| Age | Commit message (Collapse) | Author |
|---|
|
disk handling. (#949)
* Renamed all_disks() to all_blockdevices() as it's technically a bit more correct. As it would return partitions as well. And it now distinguishes between BlockDevice() and Partition() when returning values. Also lsblk has been replaced with blkid and glob.glob() on /sys/class/block/
* Added handling of loop devices
* Added device information enrichment
* Removed old code
* Updated the usage of blockdevice.info['type'] to 'DEVTYPE' as per returned by blkid instead of lsblk
* Created a MapperDev() and DMCryptDev() 'disk' classes. This should help differentiate between BlockDevice(), Partition() and crypt-specific devices. Due to some new helper functions (mentioned later) BlockDevice().device_or_backfile has been simplified thanks to the information being pre-existing instead of having to call losetup. BlockDevice().mountpoint has been added as a dummy function. archinstall.enrich_blockdevice_information() has been added to enrich information extracted from blkid. archinstall.uevent() has been created to parse blkdid data when in -o export format, also eats up /sys/class/x/y/uevent data which is of the same format. all_blockdevices() now returns mapper devices found under /dev/mapper/ and properly returns the different types, not just BlockDevice() for everything. archinstall.get_parent_of_partition() has been added, which can translate /dev/sda1 to /dev/sda via strings by using /sys/class/block/ - note here tho that it has to be a blockdevice. Other parents won't return properly. archinstall.find_mountpoint() was created to be able to find mountpoints for a device, rather than the other way around which get_mount_info() handles. find_mountpoint() will convert /dev/sda1 -> /boot for instance, whereas get_mount_info('/boot') will convert /boot -> /dev/sda1. archinstall.get_partitions_in_use() will now properly traverse the entire structure, not just one level deep. This called for the creation of get_all_targets() which will take a dictionary structure and look for 'target': '/path' and return all '/path' values, recursively without being trapped in get_partitions_in_use() which has recursive targets. get_partitions_in_use() now returns a dictionary instead of a list, as it also returns the mountpoint and associated device. It will also prioritize MapperDev over DMCryptDev (as MapperDev is a child of DMCryptDev and can be resolved, other way around is 'trickier').
* Reverted Installer().partitions to return only the partitions. It's a slight speed defecit by not returning the mountpoints directly as {mntpoint: partition}, because the partition.mountpoint is a bit slow. But it avoids breaking legacy code :)
* Fixed a few imports, and added MapperDev.filesystem
* Fixed so that Installer().detect_encryption() uses the new separated logic of MapperDev/DMCryptDev/Partition
* Fixing a translation issue on hostname
* Added DMCryptDev.filesystem
* Added back archinstall.all_disks() for legacy reasons.
* Added a deprecation warning to archinstall.all_disks()
* Added a enrichment check for dm-crypt specific devices, which in turn sets DMCRYPT_NAME on the device information, making it easier to detect and target DMCryptDev when needed. This should avoid issues with loopdevices being detected as DMCryptDev. Also some minor tweaks to variable names on unencrypted setups
* Made debug output hidden without --verbose. Also added get_subvolume_info() which takes a path and returns any 'btrfs subvol show' information on the path
* Partition().subvolumes has been added. Which generates an iterator over potential subvolumes based on mountpoint. Partition().mount_information is now a list of mount-destinations, as one partition should be allowed to be mounted to several places (more commonly used in btrfs subvolumes). BtrfsSubvolume() has been added as a 'device type', which should make it more easy to recognize a BtrfsSubvolume object right out of the gate. Only found from Partition().subvolumes currently.
* Parameter fix for --verbose in log()
* Made sure loopdev devices have a PATH key as well, to make dealing with them as blockdevice as seamless as possible.
* Added backup information grabbing on uninitated devices
* Tweaked backup option if losetup cannot find the blockdev
* looks like losetup doesn't exit with a bad code even when it should. Raising ourselves when information is empty.
