Re: Synch with 2.4.4

[Andreas Jaeger <aj at suse dot de>]

Hans-Peter Nilsson <hans-peter.nilsson@axis.com> writes:

> > From: Andreas Jaeger <aj@suse.de>
> > Date: 29 Apr 2001 14:31:08 +0200
> 
> > I merged some changes from Linux 2.4.4 with glibc.  I've committed the
> > appended patch.
> > 
> > Hans-Peter, kernel-features doesn't contain any information about
> > cris.  Can you look into this, please?
> 
> I was under the delusion that the generic checking in
> kernel-features.h was enough.  While looking at the official
> 2.4.3 and 2.4.4 I see that there's actually no usable official
> kernel release that supports the CRIS glibc port. :-(

We do some checking for special architectures for some constructs but
not for all.  So, please double check.

> Linux/CRIS 2.4.3 had an older system call ABI, and 2.4.4 has a
> lethal bug in the interrupt epilogue and anyway doesn't contain
> certain system calls.  I'm targetting a patch set that (so I'm
> told) was included in 2.4.3-ac10 but for unknown reasons was not
> included in 2.4.4.  For the record, that patch set is available
> in a tarball including other things at
> <URL:http://developer.axis.com/download/devboard_lx/R1_0_0/>.
> 
> Since IIUC glibc can't require 2.4.5 until it's released, and
> for maintenance reasons should not have the kernel-version-
> machinery target a specfic patch set, I think it's best to leave
> kernel-features.h as is, until there's a suitable official
> kernel release.  In the meantime perhaps a note be added that
> the CRIS port needs kernel patches to work.

Feel free to send a patch to the FAQ document.

> As mentioned, the glibc port as is should be considered work in
> progress.  I'll send an update after checking some more apps;
> there are a few bugs that makes it unusable for other than
> "Hello, world".  Also, all but a few of those pesky
> i386-includes can go, as Drepper noted.
> 
> > 	* sysdeps/unix/sysv/linux/kernel-features.h (__ASSUME_FCNTL64):
> > 	Define for Arm, PowerPC and SH if kernel is 2.4.4 or newer.
> 
> For 2.4.5 (hopefully), there would be a similar construct for
> CRIS.

Fine.

Thanks,
Andreas
-- 
 Andreas Jaeger
  SuSE Labs aj@suse.de
   private aj@arthur.inka.de
    http://www.suse.de/~aj