chia-blockchain 2.5.1rc1__py3-none-any.whl

chia/_tests/plotting/test_plot_manager.py

@@ -0,0 +1,781 @@
+from __future__ import annotations
+
+import logging
+import sys
+import time
+from collections.abc import Iterator
+from dataclasses import dataclass, replace
+from os import unlink
+from pathlib import Path
+from shutil import copy, move
+from typing import Callable, Optional, cast
+
+import pytest
+from chia_rs import G1Element
+
+from chia._tests.plotting.util import get_test_plots
+from chia._tests.util.misc import boolean_datacases
+from chia._tests.util.time_out_assert import time_out_assert
+from chia.plotting.cache import CURRENT_VERSION, CacheDataV1
+from chia.plotting.manager import Cache, PlotManager
+from chia.plotting.util import (
+    PlotInfo,
+    PlotRefreshEvents,
+    PlotRefreshResult,
+    add_plot_directory,
+    get_plot_directories,
+    remove_plot,
+    remove_plot_directory,
+)
+from chia.simulator.block_tools import get_plot_dir
+from chia.util.config import create_default_chia_config, lock_and_load_config, save_config
+from chia.util.ints import uint16, uint32
+from chia.util.streamable import VersionedBlob
+
+log = logging.getLogger(__name__)
+
+# These tests are not dependent on consensus rules
+pytestmark = [pytest.mark.limit_consensus_modes(reason="does not depend on consensus rules")]
+
+
+@dataclass
+class MockDiskProver:
+    filename: str
+
+    def get_filename(self) -> str:
+        return self.filename
+
+
+@dataclass
+class MockPlotInfo:
+    prover: MockDiskProver
+
+
+class Directory:
+    path: Path
+    plots: list[Path]
+
+    def __init__(self, path: Path, plots_origin: list[Path]):
+        self.path = path
+        path.mkdir(parents=True, exist_ok=True)
+        # Drop the existing files in the test directories
+        for plot in path.iterdir():
+            unlink(plot)
+        # Copy over the original plots
+        for plot in plots_origin:
+            if not Path(path / plot.name).exists():
+                copy(plot, path)
+        # Adjust the paths to reflect the testing plots
+        self.plots = [path / plot.name for plot in plots_origin]
+
+    def __len__(self):
+        return len(self.plots)
+
+    def plot_info_list(self) -> list[MockPlotInfo]:
+        return [MockPlotInfo(MockDiskProver(str(x))) for x in self.plots]
+
+    def path_list(self) -> list[Path]:
+        return self.plots
+
+    def drop(self, path: Path):
+        assert self.path / path.name
+        del self.plots[self.plots.index(self.path / path.name)]
+
+
+class PlotRefreshTester:
+    plot_manager: PlotManager
+    expected_result: PlotRefreshResult
+    expected_result_matched: bool
+
+    def __init__(self, root_path: Path):
+        self.plot_manager = PlotManager(root_path, self.refresh_callback)
+        # Set a very high refresh interval here to avoid unintentional refresh cycles
+        self.plot_manager.refresh_parameter = replace(
+            self.plot_manager.refresh_parameter, interval_seconds=uint32(10000)
+        )
+        # Set to the current time to avoid automated refresh after we start below.
+        self.plot_manager.last_refresh_time = time.time()
+        self.plot_manager.start_refreshing()
+
+    def refresh_callback(self, event: PlotRefreshEvents, refresh_result: PlotRefreshResult):
+        if event != PlotRefreshEvents.done:
+            # Only validate the final results for this tests
+            return
+        for name in ["loaded", "removed", "processed", "remaining"]:
+            try:
+                actual_value = refresh_result.__getattribute__(name)
+                if type(actual_value) is list:
+                    expected_list = self.expected_result.__getattribute__(name)
+                    if len(expected_list) != len(actual_value):
+                        return
+                    values_found = 0
+                    for value in actual_value:
+                        if type(value) is PlotInfo:
+                            for plot_info in expected_list:
+                                if plot_info.prover.get_filename() == value.prover.get_filename():
+                                    values_found += 1
+                                    continue
+                        else:
+                            if value in expected_list:
+                                values_found += 1
+                                continue
+                    if values_found != len(expected_list):
+                        log.error(f"{name} invalid: values_found {values_found} expected {len(expected_list)}")
+                        return
+                else:
+                    expected_value = self.expected_result.__getattribute__(name)
+                    if actual_value != expected_value:
+                        log.error(f"{name} invalid: actual {actual_value} expected {expected_value}")
+                        return
+
+            except AttributeError as error:
+                log.error(f"{error}")
+                return
+
+        self.expected_result_matched = True
+
+    async def run(self, expected_result: PlotRefreshResult):
+        self.expected_result = expected_result
+        self.expected_result_matched = False
+        self.plot_manager.trigger_refresh()
+        await time_out_assert(5, self.plot_manager.needs_refresh, value=False)
+        assert self.expected_result_matched
+
+
+@dataclass
+class Environment:
+    root_path: Path
+    refresh_tester: PlotRefreshTester
+    dir_1: Directory
+    dir_2: Directory
+
+
+@pytest.fixture(scope="function")
+def environment(tmp_path, bt) -> Iterator[Environment]:
+    dir_1_count: int = 7
+    dir_2_count: int = 3
+    plots: list[Path] = get_test_plots()
+    assert len(plots) >= dir_1_count + dir_2_count
+
+    dir_1: Directory = Directory(tmp_path / "plots" / "1", plots[0:dir_1_count])
+    dir_2: Directory = Directory(tmp_path / "plots" / "2", plots[dir_1_count : dir_1_count + dir_2_count])
+    create_default_chia_config(tmp_path)
+
+    refresh_tester = PlotRefreshTester(tmp_path)
+    refresh_tester.plot_manager.set_public_keys(bt.plot_manager.farmer_public_keys, bt.plot_manager.pool_public_keys)
+
+    yield Environment(tmp_path, refresh_tester, dir_1, dir_2)
+
+    refresh_tester.plot_manager.stop_refreshing()
+
+
+# Wrap `remove_plot` to give it the same interface as the other triggers, e.g. `add_plot_directory(Path, str)`.
+def trigger_remove_plot(_: Path, plot_path: str):
+    remove_plot(Path(plot_path))
+
+
+@pytest.mark.limit_consensus_modes(reason="not dependent on consensus, does not support parallel execution")
+@pytest.mark.anyio
+async def test_plot_refreshing(environment):
+    env: Environment = environment
+    expected_result = PlotRefreshResult()
+    dir_duplicates: Directory = Directory(get_plot_dir().resolve() / "duplicates", env.dir_1.plots)
+
+    async def run_test_case(
+        *,
+        trigger: Callable,
+        test_path: Path,
+        expect_loaded: list[MockPlotInfo],
+        expect_removed: list[Path],
+        expect_processed: int,
+        expect_duplicates: int,
+        expected_directories: int,
+        expect_total_plots: int,
+    ):
+        expected_result.loaded = cast(list[PlotInfo], expect_loaded)
+        expected_result.removed = expect_removed
+        expected_result.processed = expect_processed
+        trigger(env.root_path, str(test_path))
+        assert len(get_plot_directories(env.root_path)) == expected_directories
+        await env.refresh_tester.run(expected_result)
+        assert len(env.refresh_tester.plot_manager.plots) == expect_total_plots
+        assert len(env.refresh_tester.plot_manager.get_duplicates()) == expect_duplicates
+        assert len(env.refresh_tester.plot_manager.failed_to_open_filenames) == 0
+
+    # Add dir_1
+    await run_test_case(
+        trigger=add_plot_directory,
+        test_path=env.dir_1.path,
+        expect_loaded=env.dir_1.plot_info_list(),
+        expect_removed=[],
+        expect_processed=len(env.dir_1),
+        expect_duplicates=0,
+        expected_directories=1,
+        expect_total_plots=len(env.dir_1),
+    )
+
+    # Add dir_2
+    await run_test_case(
+        trigger=add_plot_directory,
+        test_path=env.dir_2.path,
+        expect_loaded=env.dir_2.plot_info_list(),
+        expect_removed=[],
+        expect_processed=len(env.dir_1) + len(env.dir_2),
+        expect_duplicates=0,
+        expected_directories=2,
+        expect_total_plots=len(env.dir_1) + len(env.dir_2),
+    )
+
+    # Add dir_duplicates
+    await run_test_case(
+        trigger=add_plot_directory,
+        test_path=dir_duplicates.path,
+        expect_loaded=[],
+        expect_removed=[],
+        expect_processed=len(env.dir_1) + len(env.dir_2) + len(dir_duplicates),
+        expect_duplicates=len(dir_duplicates),
+        expected_directories=3,
+        expect_total_plots=len(env.dir_1) + len(env.dir_2),
+    )
+    for item in dir_duplicates.path.iterdir():
+        assert item.is_file() and item in env.refresh_tester.plot_manager.get_duplicates()
+
+    # Drop the duplicated plot we remove in the next test case from the test directory upfront so that the numbers match
+    # the expected below
+    drop_path = dir_duplicates.plots[0]
+    dir_duplicates.drop(drop_path)
+    # Delete one duplicated plot
+    await run_test_case(
+        trigger=trigger_remove_plot,
+        test_path=drop_path,
+        expect_loaded=[],
+        expect_removed=[],
+        expect_processed=len(env.dir_1) + len(env.dir_2) + len(dir_duplicates),
+        expect_duplicates=len(dir_duplicates),
+        expected_directories=3,
+        expect_total_plots=len(env.dir_1) + len(env.dir_2),
+    )
+    # Drop the duplicated plot we remove in the next test case from the test directory upfront so that the numbers match
+    # the expected below
+    drop_path = env.dir_1.plots[0]
+    env.dir_1.drop(drop_path)
+    # Delete one duplicated plot
+    await run_test_case(
+        trigger=trigger_remove_plot,
+        test_path=drop_path,
+        expect_loaded=[],
+        expect_removed=[drop_path],
+        expect_processed=len(env.dir_1) + len(env.dir_2) + len(dir_duplicates),
+        expect_duplicates=len(dir_duplicates),
+        expected_directories=3,
+        expect_total_plots=len(env.dir_1) + len(env.dir_2),
+    )
+    # Remove directory with the duplicates
+    await run_test_case(
+        trigger=remove_plot_directory,
+        test_path=dir_duplicates.path,
+        expect_loaded=[],
+        expect_removed=[],
+        expect_processed=len(env.dir_1) + len(env.dir_2),
+        expect_duplicates=0,
+        expected_directories=2,
+        expect_total_plots=len(env.dir_1) + len(env.dir_2),
+    )
+    for item in dir_duplicates.path.iterdir():
+        assert item.is_file() and item not in env.refresh_tester.plot_manager.get_duplicates()
+
+    # Re-add the directory with the duplicates for other tests
+    await run_test_case(
+        trigger=add_plot_directory,
+        test_path=dir_duplicates.path,
+        expect_loaded=[],
+        expect_removed=[],
+        expect_processed=len(env.dir_1) + len(env.dir_2) + len(dir_duplicates),
+        expect_duplicates=len(dir_duplicates),
+        expected_directories=3,
+        expect_total_plots=len(env.dir_1) + len(env.dir_2),
+    )
+
+    # Remove dir_1 from which the duplicated plots are loaded. This removes the duplicates of dir_1
+    # and in the same run loads them from dir_duplicates.
+    await run_test_case(
+        trigger=remove_plot_directory,
+        test_path=env.dir_1.path,
+        expect_loaded=dir_duplicates.plot_info_list(),
+        expect_removed=env.dir_1.path_list(),
+        expect_processed=len(env.dir_2) + len(dir_duplicates),
+        expect_duplicates=0,
+        expected_directories=2,
+        expect_total_plots=len(env.dir_2) + len(dir_duplicates),
+    )
+
+    # Re-add the directory. Now the plot seen as duplicate is from dir_1, not from dir_duplicates like before
+    await run_test_case(
+        trigger=add_plot_directory,
+        test_path=env.dir_1.path,
+        expect_loaded=[],
+        expect_removed=[],
+        expect_processed=len(env.dir_1) + len(env.dir_2) + len(dir_duplicates),
+        expect_duplicates=len(dir_duplicates),
+        expected_directories=3,
+        expect_total_plots=len(env.dir_1) + len(env.dir_2),
+    )
+    # Drop the duplicated plot we remove in the next test case from the test directory upfront so that the numbers match
+    # the expected below
+    drop_path = env.dir_1.plots[2]
+    env.dir_1.drop(drop_path)
+    # Remove the duplicated plot
+    await run_test_case(
+        trigger=trigger_remove_plot,
+        test_path=drop_path,
+        expect_loaded=[],
+        expect_removed=[],
+        expect_processed=len(env.dir_1) + len(env.dir_2) + len(dir_duplicates),
+        expect_duplicates=len(env.dir_1),
+        expected_directories=3,
+        expect_total_plots=len(env.dir_2) + len(dir_duplicates),
+    )
+    # Remove dir_duplicates, this drops the duplicates and loads all plots from dir_1
+    await run_test_case(
+        trigger=remove_plot_directory,
+        test_path=dir_duplicates.path,
+        expect_loaded=env.dir_1.plot_info_list(),
+        expect_removed=dir_duplicates.path_list(),
+        expect_processed=len(env.dir_1) + len(env.dir_2),
+        expect_duplicates=0,
+        expected_directories=2,
+        expect_total_plots=len(env.dir_1) + len(env.dir_2),
+    )
+    # Remove dir_2
+    await run_test_case(
+        trigger=remove_plot_directory,
+        test_path=env.dir_2.path,
+        expect_loaded=[],
+        expect_removed=env.dir_2.path_list(),
+        expect_processed=len(env.dir_1),
+        expect_duplicates=0,
+        expected_directories=1,
+        expect_total_plots=len(env.dir_1),
+    )
+    # Remove dir_1
+    await run_test_case(
+        trigger=remove_plot_directory,
+        test_path=env.dir_1.path,
+        expect_loaded=[],
+        expect_removed=env.dir_1.path_list(),
+        expect_processed=0,
+        expect_duplicates=0,
+        expected_directories=0,
+        expect_total_plots=0,
+    )
+
+
+@pytest.mark.anyio
+async def test_initial_refresh_flag(environment: Environment) -> None:
+    env: Environment = environment
+    assert env.refresh_tester.plot_manager.initial_refresh()
+    for _ in range(2):
+        await env.refresh_tester.run(PlotRefreshResult())
+        assert not env.refresh_tester.plot_manager.initial_refresh()
+    env.refresh_tester.plot_manager.reset()
+    assert env.refresh_tester.plot_manager.initial_refresh()
+
+
+@pytest.mark.anyio
+async def test_invalid_plots(environment):
+    env: Environment = environment
+    expected_result = PlotRefreshResult()
+    # Test re-trying if processing a plot failed
+    # First create a backup of the plot
+    retry_test_plot = env.dir_1.path_list()[0].resolve()
+    retry_test_plot_save = Path(env.dir_1.path / ".backup").resolve()
+    copy(retry_test_plot, retry_test_plot_save)
+    # Invalidate the plot
+    with open(retry_test_plot, "r+b") as file:
+        file.write(bytes(100))
+    # Add it and validate it fails to load
+    add_plot_directory(env.root_path, str(env.dir_1.path))
+    expected_result.loaded = env.dir_1.plot_info_list()[1:]
+    expected_result.removed = []
+    expected_result.processed = len(env.dir_1)
+    expected_result.remaining = 0
+    await env.refresh_tester.run(expected_result)
+    assert len(env.refresh_tester.plot_manager.failed_to_open_filenames) == 1
+    assert retry_test_plot in env.refresh_tester.plot_manager.failed_to_open_filenames
+    # Give it a non .plot ending and make sure it gets removed from the invalid list on the next refresh
+    retry_test_plot_unload = Path(env.dir_1.path / ".unload").resolve()
+    move(retry_test_plot, retry_test_plot_unload)
+    expected_result.processed -= 1
+    expected_result.loaded = []
+    await env.refresh_tester.run(expected_result)
+    assert len(env.refresh_tester.plot_manager.failed_to_open_filenames) == 0
+    assert retry_test_plot not in env.refresh_tester.plot_manager.failed_to_open_filenames
+    # Recover the name and make sure it reappears in the invalid list
+    move(retry_test_plot_unload, retry_test_plot)
+    expected_result.processed += 1
+    await env.refresh_tester.run(expected_result)
+    assert len(env.refresh_tester.plot_manager.failed_to_open_filenames) == 1
+    assert retry_test_plot in env.refresh_tester.plot_manager.failed_to_open_filenames
+    # Make sure the file stays in `failed_to_open_filenames` and doesn't get loaded in the next refresh cycle
+    expected_result.loaded = []
+    expected_result.processed = len(env.dir_1)
+    await env.refresh_tester.run(expected_result)
+    assert len(env.refresh_tester.plot_manager.failed_to_open_filenames) == 1
+    assert retry_test_plot in env.refresh_tester.plot_manager.failed_to_open_filenames
+    # Now decrease the re-try timeout, restore the valid plot file and make sure it properly loads now
+    env.refresh_tester.plot_manager.refresh_parameter = replace(
+        env.refresh_tester.plot_manager.refresh_parameter, retry_invalid_seconds=uint32(0)
+    )
+    move(retry_test_plot_save, retry_test_plot)
+    expected_result.loaded = env.dir_1.plot_info_list()[0:1]
+    expected_result.processed = len(env.dir_1)
+    await env.refresh_tester.run(expected_result)
+    assert len(env.refresh_tester.plot_manager.failed_to_open_filenames) == 0
+    assert retry_test_plot not in env.refresh_tester.plot_manager.failed_to_open_filenames
+
+
+@pytest.mark.anyio
+async def test_keys_missing(environment: Environment) -> None:
+    env: Environment = environment
+    not_in_keychain_plots: list[Path] = get_test_plots("not_in_keychain")
+    dir_not_in_keychain: Directory = Directory(env.root_path / "plots" / "not_in_keychain", not_in_keychain_plots)
+    expected_result = PlotRefreshResult()
+    # The plots in "not_in_keychain" directory have infinity g1 elements as farmer/pool key so they should be plots
+    # with missing keys for now
+    add_plot_directory(env.root_path, str(dir_not_in_keychain.path))
+    expected_result.loaded = []
+    expected_result.removed = []
+    expected_result.processed = len(dir_not_in_keychain)
+    expected_result.remaining = 0
+    for i in range(2):
+        await env.refresh_tester.run(expected_result)
+        assert len(env.refresh_tester.plot_manager.no_key_filenames) == len(dir_not_in_keychain)