* Fixed structure returned from get_blockdevice_uevent()
* Made sure that fallback to blkid and loopdev info returns a PTTYPE (Partition Table Type) that is None, since we can't locate it yet (empty drive). But it's not False either, which is what logic after is looking for
* Deprecated split_bind_name() and added more debugging
* get_partitions_in_use() now only iterates over Partition() objects. And properly checks all mountpoints, not only the first.
* Flipped SysCallError to show the last 500 bytes if data rather than the first, to catch the actual errors.
* Removed anonymous parameters to all_blockdevices()
* Added .mount_information to MapperDev
* Added typing annotations
* Partition().mountpoint is now a @property that will live-report the current mountpoint.
* Removed the need for setting Partition().mountpoint = X when calling .mount() and .unmount(), as it's live-reported instead.
* Added .subvolumes to MapperDev()
* Added debug information
* Muting F841 in helpers.py
* Tweaked debug information in get_mount_info()
* Minimized log output by moving Partition().mountpoint to DEBUG, as partitions are allowed to not be mounted
* Simplified Installer().add_bootloader() to act as a router to the new split function, one for each bootloader type. This since flake8 complained about complexity and I agree that it became to wild.
|
|
* Correct definition of btrfs standard layout
* New version of the FlexibleMenu
* Added new functionality to Selector
* Created a GeneralMenu class
* GlobalMenu is made a child of GeneralMenu
* Some refining in GeneralMenu
secret is now a general function
* log is invoked in GeneralMenu directly
* Correction at GeneralMenu
* Materialize again _setup_selection_menu_options. Gives more room to play
* Callbacks converted as methods
Synch() (data area and menu) decoupled from enable()
and made general before any run
* script swiss added to the patch set
* Only_hd gets a new implementation of the menu
flake8 corrections
* swiss.py description added
* New version of the FlexibleMenu
* Added new functionality to Selector
* Created a GeneralMenu class
* GlobalMenu is made a child of GeneralMenu
* changes from the rebase left dangling
* Modify order of execution between exec_menu and post_processing.
Added selector_name as parameter for exec_menu
* minor corrections to the scripts
* Adapt to PR #874
* Solve issue #936
* make ask_for_a_timezone as synonym to ask_timezone
* Adapted to nationalization framework (PR 893).
String still NOT adapted
* flake8 complains
* Use of archinstall.output_config instead of local copy at swiss.py
* Problems with the last merge
* more flake8 complains. caused by reverted changes re. ask*timezone
* git complains
Co-authored-by: Anton Hvornum <anton@hvornum.se>
|
|
* flexibilize the definition of execution locale for OS commands executed via the SysCommand* interface.
Defined a storage argument which holds the default
Added functions to
unset the program own locales
reset to the program default locales
set a specific locale
A decorator to execute functions in the host locale environment
* rename decorator local_environ to host_locale_environ
created a simmetric decorator c_locale_environ, to make a routine work with the C locale whatever is set
* Correct definition of btrfs standard layout
* Added error handling
* Fixed issue where archinstall.Boot() would raise an exception in vain
* Added debugging for SysCommandWorker()
* Added some debugging
* Tweaking debug a bit
* Tweaking debug
* Adding more debug
* Adding more debug
* Removed some debugging
* Adding more debug
* Adding more debug
* Adding more debug
* Adding more debug
* Adding more debug
* Adding more debug
* Adding more debug
* Adding more debug
* Adding more debug
* Adding more debug
* Adding more debug
* Removed soem debugging
* Removed soem debugging
* Testing a revert
* Adding back the reverted change, adding lofile
* Redirecting stdout to /dev/null for testing (to avoid interrupting the fork)
* Reverted debug changes
* Testing os.system()
Co-authored-by: Anton Hvornum <anton@hvornum.se>
|
|
* Fixed SysCommandWorker() so that it removes ANSII VT100 escape codes. I also moved package.py into it's own folder, as that's something I want to expand on a lot, so package related stuff should go in there. I created a installed_package() function which gets information about the locally installed package. I changed so that find_packages() and find_package() returns a data-model instead for the package information. This should unify and make sure we detect issues down the line.
* Working on structuring .version constructor that works with BaseModel
* Added version contructors to VersionDef(). Also added __eq__ and __lt__ to LocalPackage() and PackageSearchResult().
* removed debug and added a TODO
* Removed whitespace
* Removed mirror-database function from myrepo
|
|
context, which means SysCommandWorker() never closed the pty, hogging all system handles for it.
|
|
* Btrfs with encrypted partitions.
We have changed installer.mount_ordered_layout into a series of loops
* open the encrypted devices
* manage btrfs subvolumes
* mount whatever
* create kyefiles for encrypted volumes
We have simplified the btrfs subvolume manager
We merged the locale branch as it is needed here
* We allow only the creation of keyfiles if the partition does not contain the root mount point.
Also, adapt examples/only_hd to the new __init__.py
Also, assorted flake8 warnings
* Cleanup code
* Naming schema for encrypted volumes
revert global locale association (provisional)
* We introduce the option of defining mount options in the partition dictionary.
It has forced us to define two new entries in this dictionary:
* format_options (formerly options) for mkfs options and
* mount_options for mount -o ones.
The different meaning of compress between partition and subvolumes is treated
* Function lib/disk/btrfs.py mount_subvolume marked as deprecated
Code cleanup.
* format_options now filesystem.options
* format_options now filesystem.format_options
mount_options nof filesystem.mount_options
* flake8 uncovered a slip in the code
|
|
* Define an standard locale for all the commands executed inside the application
* Allow LC_ALL to be overriden during call
* flake8 corrections
|
|
* Added type annotations for 1/5 of the files.
There's bound to be some issues with type miss-match, will sort that out later.
* Added type hints for 4/5 of the code
* Added type hints for 4.7/5 of the code
* Added type hints for 5/5 of the code base
* Split the linters into individual files
This should help with more clearly show which runner is breaking since they don't share a single common name any longer. Also moved mypy settings into pyproject.toml
* Fixed some of the last flake8 issues
* Missing parameter
* Fixed invalid lookahead types
* __future__ had to be at the top
* Fixed last flake8 issues
|
|
* An update to PR 715, making the handling of the *--mount-point* parameter less error prone.
I added a synomym (accepting the name both with underscore and dash) and ignoring when no value specified
I added it explicitly to the list to accept both the --parm value and --parm=value syntax
DOES NOT check the contents of the parameter
* Explicitly set all the know parameters
* Define explictly all parameters.
Make all non explicitly defined parameters behave as standard parameters, with on exception, names are not changed
Some cleanup of the code
In guided.py the reference to the dry_run parameter is updated to the standard naming convention for parameters
* Linter with flake8. corrections
* Linter with flake8. corrections (II)
* Linter with flake8. corrections (and III)
* Added --disk_layout argument. Was missing
I moved its loading from guided.py to __init__.py as it happens to the other json related arguments
* Better handling of errors during processing of the --disk_layouts parameter.
I define a routine to read an store a JSON file or stream. Tested on disk_layout
* Expand the former commit to all JSON file arguments
* Moved the function we created to read json files/streams to general.py. Add some comments
* flake8. A reference now unneded
* The merge process for the dry-run argument was causing the issue, not solving it
The del is just a cleanup for version upgrade without consequence (I hope)
* flake8 warning
* Correcting the last correction . Worked for old config files, but only for them
* New parameter parsing algorithm. More flexible and accepts multiple arguments (optionallY)
plus some documentation effort
* flake8 warning. For once is significant ( != None to not None)
|
|
mentioned in #819.
|
|
* Fixed some mypy issues regarding SysCommand* and logging
* Fixed imports and undefined variable
|
|
* Added support for storing disk encryption keyfiles and add them to a keyslot.
