Re: The current kernel for a Jornada 680 SH3

Paul Mundt <[email protected]> wrote:
> I've built 2.6 from gcc 3.0.x up to gcc 3.4 from CVS. There really aren't too
> many surprised remaining at this point, so any reasonably recent toolchain
> should work for the kernel just fine.

Although Paul has said all about the toolchain for SH kernel, I'd like
to say that even for the newer x86 kernels, some folks recommend
gcc-2.95.x for the kernel :-)

I believe that Dan's crosstool is not only just fine but also very
instructive.  If you gaze at what goes on by crosstool, you can find
a nice documentation with animation for cross toolchain and will
see what is "building cross toolchain".  Even if you understood it
well already, crosstool would be impressive.

BTW, it'd be better to report about gcc-3.3/3.4 status for sh*-linux*
targets.  Now almost all our local patches are merged to the mailine
FSF gcc.  Such merge will never occur for gcc-3.2 and lower because
they are completely frozen.  We have to live with local patches about
3.2.

gcc-3.3.3 was already released and gcc-3.3.4 will be released in very
near future.  gcc-3.3.4 includes all local patches I know for 3.3.
Unfortunately, there is a few bugs which have solved very recently and
is too late on 3.3.4 release:

  http://gcc.gnu.org/bugzilla/show_bug.cgi?id=15647
  http://gcc.gnu.org/bugzilla/show_bug.cgi?id=13250

I think gcc-3.3.4 will be enough stable and ok for userland work. Of
course there are some known ICE problems but almost of them can be
avoided by lowering the optimization level.

gcc-3.4.0 was released at Apr.20 and 3.4.1 is planed to ship in June.
Unfortunately again, there was one non SH specific bug found in 3.4.0
after freezing to release.  It causes a hang in native gcc bootstrap
on sh-linux.
3.4.1 will be the first sane compiler about java (gcj) for sh-linux,
though gij (interpreter) doesn't work on this target without a tiny local
patch.  I'll repost it after 3.4.1 released.  It isn't the right fix and
will be never apply to mainline.  Sadly it seems the right fix is beyond
me.  BTW, 3.4.1 also has the new feature for TLS (thread local storage)
support which is needed for NPTL in glibc.  CVS glibc has the SH specific
NPTL stuff and now my 2.6 system is working with NPTLed glibc-2.3.3,
though I need -O1 or -O2 -fno-unit-at-a-time -fno-reorder-blocks when
building this glibc.  It means there are wrong code generation problems
in gcc-3.4 -O2 on sh4-unknown-linux-gnu target.

Regards,
	kaz


-------------------------------------------------------
This SF.Net email is sponsored by: Oracle 10g
Get certified on the hottest thing ever to hit the market... Oracle 10g. 
Take an Oracle 10g class now, and we'll give you the exam FREE.
http://ads.osdn.com/?ad_id=3149&alloc_id=8166&op=click