Follow-up question: Re: [lug] g++ question: dynamic casts?

D. Stimits stimits at
Thu Jan 31 10:27:20 MST 2002

"Scott A. Herod" wrote:
> Hello,
> Thanks for all of the responses about dynamic casts.  For the
> sake of curiosity, below is a bit of the stack trace for a sample app.
> #0  __strtol_internal (nptr=0x805e654 "\b\201\005\b´ò\017@", endptr=0x0,
>     base=134604372, group=134768600) at eval.c:36
> #1  0x40218598 in __user_type_info::dyncast (this=0x805e654, boff=0,
>     target=@0x805e654, objptr=0x80867d8, subtype=@0x805e66c,
> subptr=0x80867d8)
>    from /usr/lib/
> #2  0x4021a033 in __dynamic_cast_2 (
>     from=0x8055618 <itvGuiPanel type_info function>,
>     to=0x8055618 <itvGuiPanel type_info function>, boff=0,
> address=0x80867d8,
>     sub=0x8056a5c <itvGuiObject type_info function>, subptr=0x80867d8)
>    from /usr/lib/
> A follow-up question: What do people think of 3.0?  We're starting
> a new, big project and are discussing the build environment.  There
> is resistance to moving to gcc 3 for development until it stabilizes
> more.  Comments?

I would also resist it *if* the end users are going to be required to
upgrade their compiler. Despite all the versions and people suggesting
*their* code is fine with the version, try installing multiple versions
of your compiler in order to support every app that wants its own
particular version. If someone has to upgrade their compiler for this
software, and does not choose multiple version installs...imagine what
happens when they discover their is other software they can no longer
compile and they must now downgrade. It's a real mess if you want to
actually use a feature or behavior that would break an older
compiler...or if someone depends on their compiler to have an older
feature/behavior to compile another source project other than yours. If
you expect people to compile the source on FreeBSD or other UNIX-like
systems, it gets worse, often other systems do not go to newer gcc/g++
as soon as Linux does.

But of course if you are releasing binary packages, that isn't a

D. Stimits, stimits at

> Thanks,
> Scott
> _______________________________________________
> Web Page:
> Mailing List:

More information about the LUG mailing list