* Added a luks2().add_key() function in order to inject a keyfile into a keyslot on a encrypted volume.
* Simplified 'missing encryption password' logic in Filesystem(). Added a call to luks2().add_key() after the root-password is set on the volume, to add the keyfile in slot 2
* Adding in password handling in luks2().add_key(). It's required to enter a previous passphrase to unlock the volume and add a new keyslot. Also simplified the handling of partition in Installer().mount_ordered_layout()
* Adding in encryption on all partitions except /boot when encryption is opted in
* Removed setting size on Partition() as it's a read only value. No idea how Partition().size = size hasn't caused an issue before. Removed size=X argument to Partition()
* Added a uniqueness to the loopdevice name. This should ensure that multiple encrypted volumes can be opened at the same time, except for Partition().detect_inner_filesystem() operations which can only happen one at a time since they share namespace. This should never be an issue since archinstall is single threaded and no concurrent operations can/should happen.
* Added partprobe() as part of disk/helpers.py, added a /dev/ -> UUID mapper function called convert_device_to_uuid(path). Added a luks2().crypttab() function that sets up a /etc/crypttab entry.
* Moved the responsability for telling archinstall to generate a keyfile from Filesystem() to user_interaction.py. This should in the future be a user-input based value, and not something the Filesystem() automatically dictates.
* Added a retry mechanism to luks2().encrypt() to avoid having to re-start the installation when a device simply wasn't up yet.
* Swapping UUID= lookup from loopdev to physdev.
|
|
|
|
* Adding a cmd_history.txt log under /var/log/archinstall/ to get a clear picture of which commands was executed.
|
|
* Added multiple `partprobe` calls and added a `.partprobe()` function on partitions, filesystem and blockdevice.
* Adding retry attempts to all UUID related operations tied to the boot process
* Tweaked logging for mounting and disk related operations
* Removed potential SysCall exception disruptor causing exceptions to go by unnoticed
* Increased the start position from 1MiB to 5MiB of /boot partition
* Optimized the GRUB installation & config code
* Improved Partition().uuid to never return None. Instead it will raise an exception if it can't get a PARTUUID within X retries with Y delay per attempt.
* Increased sleep timer for partition uuid retrieval, because even with a 3 second sleep it wasn't long enough even on fast devices.
* Make GRUB install to /dev/sda instead of /dev/sda1.
* Added 10 retries for retreiving PARTUUID with a one second sleep. Instead of increasing the sleep simply add more retries until we find a good balance on slower disks.
|
|
|
|
from user_configuration.json into user_credentials.json. As well as the JSON serializer will omit the credentials from the user_configuration.json by detecting ! in the dictionary keys (which is why they are important). UNSAFE_JSON will leave those keys in there, good for storing credentials in a separate file."
|
|
mostly how we called things in guided.py but also some SysCommand calls
|
|
|
|
|
|
|
|
positions of partitions to 1MiB in.
|
|
|
|
Introduces the use of HTTP mirrors additionally
|
|
Attempt to fix issue regarding broken output
|
|
the size creation so that they don't overlap
|
|
|
|
|
|
Since pacman and some other commands these days write multiple lines and goes back and fourth, it's not reliable to say the "peaked" line is a single line. And if it's not, it will make the output look like garbage.
So instead, we'll write any output - as is - and let the command deal with fancy printing.
|
|
This fix introduces changes so that development can be done (and tested) on other platforms than Linux. This is a convenience fix and shouldn't break anything (simply a few Linux-specific imports that have moved into the functions where they are used).
