From: Vladimir Prus (ghost_at_[hidden])
Date: 2003-11-04 11:37:15
Robert Ramey wrote:
>>I think I've found out why all warchive tests failed with gcc 3.3. It
>>seems there were two bugs in codecvt_null implementation. Both was causing
>>it to return 'partial', when strictly speaking it should return 'ok'.
> In a million years I never would have found that !
Who knows? When I started looking, it was pretty obscure to me.
>>I don't know
>>why tests passed with earlier gcc version, maybe they had sloppier logic.
> I wonder if BOOST_NO_INTRINSIC_WCHAR_T is not set differently
> in our versions of gcc. I'm using gcc 3.2 under Cygwin.
It's unset for me. I though a likely cause was that earlier libstdc++ tried
to convert several times, and after receiving 'partial' decided that it's
>>look even better.
> very nice work. Its clear to me that things are progressing very rapidly
> on this front.
Well, since I'm planning to use serialization in my own project I'd like to
have all failures cleared up ASAP.
>>P.S. And BTW, if serialization was hosted in some version control system,
>>e.g. in boost sandbox, creation of all those patches would be much easier
> Up until now its been me basically working alone. As that's beginning to
> maybe this is a good idea. Right now tailing out things on the To Do list
> and rolling in changes from you and Pavel Vozenilek (Intel 7.0 and by
> implication VC 6.0).
> Maybe we'll look at this with draft #14. I've asked to be put into the
> review queue so this might affect things as well.
Let's see how it goes. Maybe, after fixing all gcc failures I can, I'll no
longer send any patch, so this will be non-issues for me.
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk