C++, libstdc++-v3 and, well, error messages

[Gabriel Dos_Reis]

Gerald Pfeifer <pfeifer@dbai.tuwien.ac.at> writes:

[...]

| > Furthermore, the Subject: field seems to imply that the problem you
| > are reporting is related to libstdc++-v3.  The problem has nothing to
| > do with V3, it is entirely front-end specific as exemplified by this:
| 
| Sorry, I didn't mean to finger-point the libstdc++-v3 folks (which
| have done a great job).

Well, apologies if my reply sounded harsh.

| What I am worried about is that due to the switch to a libstdc++ that
| enforces proper namespace treatment -- Which is really a good thing! --
| lots of users will get exposed to this problem with our C++ frontend.

I understand.

| That is, while this is not a problem with libstdc++-v3 at all, I'd expect
| that most of our users will encounter this in the context of *using* this
| library.

Yes, I agree with your sentiment.  Maybe Mark would like to investigate
that area :-)

| > But, I agree with you that the diagnostic message could (or has to) be
| > improved.
| 
| Perhaps there might be a hack that works most of the time?

Certainly.  But there is another way to see it: it is a regression with
respect to gcc-2.95.2 and as such some actions need to taken in that
area :-)

| ... Unless that's
| feasible, we should put a BIG note into the GCC 3.0 release announcement
| concerning "using namespace std".

This is certainly a good idea anyway, irrespectively of the
unfriendly diagnostic message.

-- Gaby