Status of sparc in current git?

Rob Landley rob at landley.net
Thu Nov 12 04:36:02 UTC 2009


On Monday 02 November 2009 10:12:29 Austin Foxley wrote:
> On 11/01/2009 02:37 PM, Rob Landley wrote:
> > Somebody on IRC (skalpren?) told me that sparc works for them in current
> > git- master, and even sent me the .config they were using, so I tried it.
>
> It was me.

Sorry for the delay, I had to get an FWL release out and then track down and 
revert commit e0ac4efbdb4 (which makes _nothing_ work with my config, and has 
been in the repository since October 22).

That actually seems to make most of it behave.  It looks like most of what's 
left is either the kernel, qemu, or some really subtle toolchain mismatch:

/ # cat /proc/mounts
rootfs / rootfs rw 0 0
/dev/root / squashfs ro,relatime 0 0
sys /sys sysfs rw,relatime 0 0
dev /dev tmpfs rw,noatime 0 0
proc /proc proc rw,relatime 0 0
/ # ls -l /dev
ls: can't open '/dev': Cannot allocate memory
/ # ls -ld /dev
drwxrwxrwt    2 root     root           40 Jan ?? ??:?? /dev
/ # ls -l /sys
Illegal instruction

It's not a happy system, but it is _sort_ of running...

Well, statically linked, anyway:

  [    3.700000] Freeing unused kernel memory: 124k freed

  /bin/sh: symbol '_dl_loaded_modules': can't resolve symbol

  /bin/sh: symbol '_dl_app_init_array': can't resolve symbol
  [    4.052000] Kernel panic - not syncing: Attempted to kill init!

Dynamic linking still seems unhappy, need to work out why...

> > distcc[28] (dcc_ignore_sigpipe) Warning: signal(SIGPIPE, ignore) failed:
> > Invalid argument
> > distcc[29] (dcc_ignore_sigpipe) Warning: signal(SIGPIPE, default) failed:
> > Invalid argument
>
> The pipe syscall was broken on 0.9.30.1 but I definitely fixed it in
> git, so this looks suspiciously like you were somehow running the old
> release? Otherwise I don't really have an explanation. It was fixed in
> commit c033aa2af

With this morning's -git (and e0ac4efbdb4 reverted) I'm not getting that.

> Can you look at your source and verify that you indeed have these fixes?

It seems to have improved since last time I checked.

> -Austin

Thanks,

Rob
-- 
Latency is more important than throughput. It's that simple. - Linus Torvalds



More information about the uClibc mailing list