espresso formats partitions it wasn't asked to

[Martin Pitt]

in gparted, I removed my spare swap partition on /dev/hdc1 (due to bug 35074), and created a new vfat one on that instead.

Now, when going forward to the mount point setup, I have two entries:

/dev/hda1 swap
/dev/hdc2 /

When clicking 'Forward', the dialog tells me that not only /dev/hda1 and /dev/hdc2 are formatted, but also :

- /dev/hdc1 (the new partition I just created) as swap (that's the old type, I just recreated it as type vfat)
- /dev/hdc4, which contains all my multimedia stuff; I never touched /dev/hdc4 in the partitioner or mount point setup

[Martin Pitt (pitti)]

espresso log with ESPRESSO_DEBUG=1

[Martin Pitt (pitti)]

Oh, /dev/hdc4 was the default for / in the mount point setup. That's where it came from, probably, I changed that to hdc2.

[Martin Pitt (pitti)]

This persists even after quitting and restarting espresso.

[Barrakketh (barrakketh)]

Something annoying (though harmless AFAIK) is that Espresso will try to format all swap partitions regardless of whether they were selected or not. I have an existing Ubuntu install on another physical drive and it formatted that installation's swap partition. It also adds that installation's swap partition to the new installation's fstab.

[Matt Zimmerman (mdz)]

This sounds rather serious...

[Robin Sonefors (ozamosi)]

Solaris uses the same partition type as Linux SWAP, right? Would that mean that it would format any Solaris-install found on disc?

[Holger Arnold (holgerarnold)]

This bug is *really* serious: the installer happily overwrites existing partitions (including an existing Windows NTFS partition).

Steps:
1. Resize NTFS partition /dev/hda1.
2. Add extended partition and logical partitions for /, /home (both ext3), and swap.
3. Add mount points for / and /home.
4. /dev/hda1 gets formatted with ext3...

[Walter Neto (kublin)]

Traceback (most recent call last):
  File "/usr/lib/python2.4/site-packages/ubiquity/frontend/kde-ui.py", line 536, in on_next_clicked
    self.dbfilter.ok_handler()
  File "/usr/lib/python2.4/site-packages/ubiquity/components/partman.py", line 211, in ok_handler
    disk_choice = self.frontend.get_disk_choice()
  File "/usr/lib/python2.4/site-packages/ubiquity/frontend/kde-ui.py", line 752, in get_disk_choice
    print " get_disk_choice (self): " + str(self.part_disk_buttongroup.selected().text())
UnicodeEncodeError: 'ascii' codec can't encode characters in position 37-38: ordinal not in range(128)

[Colin Watson (cjwatson)]

Walter, please file a new bug rather than attaching your crash dump to this unrelated bug. Filing a new bug makes your problem easier for me to deal with, and makes it less likely that I will forget about your problem.

[Tormod Volden (tormodvolden)]

Similar issues have been resolved in Dapper Beta 2. Can you please try Beta 2 or newer and report back if it is still happening?

[Colin Watson (cjwatson)]

I'm not aware of anything much having changed with respect to this bug. Perhaps Tormod knows something I don't :-)

[Tormod Volden (tormodvolden)]

Sorry, I'm afraid my optimism was not rooted in reality :)
I guess I mixed up this bug with bug #40464. I hope I didn't cause too much inconvience or disasters :/ I realize the potential of doing so and apologize.

[Ublis (ublis)]

I can verify it still occurs. First I saw in the overview that the swap partition was going to be formatted as swap, though I requested it not to. The other partitions which were used showed "partition <bla> is going to be formatted as" - without mentioning as what. I had no idea what this was supposed to mean: will they be formatted or not? I went back to the mounting points definition, checked I did the right thing and went forward to the overview. Now it said they'll be formatted as ext3.
Fortunately I had a new HD, so nothing was lost - except I was forced to use ext3 instead of xfs, which is what I was aiming for.

[Colin Watson (cjwatson)]

I think I've got a fix for this now.

[Colin Watson (cjwatson)]

ubiquity (0.99.76) dapper; urgency=low

  * Make deselection of mount points actually have a useful effect on
    partman_commit (closes: Malone #35079).

 -- Colin Watson <email address hidden> Sat, 6 May 2006 00:09:22 +0100