From: David Abrahams (dave_at_[hidden])
Date: 2006-05-12 18:01:34
"Arkadiy Vertleyb" <vertleyb_at_[hidden]> writes:
> "David Abrahams" <dave_at_[hidden]> wrote
>> >> Furthermore, it's _my_ library that needs result_of, so basically I
>> >> would be telling users, "if you get a cryptic error message that
>> >> looks like ___________, find some of the types in it and register
>> >> them with this cryptic macro."
>> > The error messages re lack of registration are not that bad IMO.
>> Bad for you and bad for a user are two totally distinct things.
> Just a couple more questions:
> 1) Do you consider typeof error messages more cryptic than error messages
> produced by other boost libraries? Or STL? Is there such a thing as
> (portable) non-cryptic error message in template programming?
Irrelevant. The difference between a cryptic error message and *no*
error message is dramatic.
> 2) What is *cryptic* about registration macros?
To the user, they have no obvious relationship to the user's immediate
> Also note that the user doesn't have to "find" the type in the message --
> it's right there, saying something like TypeNotRegistered=blah<...; and
> it's always about user's own type if lambda types are indeed registered (if
> we are talking about lambda). And it seems to me that it's much easier to
> register lambda types than implement what you propose.
Maybe so. I'm beginning to think that I may have overestimated the
amount of registration that would be required here.
-- Dave Abrahams Boost Consulting www.boost-consulting.com
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk