This refines `useHasIsoTargetOnly` hook into `useHasSpecificTargetOnly` hook which takes an image type as an argument and can be re-used for non-iso targets as well.
Previously created blueprints didn't take into account what the current latest RHEL release is and built the specific minor release that was included in the blueprint. This should fix the bug and also fix snapshots.
This updates the logic of parsing bytesize into size/unit pair.
Currently 'GiB' is set as a default which causes problem with smaller partitions on Edit.
A validation for the min_size is in place and the rule is that the size cannot be smaller than 0 of a given unit. If there was a partition with smaller size in KiB it will break the rule when converted to GiB.
Edit mode is now fully tested and working for OpenSCAP profiles. A
handler for the PUT request was added and the fixtures were updated to
support this.
`EditImageWizard.tsx`: Previously we dispatched `initializeWizard()`
while waiting for the blueprintDetails to load. This meant that the
state was incorrect (empty) while the blueprintDetails request was
in flight.
`requestMapper.tsx`: Correctly populate state in edit mode if the
blueprint contains a custom file system – previously custom mountpoints
were dropped and automatic mode was selected.
`spyOnRequest()`: Differentiate between request types (e.g. GET, PUT)
for the same endpoint.
This hides the "Manually configure partitions" option from the FSC step for blueprints which have only ISO selected as a target.
There is already an alert in place for cases when ISO is combined with another target.
The snapshot_date in the image_request must be a date. snapshot_date is
an optional field and if no snapshot is used, the field should be
undefined. Previously it was an empty string and this caused image
builder to return a code 400 with message "Snapshot date is not in
DateOnly (yyyy-mm-dd) format".
OpenSCAP profile info is now correctly loaded into the state and
verified to be in the final request with tests.
I modified the fixtures for the OpenSCAP profiles. My changes ensure we
have a Venn-diagram like overlap when changing from one profile to
another where one package is the same, one package is removed, and one
package is added. The same is true for the services and partitions. For
kernel args it is not so important as that is just a string (as opposed
to an array), so it is enough to be different.
I was able to eliminate a useEffect by replacing it with a lazy query
trigger function. Setting the second arg `preferCacheValue` to `true`
means that a request is only made if cached data is not available.
I modified the Redux reducers a bit to add some additional safety.
`changeFileSystemPartitionMode` is now responsible for initializing the
partitions field in the state by adding the root partition – previously
this was done in the components themselves by dispatching
`addPartition`. This reducer can always be safely dispatched – if the
mode is ‘manual’, dispatching it with a payload of ‘manual’ will not
result in any changes to the state.
`addPackage` is also safer now. When a package is added, the list of
packages is checked. If there is a package with an identical name, the
new package overwrites the previous package. This is useful because the
description may be different for the same package – for instance, when
adding an OpenSCAP package, we use a custom description (‘this package
required by OpenSCAP’).
This adds a new field called `wizardMode` to the state. This field can have two possible values: `create` and `edit` and it indicates whether the Wizard was opened in Create or Edit mode.
Store validation status in redux state.
This is bit complex on the redux side, but pretty simple on the components.
It allows for reuse of the validation state instead of revalidating wherever needed.
This initializes packages upon a selection of an OpenSCAP profile.
Current logic also fixes a bug previously present in V1 - when changing an OpenSCAP profile with other previously chosen packages, the packages outside of the profile persist and don't get removed.
The FSC step is added to the wizard and takes full advantage of Redux
for state management.
This is still a work in progress.
Supported features:
1. Select partition mountpoint prefix (e.g. /var, /home)
2. Edit partition mountpoint suffix (e.g. /home/videogames)
3. Change displayed units (KiB, MiB, GiB)
Supported but buggy features:
1. Edit partition size
Unsupported features:
1. Add partitions
2. Remove partitions
3. Validation
Previously the V2 Wizard's request mapper was only adding the
custom_repositories field. We also need to add a nearly duplicate
payload_repositories field due to how the image-builder API works...
which is admittedly not intuitive.
Tests are also added to ensure that requests are generated correctly
when using the custom repositories feature.
We need to start using `undefined` as the default state for when a value
has not been defined. Previously we had used things like `’’` for string
typed values. But this causes problems later when generating the request
we send to image-builder. Using `undefined` is explicit and will make
generating the requests much easier (as we don’t need to check for `’’`,
determine the intent, and convert it to undefined if necessary).
Explicit is better than implicit.
With that in mind, tests have been added to ensure that the correct
request is sent to the API for every option on the Registration step.
This is facilitated using a new `spyOnRequest()` function. In the
future, we will have similar tests for the rest of the steps.
A few other minor things:
1. We need to get the `store` using `useStore()`
for when we later call `store.getState()` because the tests use a different
store that is configured in the renderer than the one we were importing.
2. In the wizardSlice, a new type RegistrationType is added that provides
additional type safety instead of using `string`.
