Re: C++ with gcc 3.0 branch

[Jo Dillon <jo at trolltech dot com>]

On Thursday 15 February 2001 11:36, Rob Taylor wrote:
> > On Thursday 15 February 2001 03:13, Alexandre Oliva expostulated:
> > > On Feb 14, 2001, Jo Dillon <jo@trolltech.com> wrote:
> > > > Mostly it works fine but at the link stage I get a lot of undefined
> > > > references to 'non-virtual thunk to <some member method>'. A small
> > > > test program I wrote with multiple inheritance, virtual methods and
> > > > so on failed to show any problem. I was wondering if this was
> > > > something yet to be resolved in the GCC sources or something that
> > > > might indicate bugs in my program?
> > >
> > > I don't know of any such problems.  It would be really nice if you
> > > could post a bug report that would let us duplicate the bug.
> >
> >   As I say I have problems isolating it to a test case, although I'll
> > keep trying. It does turn up with a released, freely available version of
> > Qt (which is a large GPLed library, about 400,000 lines - the
> > distribution includes various example programs that show the problem) -
> > would that be sufficient?
>
> which qt version are you compliling?

  Our internal build tree, but the same happens with Qt 2.2.4/Embedded
too (ftp://ftp.trolltech.com/qt/source/qt-embedded-2.2.4.tar.gz)

> what do you get if you do gcc -v?

  Reading specs from /usr/local/lib/gcc-lib/mipsel-linux/2.97/specs
Configured with: ../gcc/configure --target=mipsel-linux 
--program-prefix=mipsel-linux-
gcc version 2.97 20010212 (experimental)

(i.e. a just-checked-out copy of the 3.0 branch)

> I saw these problems a while ago (sometime round november/december) but
> haven't seen anyhting since. What flags is the qt build system passing into
> gcc?

  -pipe for compiling, nothing unusual for linking statically.
It's not a qt problem though as far as I can see, I reproduced it
with two files, about 60 lines of code and with no flags - see my GNAT 
report just added.

-- 

	Jo