RE: [boost] VC warnings and robustness (was: any_function (callback) library)

From: Paul A. Bristow (pbristow_at_[hidden])
Date: 2001-04-12 06:31:20

> -----Original Message-----
> From: Ed Brey [mailto:brey_at_[hidden]]
> Sent: Wednesday, April 11, 2001 2:46 PM
> To: boost_at_[hidden]
> Subject: [boost] VC warnings and robustness (was: any_function
> (callback) library)
>
...
> Boost code, however, is written at a higher level, and ought not be
> treated as black magic.  It should compile without warning under the
> tightest reasonable warning level to prevent bugs, using notation where
> necessary to indicate that something like a loss of significant digits
> is truly a feature and not a bug (like using [sic] in text).

I strongly agree with this.

>
> There is a catch with this approach: When boost files include system
> headers that have not been previously included by the translation unit,
> any of those system headers can trigger non-bogus warnings.
> Unfortunately, the C++ standard headers shipped with VC and the STLport
> 4 headers do this.  Fortunately, the solution is simple: include such
> headers between the push and pop.

Newer versions of the Dinkumware STL are steadily reducing the number
of spurious warnings (at some considerable effort in more careful coding).
Hopefully their latest version will be shipped with MSVC version 7, if it
ever appears,
or perhaps SP6 if that appears.

This will reduce, perhaps negate, the need for the above methods of warning
suppression.

> Longer term, I think it would be nice to have boost be completely
> warning level friendly.

Agreed.

>
> If there is interest in this, I'd be happy to lay out a more detailed
> implementation, including the documented list of overzealous warnings
> and documentation of guidelines for library writers.

This would be useful meanwhile.
>
> Dr Paul A Bristow, hetp Chromatography
Prizet Farmhouse
Kendal, Cumbria
LA8 8AB      UK
[phone]
mailto:pbristow_at_[hidden]