Do you need the vtable bug fixed?

[Franz Sirl]

Am Thu, 03 Jun 1999 schrieb H.J. Lu:
>> 
>> As some of you know, there is an old bug in g++ involving virtual base
>> classes, and virtual function calls in constructors, on Linux
>> (i.e. with vtable thunks).
>> 
>> I have proposed a patch some months ago, and asked for inclusion of
>> this patch in gcc 2.95. This patch would fix the bug, at the expense
>> of binary compatibility: Code using the bug fix cannot be linked with
>> old code, anymore (in C++ applications where the bug potentially
>> occurs).
>> 
>> Therefore, this patch introduces the bug fix as an option: people that
>> need the bug fix can recompile everything with -fvtable-thunks=2. The
>> default would be the same as in egcs 1.1 (i.e vtable-thunks=1 on
>> Linux, no vtable-thunks elsewhere).
>> 
>> So far, the g++ maintainers have not decided whether to include that
>> fix or not. If people interested in that issue could speak up, maybe
>> that would help in finding a decision - if nobody is interested, this
>> would likely also help in finding a decision.
>> 
>
>I don't care too much binary incompatibility when I have
>the correct code. I WANT to recompile all my C++ programs
>so that I don't have to worry about when my C++ programs
>will break silently.

Hmm, is there C++ binary compatibility between egcs-1.1.2 and
gcc-2.95 anyway? Eg. I couldn't link gcc-2.95 compiled KDE stuff against a
egcs-1.1.2 compiled qt-1.44 library on powerpc-linux-gnu. If there's no
compatibility possible, someone should update the versioning stuff, otherwise
I'll prepare a bug report.

Franz.