RFA: Please review the new C++ demangler patch

H. J. Lu hjl@lucon.org
Thu Jun 26 18:57:00 GMT 2003


On Thu, Jun 26, 2003 at 01:41:04PM -0500, Benjamin Kosnik wrote:
> 
> >Think a developer who has to support different platforms or
> >a package which has no real maintainer, compatibility is very
> >important. We shouldn't break compatibility just because it
> >may make our lives easier.
> 
> I think there are problems no matter what is done. This includes doing
> nothing.
> 
> The idea proposed seems the best solution to me. Obviously, opinions
> differ. Perhaps there are other ideas: Joe had some others (one similar,
> the others less satisfactory.)

I am not against this approach. But I can image how users will react
when it happens.

> 
> >Are you suggesting libdemangle should depend on a C++ library?
> 
> If
> 
> libiberty
> libstdc++
> libdemangle
> 
> is the build order, and libstdc++ is built, then yes: libdemangle should
> use __cxa_demangle. 
> 
> If gcc is configured without C++ support, but demangling C++
> symbols is desired (comment withheld), then the current "C" libiberty
> demangling code could be used as a fall-back (ie cp-demangle).

I don't think it is a good idea. There are cases you want to use
demangler in C. I don't want to link ld with a C++ library. libdemangle
should be a standalone library with a C interface.


H.J.



More information about the Gcc mailing list