Re: [boost] Boost regression notification (2005-10-06 [RC_1_33_0])

From: Jonathan Wakely (cow_at_[hidden])
Date: 2005-10-08 13:35:31

(re-sending as the server that hosts this mail account rebooted as I
sent the mail and it hasn't shown up in the archives yet ...)

Martin Wille wrote:

> Arkadiy Vertleyb wrote:
> > "Douglas Gregor" <doug.gregor_at_[hidden]> wrote
> > 
> > 
> >>>Is GCC's debug
> >>>mode
> >>>a release-critical platform ?
> >>
> >>No.
> > 
> > 
> > How can this be?  How useful is a feature that can be used in the release
> > mode oinly?  I think its usefullness is pretty close to zero...

> I agree with your questioning, Arkadiy.

I completely fail to understand the question.  The debug mode cannot be
used in "release mode" because it makes the stdlib non-conforming.  It's
useful for tracking down problems in your code, and as such is immensely
useful, but it's not meant for production code in "release mode"

What do you mean "a feature that can be used in the release mode only" ?

> People will want to use the debug mode. It's probably the best thing 
> added to recent libstdc++ distributions. Severals bugs in Boost code 
> were discovered using _GLIBCXX_DEBUG. I'd consider it a big loss not to 
> be able to use it anymore.

I'm not saying "you can't use it anymore" - just that the Boost release
shouldn't be held up by debug-mode failures that don't affect any major
platforms in their normal configuration.

If you're willing, try running the tests with libstdc++ concept-checks
enabled as well (#define _GLIBCXX_CONCEPT_CHECKS) and then let me know
if you think the release should be held up for those bugs too.

Personally I think they're more important than the debug-mode failures,
since it's possible to configure GCC so that concept-checks are enabled
by default, making parts of Boost unusable.  But debug mode is always
off by default.

jon