Re: [boost] [guidelines] why template errors suck

Subject: Re: [boost] [guidelines] why template errors suck
From: Jeremy Maitin-Shepard (jeremy_at_[hidden])
Date: 2010-09-24 19:34:36

On 09/24/2010 09:00 AM, Eric Niebler wrote:
> This is borderline self-promotion and I fretted posting this. But we a
> community of advanced C++ library developers and I'd really like
> feedback about these issues. And they really do involve Boost library
> development. If we could reach agreement about these techniques, we
> might want to integrate them into Boost's coding guidelines.
>
> http://www.reddit.com/r/programming/comments/diddg/expressive_c_why_template_errors_suck_and_what/
>
> Comments?
>

Obviously actual timing data would be useful, but it seems that in some 
cases these techniques might result in increased compilation time, which 
is at least as serious a problem as the error messages.

Also, rather than use the technique mentioned at the end of the article 
for ensuring that compilation does not continue (and produce additional 
errors) after certain static assertions fail, at the cost of increased 
code complexity (and likely increased compile time), this purpose may be 
better served by a compiler option that causes compilation to abort 
after the first error (e.g. -Wfatal-errors in the case of gcc).