[boost] [tuple] regression with borland-5_6_4

From: Joaquín Mª López Muñoz (joaquin_at_[hidden])
Date: 2004-10-14 11:28:36

Hi, I've noted that tuple_test_bench is marked as a regression
for BCB 564: http://tinyurl.com/62plc

However, in the the following page listing Boost 1.31 issues,
this test seems to appear as a failure: http://tinyurl.com/6j68r

The problem seems to be related with BCB having a hard time
with specializations on const types.

So my questions are:

1. Is this really a regression or an expected failure as it seems
to be present back in Boost 1.31?
2. Scanning the list, some folks claim that the problem goes away with
BCB 5.6.4, yet what we have have does not indicate so. Anyone knows
the status of this problem wrt to the version of the Borland compiler?
3. tuple already has an alternative implementation for those compilers
without PTS, driven by BOOST_NO_TEMPLATE_PARTIAL_SPECIALIZATION
(line 26 of boost/tuple/tuple.hpp). Could someone with local access to
the Borland compiler try to apply the alternative implementation by
replacing that line with the following?

#if defined(BOOST_NO_TEMPLATE_PARTIAL_SPECIALIZATION)||
  defined(BOOST_BCB_PARTIAL_SPECIALIZATION_BUG)

(I don't know if BOOST_BCB_PARTIAL_SPECIALIZATION_BUG
really relates to this particular problem, but its name fits perfectly
well.)

Thank you,

Joaquín M López Muñoz
Telefónica, Investigación y Desarrollo