Releases: zen-fs/core
Releases · zen-fs/core
0.18.0
- Implemented the watch API/features (#6)
- Permissions are now handled in the emulation layer, not by individual backends
- Removed redundant permissions checks in backends
- Credentials are no longer passed to
FileSystem
methods- This change greatly simplifies the
FileSystem
API, making it easier to write backends - It is also a major breaking change for the internal API
- This change greatly simplifies the
- Backends that need them can import
credentials
- Simplified credentials handling to use the exported
credentials
object - Credentials were already managed globally
- Corrected the
syscall
in someErrnoError.With
calls _openSync
now truncates the file instead of deleting and recreating it (when the flag is truncating)- Clarified
link
parameter names PortFile.stat
now returns aStats
instancelink
correctly throws an error when attempting to make a cross "device"/backend link- Simplified directory listing encoding/decoding
- Added deadlock avoidance to
Mutexed
- Fixed
Dir._entries
initialization - Added missing early return in
mkdir
0.17.1
0.17.0
- Changed
LockedFS
into a mixin namedMutexed
- Removed
Locked
backend - Removed
unlock
method
- Removed
- Fixed race condition after
Mutex
unlock (#78, Thanks @terryluan12) - Removed
OverlayFS
methods:getOverlayedFileSystems
,resDeletionLog
,unwrap
- Changed
OverlayFS
protected_readable
and_writable
properties to public readonlyreadable
andwriteable
- Renamed
UnlockedOverlayFS
toUnmutexedOverlayFS
- Backends' types are no longer expanded
- This means a much cleaner and readable error message with an incorrect configuration for a mount in
configure
andconfigureSingle
- This is exclusively an aesthetic change
- This means a much cleaner and readable error message with an incorrect configuration for a mount in
- Implemented
fs.watch
andfs.promises.watch
- Note that the internal events have not been added, so no events will be emitted
- Fully added
File.close
functionality- After closing a file, most operations no longer work.
Index
constructor now accepts entriesopendir
now preloadsDir
entries- Fixed double close in
_readFileSync
- Fixed being able to
abort
aTransaction
after it was alreadycommit
ed - Changed
Transaction
disposal toabort
instead ofcommit
- Updated
StoreFS
methods to useusing
on transaction instances
File structure changes:
- Consolidated
backends/index/*
intobackends/file_index
- Organized FS mixins into
mixins
0.16.4
0.16.3
0.16.2
- Fixed unavailable backends throwing the backend object instead of the name
- Fixed
resolveMountConfig
notawait
ingcheckOptions
- Changed internal
Offset
enum tooffsets
object (forInode
) UnlockedOverlayFS.restoreDeletionLog
anddeletePath
are nowasync
- Added runtime checks for
IndexFS
directory contents - Fixed some types being
any
in callbacks API emulation - Added a runtime check for
globalThis.ReadableStream
toFileHandle.readableWebStream
- Fixed sync tests being
async
functions - A shared Eslint config is now exported
- Upgraded to the flag Eslint config