On Tue, 19 Dec 2000, Michael R. Jinks wrote:
> Hi all, terribly sorry for bringing this up yet again, but it seems that
> when Linus Hisself says this:
>
> http://news.linuxprogramming.com/news_story.php3?ltsn=2000-12-14-008-05-CD
>
> ...we should maybe take notice? Or not?
The problem with Linus is that he doesn't do userland programming.
gcc 2.96 is not suitable for compiling 2.2.x kernels. We never claimed
anything else.
Someone who usually doesn't leave kernel space doesn't notice all the
improvements 2.96 has, especially in the C++ part.
As for the binary incompatibility part, this is true, BUT, it is true for
ANY version of gcc.
Ever since gcc started supporting C++, the ABI (binary interface) of C++
has changed with every major release: 2.7, 2.8, egcs 1.0, egcs 1.1, gcc
2.95, gcc 2.96, gcc 3.0.
Neither of them is binary compatible with any other one.
Yes, this is a problem, but it's not a problem specific to gcc 2.96 and
the only fix is waiting for gcc 3.0, which will finally (hopefully ;) )
stabilize the C++ ABI.
The ABI changes have been necessary because gcc's early c++ implementation
was incomplete, and it's still necessary to change some things for full
ISO compliance.
> BTW, we don't develop C or C++, we work in Java; I mention this because,
> while it may sidestep the direct gcc issue, Java has all kinds of its
> own nutty platform dependency issues
You'll definitely want to try whatever JDK you're using on a different box
first.
Since none of the major JDKs are free, we can't support them.
> when the cause is actually "broken libraries" in our Linux
> setup.
7 doesn't have any seriously broken libraries. There are a couple of bugs
in glibc and libstdc++, but they're fixed in the errata.
LLaP
bero
_______________________________________________
Redhat-list mailing list
[EMAIL PROTECTED]
https://listman.redhat.com/mailman/listinfo/redhat-list
