gcc-3.4/gcc-4.0 archive [Was: Re: amd64(gcc4): aptitude dumps core]
On Wed, Jan 12, 2005 at 11:44:33PM +0100, Harald Dunkel wrote:
> Peter Nelson wrote:
> |
> | The gcc-3.4 (well, I guess it's now 4.0 but still has the old name)
> | distrobution should be considered experimental. If you just want
> | something that works use the pure64 version.
>
> But the compiler used to build pure64 doesn't work for amd64.
> Thats why a lot of packages in pure64 have been built with
> gcc 3.4, too. The problem is: You never know.
No it's not. There are maybe 10 packages build using gcc-3.4.
We have less problems using gcc-3.3 then gcc-3.4. There are even
more packages available in the pure64 archive then in the gcc-3.4
archive.
Note that currently I will not even look at problems if you're
using the gcc-3.4 archive. I also hope that people don't go and
file bugs when using the gcc-3.4 archive since it's very likely
not a problem with the package but with the compiler.
Has anyone even read the branch status of gcc-4.0? It's at:
http://gcc.gnu.org/ml/gcc/2004-12/msg00958.html
To quote from it:
| I've spent the afternoon looking through Bugzilla. My conclusion is
| that we're making progress (down to 169 regressions open against 4.0),
| but we've still got quite a ways to go. There are a lot of
| optimization issues, and as usual, a lot of C++ issues. We've got bug
| reports about bad code generation in the Linux kernel, a variety of
| other wrong-code issues, and a lot of issues about inferior code generation.
To me that looked rather scary. And by the recent problems
reported here about it, I'd just don't touch gcc-4.0 at all until
it's atleast released. The debian gcc maintainers have put it in
experimental, and that's where it belongs. It shouldn't be used
to build unstable yet.
Kurt
Reply to: