GSoC: Making ports work with clang

Charlie Kester corky1951 at comcast.net
Sun May 2 22:29:54 UTC 2010


On Sun 02 May 2010 at 14:03:06 PDT Andrius Mork??nas wrote:
>On Sun, 02 May 2010 23:17:00 +0300, Eitan Adler <eitanadlerlist at gmail.com> wrote:
>>Good - and those 30% of ports will help improve clang++ even more.
>Some probably will, we submit a lot of bug reports for clang/llvm.
>
>>Hopefully over time that number will increase to 100% and we will be
>>able to say goodbye to gcc for good.
>That won't happen, at least not anytime soon and not until we get rid of
>[old] poorly written ports from the ports tree. Another problem is ports
>using horrible or less horrible GNU extensions for C or C++, clang will
>not support all of them. So we will still need gcc for some things, just
>like we need USE_GCC=whatever now, because some ports don't compile with
>gcc42 from base. I just hope we can get the majority of ports working
>with clang and keep the number of ports that need gcc as low as possible.
>

As things stand today, we don't know exactly which ports have the kind
of dependency on gcc that you describe. If this project gets us closer
to that list, it will have been worthwhile. 

Once we know which ports are unavoidably dependent on gcc, we can start
exploring alternatives to them.  More projects for GSOC and others
looking for ways to contribute!  Sounds like fun!


More information about the freebsd-hackers mailing list