[boost] Re: [MPL, TypeTraits] Absurd template nesting on GCC 2.95

From: Jonathan Turkanis (technews_at_[hidden])
Date: 2004-02-09 02:46:20

"Daniel Wallin" <dalwan01_at_[hidden]> wrote in message
news:40272F00.2040102_at_student.umu.se...
> Jonathan Turkanis wrote:
> > Hi all,
> >
> >     I'm working on some code which if naively implemented would
> > involve deep template nesting. Since I can't get a version using
MPL
> > algorithms to work on BCC 5.5.1 or VC6 , I wrote out the algorithm
> > with hand-coded loop unrolling.
>
> IMHO it would be better to try to get the MPL version to work rather
> than rolling your own.
>

I agree in general. After struggling a bit Borland and VC6 for a while
I decided to try a hand-writen version, which turned out to be very
simple to write -- it just took a few minutes. The workarounds for
Borland 5.6.4 and VC6 were straigforward; getting it to work on
Borland 5.5.1 and GCC 2.95 was the real pain. I've never used Borland
5.5.1 or GCC 2.95 before this week, so I'm not surprised it took me a
while to get them to work.

> What problems are you having with VC6?

About 1500 lines of errors relating to MPL internals which I don't
understand. The usual stuff: 'xxx has no member yyy', where xxx is
some undocumented template. I can usually get MPL code to work on VC6,
but not because I understand the error messages. ;-) I didn't really
try that hard to fix it today, because I remember spending a lot of
time on it several months ago.

The hand-written version seems to be quite fast and stable now. I may
ask you for help if I decide to work on it again.

I was just a bit surprised to find the enormous depth of nesting on
the old GGC.

Best Regards,
Jonathan