Re: [boost] Review: Config system

From: Ross Smith (ross.s_at_[hidden])
Date: 2001-08-20 21:17:46

I vote that the config system be accepted into Boost.

I have only a couple of minor nitpicks:

(1) The macro BOOST_NO_INTEGRAL_INT64_T is listed under "Macros that
describe defects", i.e. macros that are only defined if a feature of the
C++ standard is missing. int64_t is not part of the C++ standard (it's
part of the C99 standard, I think), so this macro doesn't belong here.
It's listed again further down, under "Boost informational macros", so I
suspect its inclusion in the earlier section is just a cut-and-paste
error rather than a misunderstanding of the standard.

(2) I brought this one up before: BOOST_NO_STRINGSTREAM is defined for
GCC 2.95.3, which does have stringstreams (not 100% conforming, but no
worse than the rest of its iostream library, which seems to be
acceptable to Boost). The problem here is that previous 2.95.x versions
didn't have them, and (before 3.0) GCC provided no way to detect the
third part of the version number.

There doesn't seem to be any "under the counter" way of detecting the
difference, either. I have both 2.95.2 and 2.95.3 installed on this box,
and some work with diff on their respective include directories shows
nothing that could be detected by the preprocessor (apart from
<sstream>, the only changes are a few small tweaks to deque and rope).

I suspect that most people who were using 2.95.[0-2] have now updated to
2.95.3 or 3.x, or could do so painlessly. I suggest that it's reasonable
now to define BOOST_NO_STRINGSTREAM only for GCC <2.95, and ask anybody
still using 2.95.[0-2] to upgrade, rather than unnecessarily crippling
the library when used with 2.95.3.

(It would be different if 3.x was an acceptable replacement for 2.95.3;
then I think it would be reasonable for 2.95.3 users to accept the
upgrade-or-lose-functionality burden. But it's clear from the GCC
mailing list that 3.0.[01]'s C++ performance is still very inferior to
2.95.3 and unlikely to improve before 3.1 at the earliest, so I don't
think that's currently a practical approach.)

-- 
Ross Smith <ross.s_at_[hidden]> The Internet Group, Auckland, New Zealand
========================================================================
"Unix has always lurked provocatively in the background of the operating
system wars, like the Russian Army."                  -- Neal Stephenson