+ BTRFS support
+ Refactoring of `disk.py` into `lib/disk/*.py`
+ `get_mount_info()` was reworked to support traverse backwards until mountpoint information is found (reverse-recursiveness)
+ `suggest_single_disk_layout()` was added/fixed to support BTRFS options
+ - `suggest_multi_disk_layout()` does not yet support BTRFS subvolumes in suggestive mode
+ `Installer()` now calls `create_subvolume()` and `mount_subvolume()` during loading of file structure on `mount_ordered_layout()`
File: lib/user_interaction.py
When function manage_new_and_existing_partitions() is used, and
'Suggest partition layout' is selected, the installer crashes.
REASON
Bug was introduced in commit 9e67ce3, when partition layout was
changed to use device.path as keys (instead of device).
It seems all necessary changes were made for this, except this one.
File: lib/disk.py
When installing on a loopback device (a.k.a loop device), function
Filesystem.partuuid_to_index() crashes with a JSON parsing error.
REASON
1) For loop devices, the property BlockDevice.device returns the
actual image file (back-file) of the loop device instead of the
/dev/X device.
2) Function Filesystem.partuuid_to_index() executes `lsblk --json`
against BlockDevice.device .
3) `lsblk` fails and prints the error "not a block device" to stderr.
This causes the output to not be valid JSON.
4) Code crashes when JSON parser tries to parse the output.
SOLUTION
- Make sure property BlockDevice.device only returns a valid block
device.
- Create new function BlockDevice.device_or_backfile that mimics
the present behaviour of BlockDevice.device.
- Use BlockDevice.device_or_backfile in function
BlockDevice.__repr__().
SOLUTION REASONING
I can only see one reason behind BlockDevice.device returning
the back-file of a loop device, and that is to show the back-file
to the user (instead of /dev/X) when printing the string
representation of a BlockDevice.
All other parts of the code can use the /dev/X file just fine.
And IMO it makes more sense that a property named `device` only
returns devices, and not normal files.
File: lib/user_interaction.py
When function manage_new_and_existing_partitions() is used, and
no partitions are configured, and 'Set desired filesystem for a
partition' is selected, the installer crashes.
File: lib/user_interaction.py
When function manage_new_and_existing_partitions() is used, and
'Suggest partition layout' is selected, the partition info is not
correctly stored in the config.
Instead of: {"partitions": [{...}, {...}]}
You get: {"partitions": {"partitions": [{...}, {...}], "wipe":True}}
Bug affects normal interactive usage (example/guided.py).
The timezone configured in the installer is not the timezone that
ends up in the new installed system. Instead, the timezone
used in the host system (from where the installer is run) is the
one that finally ends up being used.
Reason: systemd-nspawn by default copies the host timezone into
the target. And systemd-nspawn is run when keyboard-layout is
changed (which is done after changing the timezone).
Solution: Add option `--timezone=off` to systemd-nspawn, which
hinders affecting the timezone in the target.
I'm making sure that the JSON structure of the user config can get a say in how the subvolumes should be used later on. As well as splitting up where the logic should be to make it easier to maintain.
Anton Hvornum
Anton Hvornum
Oleksandr Zinkevych
Richard Neumann
theaeonsolution
JakobDev
Hugo Ankarloo
SecondThundeR
Falieri