Re: [boost] type_traits bugs detected by CWPro8?

From: David Abrahams (david.abrahams_at_[hidden])
Date: 2002-04-16 06:25:30

----- Original Message -----
From: "John Maddock" <john_maddock_at_[hidden]>

> > 201 tests completed, 21 failures found, 19 failures expected from
this
> > compiler.
>
> I would "expect" all those failures from MWCW:

In that case, can you update the "expected failures numbers" for this
compiler so that it doesn't squawk?
I'm also really interested in answers to my other questions about
testing and maintenance methodology. I have made *lots* of fixes to type
traits in the past few months but I want to be sure I'm doing the right
thing.

> these are either failures
> from "we don't know how to implement this without compiler help" - for
> example the is_POD, has_trivialXXX and has_nothrowXXX templates

Hmm, in theory we could use the MSL lib for some of those, if we cared.

> or else
> failures from is_convertible, and hence is_enum and hence anything
dependent
> upon that: has_trivial_assign<empty_UDT>, is_base_and_derived<T>,
> is_empty<enum_UDT> etc.

Understood.

<snip errors>

> Looks like is_class<function_type>::value is true (wrongly), and hence
> is_empty fails to compile for that type - everything else then
cascades off
> it.

Well, things are back to normal for some definition of normal, now. The
big issues left for you are:
1. Config macros
2. Testing/maintenance methodology
3. Header reorganization

-Dave