+        for path in env.refresh_tester.plot_manager.no_key_filenames:
+            assert path in dir_not_in_keychain.plots
+    # Delete one of the plots and make sure it gets dropped from the no key filenames list
+    drop_plot = dir_not_in_keychain.path_list()[0]
+    dir_not_in_keychain.drop(drop_plot)
+    drop_plot.unlink()
+    assert drop_plot in env.refresh_tester.plot_manager.no_key_filenames
+    expected_result.processed -= 1
+    await env.refresh_tester.run(expected_result)
+    assert drop_plot not in env.refresh_tester.plot_manager.no_key_filenames
+    # Now add the missing keys to the plot manager's key lists and make sure the plots are getting loaded
+    env.refresh_tester.plot_manager.farmer_public_keys.append(G1Element())
+    env.refresh_tester.plot_manager.pool_public_keys.append(G1Element())
+    expected_result.loaded = dir_not_in_keychain.plot_info_list()  # type: ignore[assignment]
+    expected_result.processed = len(dir_not_in_keychain)
+    await env.refresh_tester.run(expected_result)
+    # And make sure they are dropped from the list of plots with missing keys
+    assert len(env.refresh_tester.plot_manager.no_key_filenames) == 0
+
+
+@pytest.mark.anyio
+async def test_plot_info_caching(environment, bt):
+    env: Environment = environment
+    expected_result = PlotRefreshResult()
+    add_plot_directory(env.root_path, str(env.dir_1.path))
+    expected_result.loaded = env.dir_1.plot_info_list()
+    expected_result.removed = []
+    expected_result.processed = len(env.dir_1)
+    expected_result.remaining = 0
+    await env.refresh_tester.run(expected_result)
+    assert env.refresh_tester.plot_manager.cache.path().exists()
+    unlink(env.refresh_tester.plot_manager.cache.path())
+    # Should not write the cache again on shutdown because it didn't change
+    assert not env.refresh_tester.plot_manager.cache.path().exists()
+    env.refresh_tester.plot_manager.stop_refreshing()
+    assert not env.refresh_tester.plot_manager.cache.path().exists()
+    # Manually trigger `save_cache` and make sure it creates a new cache file
+    env.refresh_tester.plot_manager.cache.save()
+    assert env.refresh_tester.plot_manager.cache.path().exists()
+    refresh_tester: PlotRefreshTester = PlotRefreshTester(env.root_path)
+    plot_manager = refresh_tester.plot_manager
+    plot_manager.set_public_keys(bt.plot_manager.farmer_public_keys, bt.plot_manager.pool_public_keys)
+    plot_manager.cache.load()
+    assert len(plot_manager.cache) == len(env.refresh_tester.plot_manager.cache)
+    for path, cache_entry in env.refresh_tester.plot_manager.cache.items():
+        cache_entry_new = plot_manager.cache.get(path)
+        assert bytes(cache_entry_new.prover) == bytes(cache_entry.prover)
+        assert cache_entry_new.farmer_public_key == cache_entry.farmer_public_key
+        assert cache_entry_new.pool_public_key == cache_entry.pool_public_key
+        assert cache_entry_new.pool_contract_puzzle_hash == cache_entry.pool_contract_puzzle_hash
+        assert cache_entry_new.plot_public_key == cache_entry.plot_public_key
+    await refresh_tester.run(expected_result)
+    for path, plot_info in env.refresh_tester.plot_manager.plots.items():
+        assert path in plot_manager.plots
+        assert plot_manager.plots[path].prover.get_filename() == plot_info.prover.get_filename()
+        assert plot_manager.plots[path].prover.get_id() == plot_info.prover.get_id()
+        assert plot_manager.plots[path].prover.get_memo() == plot_info.prover.get_memo()
+        assert plot_manager.plots[path].prover.get_size() == plot_info.prover.get_size()
+        assert plot_manager.plots[path].prover.get_compression_level() == plot_info.prover.get_compression_level()
+        assert plot_manager.plots[path].pool_public_key == plot_info.pool_public_key
+        assert plot_manager.plots[path].pool_contract_puzzle_hash == plot_info.pool_contract_puzzle_hash
+        assert plot_manager.plots[path].plot_public_key == plot_info.plot_public_key
+        assert plot_manager.plots[path].file_size == plot_info.file_size
+        assert plot_manager.plots[path].time_modified == plot_info.time_modified
+    assert plot_manager.plot_filename_paths == env.refresh_tester.plot_manager.plot_filename_paths
+    assert plot_manager.failed_to_open_filenames == env.refresh_tester.plot_manager.failed_to_open_filenames
+    assert plot_manager.no_key_filenames == env.refresh_tester.plot_manager.no_key_filenames
+    plot_manager.stop_refreshing()
+    # Modify the content of the plot_manager.dat
+    with open(plot_manager.cache.path(), "r+b") as file:
