost.Optional, for this compiler, passes the
tests that involve "none", so is not that "none" 0doesn't work. I'm not sure
why it fails in *this* (new) test (and I don't have the compiler to check
for myself).

I suspect that the problem is that the reciever function takes the argument
by const reference, so it requires a temporary. But this is not strictly
neccessary for testing purposes, so I'll change it.

Best

Fernando Cacciola