[MAJOR BUG] g++ 2.95.x generates invalid code for KDE 3.3.0

[Grzegorz Jaśkiewicz]

On Sun, 19 Sep 2004 12:51:50 -0300, Thiago Macieira
<thiago.macieira at kdemail.net> wrote:
> Guillaume Laurent wrote:
> >On Sunday 19 September 2004 15:24, Thiago Macieira wrote:
> >> - a constructor for one of the classes inheriting the virtual base calls
> >> one function which in turn calls an overridden virtual function
> >
> >Virtual functions do not work in ctors in general.
> 
> They work fine in g++ 3.3/3.4 and there's nothing in the standard that
> prevents us from doing that.
3.4.x is very stable for me, and produces much faster code for c++
apps than 2.x branch.

> g++ 2.95.4 on Linux is known now to produce invalid code. It should no
> longer be used. As a compromise till people upgrade their compilers (to a
> 2.95.5, for instance, if that ever comes along), we should provide code
> that doesn't trigger the bug for a while, but not for long.
> 
> As for its use on other platforms, I said "not supported" meaning that the
> user's problem of making it work, not ours.
I probably shouldn't ask that quesiton, but did anyone contact gcc ppl
acctualy ?
I am lurking gcc stuff acctively, and I remember ongoing discussin on
their list about 2.x branch, when 3.4 was released. I know also that
there is planty of bugs fixed for 2.x branch, but no new release was
made for looong time. Maybe we should persuade them to release 2.97 or
so version that will contain few more fixes. Possibly fix for our
problem ?
It's strange that bug doesn't appear on fbsd, as i386 code generator
is more-more-than-less the same for every system platform.

-- 
GJ