+        file.write(b"\xff\xff")  # Sets Cache.version to 65535
+    # Make sure it just loads the plots normally if it fails to load the cache
+    refresh_tester: PlotRefreshTester = PlotRefreshTester(env.root_path)
+    plot_manager = refresh_tester.plot_manager
+    plot_manager.cache.load()
+    assert len(plot_manager.cache) == 0
+    plot_manager.set_public_keys(bt.plot_manager.farmer_public_keys, bt.plot_manager.pool_public_keys)
+    await refresh_tester.run(expected_result)
+    assert len(plot_manager.plots) == len(plot_manager.plots)
+    plot_manager.stop_refreshing()
+
+
+@pytest.mark.anyio
+async def test_drop_too_large_cache_entries(environment, bt):
+    env: Environment = environment
+    expected_result = PlotRefreshResult(loaded=env.dir_1.plot_info_list(), processed=len(env.dir_1))
+    add_plot_directory(env.root_path, str(env.dir_1.path))
+    await env.refresh_tester.run(expected_result)
+    assert env.refresh_tester.plot_manager.cache.path().exists()
+    assert len(env.dir_1) >= 6, "This test requires at least 6 cache entries"
+    # Load the cache entries
+    cache_path = env.refresh_tester.plot_manager.cache.path()
+    serialized = cache_path.read_bytes()
+    stored_cache: VersionedBlob = VersionedBlob.from_bytes(serialized)
+    cache_data: CacheDataV1 = CacheDataV1.from_bytes(stored_cache.blob)
+
+    def modify_cache_entry(index: int, additional_data: int, modify_memo: bool) -> str:
+        path, cache_entry = cache_data.entries[index]
+        prover_data = cache_entry.prover_data
+        # Size of length hints in chiapos serialization currently depends on the platform
+        size_length = 8 if sys.maxsize > 2**32 else 4
+        # Version
+        version_size = 2
+        version = prover_data[0:version_size]
+        # Filename
+        filename_offset = version_size + size_length
+        filename_length = int.from_bytes(prover_data[version_size:filename_offset], byteorder=sys.byteorder)
+        filename = prover_data[filename_offset : filename_offset + filename_length]
+        # Memo
+        memo_length_offset = filename_offset + filename_length
+        memo_length = int.from_bytes(
+            prover_data[memo_length_offset : memo_length_offset + size_length], byteorder=sys.byteorder
+        )
+        memo_offset = memo_length_offset + size_length
+        memo = prover_data[memo_offset : memo_offset + memo_length]
+        # id, k, table pointers, C2
+        remainder = prover_data[memo_offset + memo_length :]
+
+        # Add the additional data to the filename
+        filename_length += additional_data
+        filename += bytes(b"\a" * additional_data)
+
+        # Add the additional data to the memo if requested
+        if modify_memo:
+            memo_length += additional_data
+            memo += bytes(b"\b" * additional_data)
+
+        filename_length_bytes = filename_length.to_bytes(size_length, byteorder=sys.byteorder)
+        memo_length_bytes = memo_length.to_bytes(size_length, byteorder=sys.byteorder)
+
+        cache_data.entries[index] = (
+            path,
+            replace(
+                cache_entry,
+                prover_data=bytes(version + filename_length_bytes + filename + memo_length_bytes + memo + remainder),
+            ),
+        )
+        return path
+
+    def assert_cache(expected: list[MockPlotInfo]) -> None:
+        test_cache = Cache(cache_path)
+        assert len(test_cache) == 0
+        test_cache.load()
+        assert len(test_cache) == len(expected)
+        for plot_info in expected:
+            assert test_cache.get(Path(plot_info.prover.get_filename())) is not None
+
+    # Modify two entries, with and without memo modification, they both should remain in the cache after load
+    modify_cache_entry(0, 1500, modify_memo=False)
+    modify_cache_entry(1, 1500, modify_memo=True)
+
+    invalid_entries = [
+        modify_cache_entry(2, 2000, modify_memo=False),
+        modify_cache_entry(3, 2000, modify_memo=True),
+        modify_cache_entry(4, 50000, modify_memo=False),
+        modify_cache_entry(5, 50000, modify_memo=True),
+    ]
+
+    plot_infos = env.dir_1.plot_info_list()
+    # Make sure the cache currently contains all plots from dir1
+    assert_cache(plot_infos)
+    # Write the modified cache entries to the file
+    cache_path.write_bytes(bytes(VersionedBlob(uint16(CURRENT_VERSION), bytes(cache_data))))
+    # And now test that plots in invalid_entries are not longer loaded
+    assert_cache([plot_info for plot_info in plot_infos if plot_info.prover.get_filename() not in invalid_entries])
+
+
+@pytest.mark.anyio
