This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [PATCH] adds powerpc-*-freebsd? to mainline


Hi Stan,

> Yeah, like me.  I've been following this along without understanding
> all the details, but it astounds me that it would seem rational to pull
> a freebsd config file into every other rs6000 config.

I take no public position in this thread since I lack information.

> What's so bizarre about freebsd that it requires such treatment?

But there is one issue that should be understood in order to
understand why FreeBSD (e.g.) might look more bizarre in its needs
than Darwin (e.g.).  Within gcc/config, many/all CPU architectures
have a consistent way of doing things for ports that are based on that
one CPU.  The problem is that although internally consistent in how a
port adheres to the model of overriding standard settings, the models
are not always very consistent accross all CPU architectures.  Thus,
an operating system like FreeBSD that wants to share common non-CPU
related setup amongst various "CPU ports" encounters an issue each
time support for a new CPU is undertaken.  We have to rework our
CPU-neutral model (the current one was last built with some input from
RTH and others when FreeBSD expanded to include support for alpha in
addition to i386) and/or update the CPU configuration files to better
support our needs without breaking others.  I know David is willing to
adjust the FreeBSD CPU-neutral model to a degree (and I am willing to
help him) but anything that makes the various CPU models more similar
might be a good thing to do as well.

You may have noticed that many of David's huge configuration patches
are to undo header nesting in gcc/config header files.  It makes for
larger tm_file settings but it adds flexibility.  One need only run
``grep tm_file gcc/config.gcc'' to see that the granularity of headers
is one area where different CPU models differ greatly.

Regards,
Loren


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]