[Server-devel] updates

[Jerry Vonau]

On Mon, 2009-01-26 at 17:08 -0500, Martin Langhoff wrote:
> On Sun, Jan 25, 2009 at 12:19 PM, Jerry Vonau <jvonau at shaw.ca> wrote:
> > latest re-roll of anaconda, srpms, rpms and patch used, updated
> > mkusbinstall for the XO at:
> >
> > http://members.shaw.ca/jvonau/pub/anaconda/
> 
> That's such good news.
> 
It's a bit broken, that addition to the anaconda file is unneeded, but
it uncovered the same bug that you ran into with kickstart on a fat 
filesystem. 


18:12:36 INFO    : moving (1) to step writeksconfig
18:12:36 INFO    : Writing autokickstart file
18:12:39 INFO    : Running kickstart %%traceback script(s)
18:12:39 INFO    : All kickstart %%traceback script(s) have been run
18:12:39 CRITICAL: anaconda 11.4.0.83 exception report
Traceback (most recent call first):
  File "/tmp/updates/anaconda", line 593, in writeMethodstr
    partition = isys.getDeviceByToken("LABEL", partition[6:])
  File "/usr/lib/anaconda/instdata.py", line 238, in writeKS
    self.anaconda.writeMethodstr(f)
  File "/usr/lib/anaconda/packages.py", line 77, in writeKSConfiguration
    anaconda.id.writeKS(fn)
  File "/usr/lib/anaconda/dispatch.py", line 208, in moveStep
    rc = stepFunc(self.anaconda)
  File "/usr/lib/anaconda/dispatch.py", line 131, in gotoNext
    self.moveStep()
  File "/usr/lib/anaconda/text.py", line 702, in run
    anaconda.dispatch.gotoNext()
  File "/tmp/updates/anaconda", line 1070, in <module>
    anaconda.intf.run(anaconda)
AttributeError: 'module' object has no attribute 'getDeviceByToken'

18:12:39 INFO    : in run, screen = <snack.SnackScreen instance at
0xb7a0d34c>
18:54:44 DEBUG   : adding drive sda to disk list
18:54:44 DEBUG   : adding drive sdb to disk list
18:54:45 CRITICAL: parted exception: Bug: Assertion
(ped_partition_is_active (part)) at disk.c:1186 in function
ped_partition_get_flag() failed.
18:54:54 CRITICAL: parted exception: Bug: Assertion
(part->disk_specific != NULL) at dos.c:1423 in function
msdos_partition_get_flag() failed.

Not sure what to file this one against, looks like parted to me, as one
of the booty functions import parted.


> > - finds kickstart file without asking where it is...
> 
> Good - does that depend on labelling the partition too? Do we have to
> change mkusbinstall to label it appropriately?
> 
> > - loader doesn't copy stage2 to RAM
> 
> Cool -
> 
> > - fix compile error re: MD_NEW_SIZE_BLOCKS(size)
> >
> > - added support for method=hd:LABEL=XSRepo at the boot prompt
> > This allows us not to ask where the install media if the usbdrive is
> > labeled as XSRepo and passed the above string.
> >
> >  - added support for mmc cards (for the XO) in loader. This is
> > incomplete, loader now finds the CF card on the XO but is unavailable to
> > partition and thus install to it. F10 has the same behavior, I filed BZ#
> > 481441. Hope they fix the issue before I have to dig it up....
> 
> I've CC'd myself on the bug. The actual number is - BZ 481431 -
> 
Fat fingered that, sorry.

> > - updated mkusbinstall added the labeling of the usbdrive for the above
> > method and creation of /boot/olpc.fth for the XO.
> 
> Ah, that's what I was alluding to earlier.
> 
> Is it possible to workaround the SD card issue by building an
> installer image with an older kernel? From what you said it's a
> regression with 2.6.27.x right?
> 
No, the issue is that anaconda, for F9, doesn't have the kernel modules
in the initrd, F10 does. Both fail to add the CF card to the
partitioning screen, thus making them unavailable as install targets.  

> I'm wondering if we can craft a just-installed image that can be dd'd
> to SD cards.
> 
That seems do-able, I'll try on the weekend to work something up.

> I guess I'll explore this tomorrow -- I'm on the road today, so
> limited in my hacking ;-)
> 
Jerry