+async def test_cache_lifetime(environment: Environment) -> None:
+    # Load a directory to produce a cache file
+    env: Environment = environment
+    expected_result = PlotRefreshResult()
+    add_plot_directory(env.root_path, str(env.dir_1.path))
+    expected_result.loaded = env.dir_1.plot_info_list()  # type: ignore[assignment]
+    expected_result.removed = []
+    expected_result.processed = len(env.dir_1)
+    expected_result.remaining = 0
+    await env.refresh_tester.run(expected_result)
+    expected_result.loaded = []
+    cache_v1: Cache = env.refresh_tester.plot_manager.cache
+    assert len(cache_v1) > 0
+    count_before = len(cache_v1)
+    # Remove half of the plots in dir1
+    for path in env.dir_1.path_list()[0 : int(len(env.dir_1) / 2)]:
+        expected_result.processed -= 1
+        expected_result.removed.append(path)
+        unlink(path)
+    # Modify the `last_use` timestamp of all cache entries to let them expire
+    last_use_before = time.time() - Cache.expiry_seconds - 1
+    for cache_entry in cache_v1.values():
+        cache_entry.last_use = last_use_before
+        assert cache_entry.expired(Cache.expiry_seconds)
+    # The next refresh cycle will now lead to half of the cache entries being removed because they are expired and
+    # the related plots do not longer exist.
+    await env.refresh_tester.run(expected_result)
+    assert len(cache_v1) == count_before - len(expected_result.removed)
+    # The other half of the cache entries should have a different `last_use` value now.
+    for cache_entry in cache_v1.values():
+        assert cache_entry.last_use != last_use_before
+
+
+@pytest.mark.parametrize(
+    ["event_to_raise"],
+    [
+        pytest.param(PlotRefreshEvents.started, id="started"),
+        pytest.param(PlotRefreshEvents.batch_processed, id="batch_processed"),
+        pytest.param(PlotRefreshEvents.done, id="done"),
+    ],
+)
+@pytest.mark.anyio
+async def test_callback_event_raises(environment, event_to_raise: PlotRefreshEvents):
+    last_event_fired: Optional[PlotRefreshEvents] = None
+
+    def raising_callback(event: PlotRefreshEvents, _: PlotRefreshResult):
+        nonlocal last_event_fired
+        last_event_fired = event
+        if event == event_to_raise:
+            raise Exception(f"run_raise_in_callback {event_to_raise}")
+
+    env: Environment = environment
+    expected_result = PlotRefreshResult()
+    # Load dir_1
+    add_plot_directory(env.root_path, str(env.dir_1.path))
+    expected_result.loaded = env.dir_1.plot_info_list()  # type: ignore[assignment]
+    expected_result.removed = []
+    expected_result.processed = len(env.dir_1)
+    expected_result.remaining = 0
+    await env.refresh_tester.run(expected_result)
+    # Load dir_2
+    add_plot_directory(env.root_path, str(env.dir_2.path))
+    expected_result.loaded = env.dir_2.plot_info_list()  # type: ignore[assignment]
+    expected_result.removed = []
+    expected_result.processed = len(env.dir_1) + len(env.dir_2)
+    expected_result.remaining = 0
+    await env.refresh_tester.run(expected_result)
+    # Now raise the exception in the callback
+    default_callback = env.refresh_tester.plot_manager._refresh_callback
+    env.refresh_tester.plot_manager.set_refresh_callback(raising_callback)
+    env.refresh_tester.plot_manager.start_refreshing()
+    env.refresh_tester.plot_manager.trigger_refresh()
+    await time_out_assert(5, env.refresh_tester.plot_manager.needs_refresh, value=False)
+    # And make sure the follow-up evens aren't fired
+    assert last_event_fired == event_to_raise
+    # The exception should trigger `PlotManager.reset()` and clear the plots
+    assert len(env.refresh_tester.plot_manager.plots) == 0
+    assert len(env.refresh_tester.plot_manager.plot_filename_paths) == 0
+    assert len(env.refresh_tester.plot_manager.failed_to_open_filenames) == 0
+    assert len(env.refresh_tester.plot_manager.no_key_filenames) == 0
+    # The next run without the valid callback should lead to re-loading of all plot
+    env.refresh_tester.plot_manager.set_refresh_callback(default_callback)
+    expected_result.loaded = env.dir_1.plot_info_list() + env.dir_2.plot_info_list()  # type: ignore[assignment]
+    expected_result.removed = []
+    expected_result.processed = len(env.dir_1) + len(env.dir_2)
+    expected_result.remaining = 0
+    await env.refresh_tester.run(expected_result)
+
+
+@pytest.mark.anyio
+async def test_recursive_plot_scan(environment: Environment) -> None:
+    env: Environment = environment
+    # Create a directory tree with some subdirectories containing plots, others not.
