Re: [boost] [mpl] gcc ICE in aux_/has_tag<>

From: Aleksey Gurtovoy (agurtovoy_at_[hidden])
Date: 2006-11-10 23:29:05

Eric Niebler writes:
> Any chance one of the MPL gurus could have a look at 
> http://gcc.gnu.org/bugzilla/show_bug.cgi?id=29653 and see if any 
> possible changes to mpl/aux_/has_tag<> could make gcc happy?

This is basically another side of yet another GCC 4.1.x regression:
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=28088 (which is now
referenced from your bug report).

I'm pretty sure I could come up with a workaround if we really, really
want to see this work, but: it's already an intricate area in the
library, specifically under GCC and _because_ of its non-conforming
behavior that got broken in 4.1, so the cost of finding and
implementing a workaround that doesn't break things is going to be
pretty high. So, until somebody convinces me otherwise, I'd say this
is a good enough answer (from http://tinyurl.com/yccfau):

  [ Caleb Epstein ] This is a regression in the gcc 4.1 series that
  will be fixed in gcc 4.1.2. See bug #28088 for details.

> Preprocessed source code is attached to the bug report. FWIW, this
> is against Boost HEAD.

It surfaces in one of the basic MPL tests in RC_1_34_0 as well (see
the above link).

-- 
Aleksey Gurtovoy
MetaCommunications Engineering