Bug 105441

Update: with the manually created partition table, the installation was successful

Please provide y2logs, see:
http://www.opensuse.org/index.php/Bug_Reporting_FAQ#YaST

The option to reread the partition table still exists, it is a menue entry
under the "Expert" button on the lower row of buttons in Expter partitioner.

To do something about the bug we need the logs of a failed installation,
not of a succesful one.

I just redid the installation from scratch and did it the way I did it the first
time, and it failed in exactly the same way. I'm attaching the yast log files
from this installation attempt

Created attachment 46453 [details]
log files from failed installation attempt

Thanks for the logs.

Jiri, I think this is a problem with bootloader, from the view of storage all
went fine. I think the problem could be caused by the new dynamic /dev
directory. Maybe there is no instance that creates device node for /dev/hda6 
in the installed system.

Last log entries are:
Y2SystemFunction.cc(useRemote):129 'initializeBootloader': switched to remote
2StdioFunction.cc(evaluateCall):137 Evaluating remote call to 'Bootloader_API::init
bootloader/routines/lib_iface.ycp:204 Error occurred while initializing bootloader
BootGRUB.ycp:572 GRUB return value: nil

Yes, missing /dev/hda6 is the problem. GRUB needs to open the device. Please, 
attach also YaST logs from the instaleld system (during 
installation /mnt/var/log/YaSTY2), they would be more useful in this case to 
proof how GRUB failed. 
 
Reassigning to udev maintainer.  

i don't understand this last remark. The logs I got were from var/log/YaST2
during the failed installation attempt. Since the root partition disappears, I
can't boot to it.

Do you mean you want me to create the partition manually again and install the
complete system to send you logs from?

Well, what exactly do you mean? Partition disappears, or the node in /dev 
disappears? 
 
While finishing basic installation, a part of YaST runs in chroot environment 
in the target syetem. And it also puts logs there. So, the output of GRUB 
command itself is logged there. 

By the way, some of the comments seem to say that you think it's only a problem
with the dynamic devs. Just to confirm, the partition does not exist on the hard
drive partition table at the point grub tries to install. It is gone

Ah, thanks for clarification. Then, it is problem for you, Thomas. Without the 
ability to determine where the partition begins and ends, GRUB can hardly 
install properly. 

I seriously doubt it is a problem of yast2-storage. I already analized the logs
and there is no problem with creating the partiton /dev/hda6, mounting it,
installing package onto it. How should installation of all packages succeed if
there is no partition /dev/hda6. 

No idea what exactly happens during system startup but to my knowledge 
yast2-storage is not involved here at all.

Related issue: when trying to create a new partion in a running system,  
creating a filesystem fails because the device node was not created until  
after the mkreiserfs was called.     
  
Thomas told me this is due to udev creating the device nodes asynchronously.  

Created attachment 46467 [details]
y2log of failed partition creation

@comment #12: 
 
I understand your concern. I must say I have no clue why the partition table 
on the disk is corrupted. Bootloader installer modifies MBR, but not the are 
where partition table resides (but for running the activate command). And, 
hda6 is logical partition, so info about it is not stored in MBR AFAIK. 

On rebooting and restarting the installation from scratch, I get an error
message saying 'the partition table is not readable by parted'

Ok, I looked at the logs once more and found the following error messages
issued by parted during partition creation:

You found a bug in GNU Parted.  Please email
        parted-maintainers@lists.alioth.debian.org
or (preferably) file a bug report at
        http://parted.alioth.debian.org/bugs/
Your report should contain the version of this release (1.6.24)
along with the following message and preferably
additional information about your setup:

Assertion (metadata_length > 0) at disk_dos.c:1983 in function
add_logical_part_metadata() failed.

Unfortunately parted still gives a exit code of 0 therefore I did not see
this on my first scan through the logs.

It seems to happen only when choosing "use entire hard disk". I just manually
wiped the partition table using fdisk, and started the installation. The
proposal then became to create hda1 and hda2 as swap and root respectively, and
this partition table worked

Sure the parted bug is in creation of a logical partitition so if you create
primariy partitions it will not show up.

Unfortunately the problem is not reproducible on the ide disk I have in my 
machine. If I try the same partitioning commands here they succeed.
 
Would it be possible to get access via ssh to the machine?

sure. Contact me on groupwise messenger (ajohansson)

Menwhile I was succefully reproducing it on another disk.
I will probably contact you if I have a possible fix for parted.
Thanks anyway.

Ok, I changed YaST2 now to avoid the parted problem. 
It would be very helpful if you could retry your installation with beta#3.

Will do, as soon as it's released :) 
 
(Unless I can find a prerelease with your fix included somewhere) 

*** Bug 105445 has been marked as a duplicate of this bug. ***