Boost logo

Boost :

From: David Abrahams (dave_at_[hidden])
Date: 2004-01-21 15:44:25


"Pavel Vozenilek" <pavel_vozenilek_at_[hidden]> writes:

> "David Abrahams" <dave_at_[hidden]> wrote in
>
> [snip regression of is_class<> on Intel]
>
>> The way to debug problems like this is to do a binary search in the
>> CVS state for the time when the problem was introduced. Can you do
>> that?
>>
> Unfortunately I cannot. I am not friend of CVS and have problems even with
> TortoiseCVS wrapper.

This is really trivial to do. At the root of your Boost CVS tree you
do:

  cvs up -D2003/12/1

for example, to get the source state on 2003/12/1.

For the particular test you're interested in, it's almost certainly
sufficient to do it in the boost/type_traits/ subdirectory, or at
least in boost/. That will speed updates.

> The is_class<> implementations in 1.31.RC1 and 1.30.2 are very different and
> probably the massive change caused regresion.

But the "massive change", as you call it, was made in at least 5 steps
in that file alone, and might really be caused by other files.

-- 
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