Add two situations where the cache is pruned on startup, removing all
the cached objects from storage:
1. the version field does not match the current one
2. the model has been changed out of band
The second one is very important as it allows to use the revng2 CLI
while editing the model manually on disk.
Ignore `OSError`s in `LocalStorageProvider` when rebuilding the list of
files for a given hash, since those paths might have become inaccessible
in the meantime.
Before this commit the `invalidate` code did invalidate all the objects
but left all the rows in `dependencies` that were paired to the produced
objects except the invalidated paths.
In this commit the logic is changed to be done in 3 steps:
1. SELECT all the objects that need to be invalidated, write down their
rowid
2. Delete all the rows in `dependencies` that were produced in
conjunction to the objects from (1)
3. Remove the objects from the table, returning invalidation data
This guarantees that the database does not accumulate invalidation data
over time due to invalidations.
Change the `get_model` and `set_model` interface of `StorageProvider` so
that it is responsibility of the `StorageProvider` to
serialize/deserialize the model before returning to the caller.
This is in preparation to the model migration being implemented, since
it's now a responsibility of the storage provider to deserialize it it
can trivially re-save it to disk if it is migrated.
Instead of executing `chdir` when using the `-C` option, which is
fragile when other paths are involved, store the option in the click
context and propagate it to the required code paths that require knowing
what the base directory is.
Allow the storage provider to be initialized with `local://?inline`.
This will set the cache directory to `$PWD/.cache` instead of
`~/.cache/revng`. This is to simplify testing for pypeline.
When running `invalidate`, use a temporary table to store the
invalidated model paths that need to be used. This saves to template
them into the query and allows them to be recycled later when the
`dependency` table needs to be cleaned up.
Move the epoch inside the storage_provider, this allows us to make the
storage_provider_factory to return an async context manager.
This way we can simplify the locking for the daemon, but forces the
CLI project commands to deal with async code.
Now the pype command has a `--verbose` argument that to enable debug
logging in the pypelien code.
Adapted the codebase to use this new logging format but replacing
`logging` with `revng.pypeline.utils.logger`, this is done in
preparation of debug-log.
Now pype has an `autocomplete` command usable to enable autocompletion
and revng2 now is based on `pype`, modifying its defalults and
injecting new commands.
Change the signature of `Container.serialize` and
`Container.deserialize` to use `Buffer` instead of `bytes`, as this
makes the functions compatible with the `revng::pypeline::Buffer`
classes.
When calling `StorageProvider.invalidate`, return a dictionary that
conveys to the caller which objects were deleted from storage as a
consequence of the paths supplied to the `invalidate` call.
Fix the invalidation logic in both `InMemoryStorageProvider` and
`Sqlite3StorageProvider`, as it did not account for propagating the
invalidation to both children objects and parent objects.
The column stored read model paths was erroneously called
`model_path_hash`, while the entire model was stored instead of a hash.
Rename the column to just `model_path` to avoid any confusion.
Giacomo Vercesi