Struct TransactionManager 

pub struct TransactionManager { /* private fields */ }

Manages copy-on-write commits and root pointer alternation.

Workflow:

1. Filesystem operations allocate new blocks and write objects.
2. When ready to commit, call commit() with the new superblock.
3. TransactionManager writes the superblock to a fresh block, then updates the next root pointer slot (A or B).
4. Old blocks can then be freed (deferred GC - not yet implemented).

Implementations

impl TransactionManager

pub fn new() -> Self

pub fn from_recovered(generation: u64, last_was_b: bool) -> Self

Initialize from recovered state.

pub fn generation(&self) -> u64

Current generation.

pub fn commit( &mut self, store: &dyn BlockStore, crypto: &dyn CryptoEngine, codec: &PostcardCodec, allocator: &dyn SlotAllocator, superblock: &Superblock, ) -> FsResult<()>

Commit a new superblock.

1. Allocate a block for the superblock.
2. Write the encrypted superblock to that block.
3. Write the root pointer to the next slot (A or B), unencrypted (root pointers are integrity-checked via BLAKE3 but not encrypted so we can read them without the key to find the superblock).

Note: root pointers are written unencrypted but with a checksum. The superblock itself is encrypted.

pub fn read_root_pointer( store: &dyn BlockStore, codec: &PostcardCodec, slot: u64, ) -> FsResult<Option<RootPointer>>

Try to read a root pointer from a slot. Returns None if the slot is empty/invalid.

pub fn recover_latest( store: &dyn BlockStore, codec: &PostcardCodec, ) -> FsResult<Option<(RootPointer, bool)>>

Recover the latest valid root pointer by reading both slots. Returns (RootPointer, was_slot_b).