Tails

#2 Updated by Anonymous 2015-06-10 10:20:43

First report:

I tested this with a self built package of liveusb-creator on Jessie (built from the feature/jessie_clean branch, arch=i386, because of the missing syslinux binary for amd64 in the current ISO).

The command I used was liveusb-creator -nvxm. I could successfully install a copy of Tails this way to a USB stick.

What did not work was:

• automatically formatting the device if it’s not in FAT initially. Thus, I formatted it myself as FAT32 before running the installation.
• partitioning the device using the -P option. I suspect https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=767457

I’ll investigate a bit more before modifying the launcher.

#3 Updated by intrigeri 2015-06-10 12:50:14

> I could successfully install a copy of Tails this way to a USB stick.

Yay! :)

> What did not work was:
> * automatically formatting the device if it’s not in FAT initially. Thus, I formatted it myself as FAT32 before running the installation.

It’s intentional that we don’t format anything unless -P is passed.
Tails Installer only supports initial install (with -P) and upgrading an existing Tails (that doesn’t require any formatting).

> * partitioning the device using the -P option. I suspect https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=767457

That’s supposedly workaround’ed in liveusb/creator.py: look for that bug number in there. Is our workaround buggy, or is anything else happening that we’re not catching?

#4 Updated by Anonymous 2015-06-10 14:58:28

intrigeri wrote:
> > I could successfully install a copy of Tails this way to a USB stick.
>
> Yay! :)
>
> > What did not work was:
> > * automatically formatting the device if it’s not in FAT initially. Thus, I formatted it myself as FAT32 before running the installation.
>
> It’s intentional that we don’t format anything unless -P is passed.
> Tails Installer only supports initial install (with -P) and upgrading an existing Tails (that doesn’t require any formatting).

thanks for confirming this.

> > * partitioning the device using the -P option. I suspect https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=767457
>
> That’s supposedly workaround’ed in liveusb/creator.py: look for that bug number in there. Is our workaround buggy, or is anything else happening that we’re not catching?

indeed, but there seems to be a bug. sometimes it works with -P, sometimes i get an list index out of range error on line 901 in creator.py which is the line choosing the system partition.
I’ll post the exact output here as soon as my VM gets its network back :)

#5 Updated by Anonymous 2015-06-10 20:16:41

[creator.py:868 (partition_device)] DEBUG: Creating partition
[creator.py:323 (popen)] DEBUG: ['/sbin/sgdisk', '--attributes=1:set:0', '--attributes=1:set:2', '--attributes=1:set:60', '--attributes=1:set:62', '--attributes=1:set:63', '/dev/sda']
[gui.py:270 (run)] ERROR: list index out of range
Traceback (most recent call last):
  File "/usr/lib/python2.7/dist-packages/liveusb/gui.py", line 203, in run
    self.live.partition_device()
  File "/usr/lib/python2.7/dist-packages/liveusb/creator.py", line 901, in partition_device
    system_partition = partitions[0]
IndexError: list index out of range

That’s the error I got multiple times in the beginning, but it stopped randomly.

#6 Updated by intrigeri 2015-06-11 07:59:41

This looks like a leftover of Feature #8290, then. Next debugging step is to reproduce, and then check if indeed the partition was created:

1. if it was created, but it’s not in the partitions list, then I guess that some caching is going on, that we need to discard;
2. if it was not created (which I doubt, otherwise self._set_partition_flags(self.drive['device'], 1, SYSTEM_PARTITION_FLAGS) would have failed earlier), then probably the call for partition_table.call_create_partition_sync has failed without us noticing (which I would be surprised by).

Once you have this info, then I’ll send it back to Alan on Feature #8290, and if he hasn’t time to fix it, oh well, I guess you or I will have to.

#7 Updated by Anonymous 2015-06-12 13:44:58

Hi,

> This looks like a leftover of Feature #8290, then. Next debugging step is to reproduce, and then check if indeed the partition was created:
>
> * if it was created, but it’s not in the partitions list, then I guess that some caching is going on, that we need to discard;

I was able to reproduce the bug. The partition was indeed created.
But I still cannot install Tails now. With the -P option I always get the same error.
Without the -P option. I get “unknown filesystem. you might need to reformat the device.”

From here, in order to make it work, I need to manually reformat the system partition as FAT. cfdisk first shows it as Linux partition. Then run mkfs.fat on it. And then I can install Tails using liveusb-creator -nmxv