Re: [boost] Compiler choice

From: Jens Maurer (Jens.Maurer_at_[hidden])
Date: 2001-05-18 13:49:51

Paul Baxter wrote:
> 
> Apologies if this is a tad offtopic, but I wondered what the current
> thoughts are on a 'compliant' C++ compiler for the PC.

If "PC" includes other operating systems than Windows on PC-style hardware,
your choice is larger.

> Comeau (+ STLPort?)

Works nice for me.

> GCC 2.9x/3.0 (will use this as a backup 2nd compiler anyway after its had a
> shakedown for a few months)

I would recommend against gcc 2.9x for standard compliance checking at this
time.  gcc 3.0 should be much better in this regard.

> Whilst Boost may shy away from actually recommending a compiler, I think its
> time to start guiding people more towards the good implementations. Although
> the test results posted for Boost do help, they are coloured by a huge
> amount of workarounds for some compilers.

People interested in the amount of workaround for a particular compiler
should have a look at config.hpp and the number of #define's there. In
conjunction with the regression test results, this appears to be an
interesting measure of compliance.

In general, I've found the boost regression tests to exercise quite a lot
of aspects of a given compiler/library, often finding bugs in compilers or
libraries.
Of course, you always start with an empty config.hpp for the compiler.

Jens Maurer