libtool, java woes

Jeff Sturm jsturm@one-point.com
Tue Apr 10 15:17:00 GMT 2001


On Tue, 10 Apr 2001, Mark Mitchell wrote:
>     >> We don't need or want libstdc++.
> 
> I think it's basically wrong to link with libsupc++ on any
> non-embedded system.  In the same way that we don't provide a separate
> malloc-and-free-but-not-popen C library, we shouldn't provide a
> separate new-and-delete-but-not-iostream C++ library.

But we're talking about java, not c++.  Because the two utilize each
other's ABI, isn't it helpful to share some implementation details?

Java users don't need iostreams, they have java.io.  They don't want to be
forced to link the entire libstdc++ runtime any more than c++ users would
want libgcj.  (Of course users could choose both, which is why Alexandre
is concerned about symbol duplication.)

For that matter I don't buy the argument that libsupc++ is a C++-only
thing.  Just as the gcj folks did, someone else could appear with a
completely new language frontend (c#?) and want transparent integration
with c++.

In any case, the use of libsupc++ is very much an internal build issue.  I
don't remember anyone suggesting it be installed.

Jeff




More information about the Java mailing list