Re: Internal compiler error

["Martin v. Loewis" <martin@mira.isdn.cs.tu-berlin.de>]

> I have to report the following problem. We are now replace the gcc 2.7.2
> with egcs.

When you do that, I strongly recommend to use gcc 2.95. There is a
good chance that the internal error has been fixed.

> hv.C: In method `unsigned char hybridViewer::cube(int, int, int)':
> hv.C:1426: warning: sorry: semantics of inline function static data `int
> cbx[8]' are wrong (you'll wind up with multiple copies)
> hv.C:1426: Internal compiler error.
> 
> The gcc-2.7.2 had no problems.

If you mean it didn't crahs - perhaps (you'll know better than I do
:-) If you mean that gcc 2.7.2 did the right thing with respect to the
warning - I doubt it. This is a subtle error: If all invocations of
the inline function are in one source file, the program will still
behave correctly.

It may be that on your system, g++ can indeed deal with such data
correctly, in which case it would be a bug to be sorry. I don't know
what the status on OSF1 is.

Unfortunately, you did not provide a complete test case for somebody
to reproduce that error.

Regards,
Martin