Re: [boost] Re: Nuemric Conversion review

From: Beman Dawes (bdawes_at_[hidden])
Date: 2003-12-19 21:21:51

At 06:54 PM 12/19/2003, Andy Little wrote:

 >I am not sure what the policy on code correctness at boost is, but in my
 >book that means I would have to give an automatic
 >"reject" vote to any design that doesnt actually work. ( I do understand
 >that the problems are in dependent libs... but... cmon guys... get it
 >sorted for VC7.1 please! ) ...

I don't think we have an explicit policy for what compilers a library has 
to work on at time of formal review, and I don't think we want such a 
policy just yet. While we should only accept libraries that we think are 
portable, we should require actual testing unless we think there is a 
serious question. It is too much work to ask of submitters who don't know 
if their submission will be accepted.

But given that 100% of Boost's current libraries pass all tests on VC++ 7.1 
with few or no workarounds, I'd hate to see a library go into a release 
that wouldn't pass VC++ 7.1.

I expect both Metrowerks and Intel will also be passing 100% very soon with 
few or no workarounds needed. As each compiler gets to that level, I think 
we should start requiring that new libraries pass too. (Unless some new 
technique, like enable_if, is discovered that doesn't work, or something 
else exceptional.)

--Beman