Description of Problem:
This is really two problems. This is on beta hardware platform #28.
/dev/sda1 and /dev/sdb1 both have FAT16 partitions defined and WINNT
installed on /dev/sda1 with /dev/sdb2 as an extra FAT16 partition. Each of
these is 2000MB in size.
When the grub/lilo boot panel comes up, /dev/sda1 is defined as FAT16 with
"DOS" as the bootup name/label. However, /dev/sdb1 is defined as "ntfs"
with **NO** label. You cannot proceed until a label is defined.
The boot loader selected is grub.
Steps to Reproduce:
1. do gui install with definition as above
OK once I defined a label for /dev/sdb1 [(hd1,0) in grubese]
1. Partition type correct.
2. Null label allowed to indicate that entry NOT to be included.
I have no problem continuing on to the next screen without a label defined for
an NTFS partition.
As to the partition misdetect, was there ever an NTFS partition starting at this
location on the disk?
I have check into this and /dev/sdb1 is a FAT16 partition with a single file on
it (the Win2k paging file).
As far as I can tell, NTFS was never on that disk.
Could you run `parted /dev/sdb` and see if it also shows the partition as ntfs?
I ran parted against both sda and sdb (they each have a FAT16 partition as the
Doing check 1 on sda did some work but displayed nothing.
Doing check 2 on sdb said it did not support ntfs.
Whatever signature it (and anaconda) look for is present.
OK, I reinstalled Roswell into a test partition (/dev/sdb3) to check everything
again and discovered some small amount of additional information:
1. Again /dev/sdb1 was identified as ntfs. The fact that this was identified
ntfs is not important to me but it may contribute to the other part of the problem.
2. /dev/sda1 was identified as DOS -- really it is win2k installed on a FAT16
sdb1 had not label. If I remove the label for sda1 (delete "DOS") then then
"next" button gets greyed out. If I have a label on with sda1 or sdb1, then I
On other systems with winnt or win2k installed on hda1 or sda1 but no sdb1
issue, I can delete the label for "DOS" and proceed fine. It is only this one
system with the strange partition that causes a problem.
If parted says it's ntfs, anaconda will since anaconda is just using parted's
python module. I've fixed the having multiple blank labels not allowing you to
go forward in CVS.
Matt -- any ideas on the ntfs misdetect?
could you dd if=/dev/hda2 of=fshead bs=1024k count=1 and attach it, please?
(I'm moving this bug to parted, since this would be a parted bug. Shame on you
for reporting two bugs in one bugzilla....)
The bugs seemed to be one bug initially.
I haven't a clue as to the ntfs misdetect but will try to get the "dd" tomorrow.
I assume you want the information from the partition identified as ntfs:
/dev/sdb1. The only ide device on the system is the cdrom drive.
The /dev/sda2 partition is swap ... did you want that??
Created attachment 26793 [details]
dd if=/dev/sdb1 of=fshead bs=1024 count=1
sdb1 (a vfat formatted disk) is still recognized as NTFS. This still happens
with RC1 although leaving the system "name" blank does not create a badgrub.conf
and this can be ignored.
Unless you RedHat want to pursue this, I suggest it should be closed.
It clearly has the NTFS magic at the top of this filesystem.
does this still happen with latest parted?
Arrgh. Unfortunately, I no longer have access to that system. If you cannot
get the info elsewhere, I can try and create the problem but this will not be
easy or guarenteed to reproduce the problem.
I'll mark this as closed since we do not have a way to reproduce this exact
situation. If it continues to be a problem please reopen this report.