Port-playstation2 archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]

Re: I'm getting better!

On Thu, Apr 03, 2014 at 05:14:14PM -0500, John D. Baker wrote:

> .if "${MACHINE}" == "playstation2"
> .endif

Actually what I did so far is slightly different:

I just used build.sh with -V HAVE_GCC=48 for target playation2 to build
all of userland (some sysinst changes not yet committed), which went as
far as when it tries to build the kernel - which was expected.

The R5900 should not need a special toolchain for userland code (maybe besides
the X server).

> The mechanism that evaluates "EXTERNAL_TOOLCHAIN" assumes that the tools
> reside in a "bin" subdirectory under the path value of the variable, so
> one does not need to specify a longer path.
> The cross tools are installed with names in the pattern:
>   mips--netbsdelf-*
> but the build system expects the tools to be named in the pattern:
>   mipsel--netbsd-*

Yeah, I was not sure what to do about that one - the toolchain supports
both endianesses, but we could easily make it configure for le only and
simplify this.

However, overriding MACHINE_GNU_PLATFORM for playstation2 should be another
option (but it did not work well so far for me).

Next important step is to build the kernel. I used a different aproach:
configure the kernel, go to the build directory, and then invoke
the shell script playstation2/conf/build.playstation2.sh instead of make
to build it. This uses TOOLCHAIN_MISSING instead of EXTERNAL_TOOLCHAIN
and supplies various tools directly.

I got the kernel libraries build, but it fails as soon a mkdep is invoked
on one of the main kernel .S files, due to CC not being passed correctly
down to mkdep.

At that point I stopped and called for help from a toolchain guru.

Once the kernel builds and boots, we can streamline the overall process,
hopefully finding some easy variation of EXTERNAL_TOOLCHAIN but still
building gcc 4.8 for userland.


Home | Main Index | Thread Index | Old Index