This commit also introduces sorting based on a list of priorities (where the default will be last if not matched).
|
|
contexts, we have to flush the output manually here to avoid newlines not being before the next output.
|
|
|
|
|
|
|
|
* added support for ingesting config
* fixed condition to check key in dictionary
* Removed redundant code, profile and desktop keys are now optional
* Added base-config.json and support for pulling credentials from .env
* added base config file and env file for users credentials
* added silent install switch
* added python-dotenv as a dependency
* Updated Readme to include argparse changes as well as config ingestion
* Updated Readme to include argparse changes as well as config ingestion
* fixed typo in pyproject.toml
* Replaced the magic __builtin__ global variable. This should fix mypy complaints while still retaining the same functionality, kinda. It's less automatic but it's also less of dark magic, which makes sense for anyone but me.
* Fixes string index error.
* Quotation error.
* fixed initializing --script argument
* added python-dotenv as a dependency
* Installation can't be silent if config is not passed
* fixed silent install help
* fixed condition for ask_user_questions
* reverted to creating profile object properly
* Cleaned up and incorporated suggestions
* added Profile import
* added condition if Profile is null
* fixed condition
* updated parsing vars from argparse
* removed loading users from .env
* Reworking SysCommand & Moving to localectl for locale related activities (#4)
* Moving to `localectl` rather than local file manipulation *(both for listing locales and setting them)*.
* Swapped `loadkeys` for localectl.
* Renamed `main` to `maim` in awesome profile.
* Created `archinstall.Boot(<installation>)` which spawns a `systemd-nspawn` container against the installation target.
* Exposing systemd.py's internals to archinstall global scope.
* Re-worked `SysCommand` completely, it's now a wrapper for `SysCommandWorker` which supports interacting with the process in a different way. `SysCommand` should behave just like the old one, for backwards compatibility reasons. This fixes #68 and #69.
* `SysCommand()` now has a `.decode()` function that defaults to `UTF-8`.
* Adding back peak_output=True to pacstrap.
Co-authored-by: Anton Hvornum <anton.feeds@gmail.com>
Co-authored-by: Dylan Taylor <dylan@dylanmtaylor.com>
Co-authored-by: Anton Hvornum <anton@hvornum.se>
Co-authored-by: Anton Hvornum <anton.feeds@gmail.com>
* fixed indent
* removed redundant import
* removed duplicate import
* removed duplicate import
Co-authored-by: Anton Hvornum <anton.feeds@gmail.com>
Co-authored-by: Anton Hvornum <anton@hvornum.se>
Co-authored-by: Dylan M. Taylor <dylan@dylanmtaylor.com>
|
|
* Moving to `localectl` rather than local file manipulation *(both for listing locales and setting them)*.
* Swapped `loadkeys` for localectl.
* Renamed `main` to `maim` in awesome profile.
* Created `archinstall.Boot(<installation>)` which spawns a `systemd-nspawn` container against the installation target.
* Exposing systemd.py's internals to archinstall global scope.
* Re-worked `SysCommand` completely, it's now a wrapper for `SysCommandWorker` which supports interacting with the process in a different way. `SysCommand` should behave just like the old one, for backwards compatibility reasons. This fixes #68 and #69.
* `SysCommand()` now has a `.decode()` function that defaults to `UTF-8`.
* Adding back peak_output=True to pacstrap.
Co-authored-by: Anton Hvornum <anton.feeds@gmail.com>
Co-authored-by: Dylan Taylor <dylan@dylanmtaylor.com>
|
|
|
|
https://docs.python-guide.org/writing/gotchas/#mutable-default-arguments
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
* Moving away from custom log levels, to something that's well defined.
* Added backward compability to log() as well.
* Added an option to force log messages out on screen even if the level is below the log level threashold.
* Added force log messages when wrong notation is used.
* Added some more length to the deprecated message
* Swapped all log levels to use logging.<level> instead.
Co-authored-by: Anton Hvornum <anton.feeds@gmail.com>
|
|
|
|
required to disable systemd coloring and paging when querying for service states. Otherwise it returns unreliable data that can cause hanging.
|
|
|
|
the peak again. Added this to any pacstrap call.
|