On Tue, 2012-02-28 at 15:25 +0100, Peter Gavin wrote: 
> Hello,
> 
> As you might know from seeing me around on IRC, I've been working on
> bringing the OpenRISC toolchain up to date with the most recent
> upstream versions.  I've also completed the work that Julius has been
> doing with binutils to make it use CGEN.  I've also changed the name
> of the architecture to or1k (from the old or32).  Things seem to be
> working pretty well, so I figured I'd do a pseudo-announcement on the
> list so people can give them a try.
> 
> There are git repositories for each component at
> <https://github.com/pgavin/>.  I've been able to get the linux kernel
> compiled and running in or1ksim with this toolchain, and if you want
> to try it for yourself, please let me know about any trouble you have
> so I can try to fix it.
> 
> The compile instructions are the same as you've already been using,
> you'll just have to use --target=or1k-linux instead of
> --target=or32-linux like it used to be.
> 
> -Pete
> ___
Hi Pete,

This is good news. I wasn't aware you were working on this. I look
forward to reviewing

The master repository is on SVN at openCores.org, so we should get your
code into there. To date we have mirrored specific FSF releases of each
tool, but if you are tracking mainline, we should create src-mainline
and gcc-mainline trees. The build script will then be rather simpler,
and merging upstream changes must easier.

I'd like to propose that Pete is given SVN write access (has to be
approved through this mailing list).

It is important that the tools, particularly GCC, pass regression on
both newlib and uClibc targets. binutils and gcc have a good track
record here (the results are on the Wiki). I don't understand why
binutils is trying to run the shared object tests (I saw your other
email) - it should know this is not supported. It might be that an
expect test is wrongly determining support is present.

One of the weaknesses of all the regression tests is the lack of target
specific tests. For binutils it is a particular weakness, since it is so
target dependent.

GDB is a lot weaker on regression testing. I've been done a fair amount
of work porting GDB 7.3 over the last few months, aiming to improve the
robustness of the code, but it looks like you have leapfrogged me. I
suggest that when you have your code checked in, I'll add any relevant
changes I've made on top.

I look forward to seeing the code.

Best wishes,


Jeremy

-- 
Tel:      +44 (1590) 610184
Cell:     +44 (7970) 676050
SkypeID: jeremybennett
Email:   [email protected]
Web:     www.embecosm.com

_______________________________________________
OpenRISC mailing list
[email protected]
http://lists.openrisc.net/listinfo/openrisc

Reply via email to