+    root_plot_dir = env.root_path / "root"
+    sub_dir_0: Directory = Directory(root_plot_dir / "0", env.dir_1.plots[0:2])
+    sub_dir_0_1: Directory = Directory(sub_dir_0.path / "1", env.dir_1.plots[2:3])
+    sub_dir_1: Directory = Directory(root_plot_dir / "1", [])
+    sub_dir_1_0: Directory = Directory(sub_dir_1.path / "0", [])
+    sub_dir_1_0_1: Directory = Directory(sub_dir_1_0.path / "1", env.dir_1.plots[3:7])
+
+    # List of all the plots in the directory tree
+    expected_plot_list = sub_dir_0.plot_info_list() + sub_dir_0_1.plot_info_list() + sub_dir_1_0_1.plot_info_list()
+
+    # Adding the root without `recursive_plot_scan` and running a test should not load any plots (match an empty result)
+    expected_result = PlotRefreshResult()
+    add_plot_directory(env.root_path, str(root_plot_dir))
+    await env.refresh_tester.run(expected_result)
+
+    # Set the recursive scan flag in the config
+    with lock_and_load_config(env.root_path, "config.yaml") as config:
+        config["harvester"]["recursive_plot_scan"] = True
+        save_config(env.root_path, "config.yaml", config)
+
+    # With the flag enabled it should load all expected plots
+    expected_result.loaded = expected_plot_list  # type: ignore[assignment]
+    expected_result.processed = len(expected_plot_list)
+    await env.refresh_tester.run(expected_result)
+
+    # Adding the subdirectories also should not lead to some failure or duplicated loading
+    add_plot_directory(env.root_path, str(sub_dir_0_1.path))
+    add_plot_directory(env.root_path, str(sub_dir_1_0_1.path))
+    expected_result.loaded = []
+    await env.refresh_tester.run(expected_result)
+
+
+@boolean_datacases(name="follow_links", false="no_follow", true="follow")
+@pytest.mark.anyio
+async def test_recursive_plot_scan_symlinks(environment: Environment, follow_links: bool) -> None:
+    env: Environment = environment
+    # Create a directory tree with some subdirectories containing plots, others not.
+    root_plot_dir = env.root_path / "root"
+    sub_dir_0: Directory = Directory(root_plot_dir / "0", env.dir_1.plots[0:2])
+    sub_dir_0_1: Directory = Directory(sub_dir_0.path / "1", env.dir_1.plots[2:3])
+
+    # Create a plot directory outside of the root plot directory
+    other_plot_dir: Directory = Directory(env.root_path / "other", env.dir_1.plots[3:7])
+
+    # Create a symlink to the other plot directory from inside the root plot directory
+    symlink_path = Path(root_plot_dir / "other")
+    symlink_path.symlink_to(env.root_path / "other")
+
+    # Adding the root without `recursive_plot_scan` and running a test should not load any plots (match an empty result)
+    expected_result = PlotRefreshResult()
+    add_plot_directory(env.root_path, str(root_plot_dir))
+    await env.refresh_tester.run(expected_result)
+
+    if follow_links:
+        expected_plot_list = sub_dir_0.plot_info_list() + sub_dir_0_1.plot_info_list() + other_plot_dir.plot_info_list()
+    else:
+        expected_plot_list = sub_dir_0.plot_info_list() + sub_dir_0_1.plot_info_list()
+
+    # Set the recursive scan flag in the config and symlink flag
+    with lock_and_load_config(env.root_path, "config.yaml") as config:
+        config["harvester"]["recursive_plot_scan"] = True
+        config["harvester"]["recursive_follow_links"] = follow_links
+        save_config(env.root_path, "config.yaml", config)
+
+    # With the flag enabled it should load all expected plots
+    expected_result.loaded = expected_plot_list  # type: ignore[assignment]
+    expected_result.processed = len(expected_plot_list)
+    await env.refresh_tester.run(expected_result)