|
|
@@ -144,8 +144,8 @@ Compatibility notes:
|
|
|
Change Log 2.x
|
|
|
==============
|
|
|
|
|
|
-Version 2.0.0b18 (not released yet)
|
|
|
------------------------------------
|
|
|
+Version 2.0.0b18 (2025-06-19)
|
|
|
+-----------------------------
|
|
|
|
|
|
Please note:
|
|
|
|
|
|
@@ -156,7 +156,29 @@ above.
|
|
|
|
|
|
New features:
|
|
|
|
|
|
-- new "buzhash64" chunker, init bh table using a 256bit key derived from ID key
|
|
|
+- experimental new "buzhash64" chunker (later, after testing, this shall become
|
|
|
+ the default chunker in borg2):
|
|
|
+
|
|
|
+ - add own cryptographically secure pseudo-random number generator (CSPRNG)
|
|
|
+ based on AES256-CTR to create deterministic random, based on a 256bit seed.
|
|
|
+ - use that to deterministically create a perfectly balanced buzhash64 table.
|
|
|
+ - "buzhash64" chunker computes 64bit hash values for the chunking decision.
|
|
|
+ - performance is similar to "buzhash" (measured on Apple M3P cpu).
|
|
|
+
|
|
|
+ That should also resolve these points of criticism about the old "buzhash"
|
|
|
+ 32bit code:
|
|
|
+
|
|
|
+ - table_base: that the bits are not randomly distributed enough
|
|
|
+ - that an XORed seed cancels out for specific window sizes
|
|
|
+ - that XORing the table with a seed is equivalent to XORing the computed hash
|
|
|
+ value with another constant
|
|
|
+
|
|
|
+ Please test the chunkers extensively (e.g. with borg create, borg transfer),
|
|
|
+ we can hardly change them "in production", because chunking differently also
|
|
|
+ means not deduplicating with old chunks. So, in case there are changes
|
|
|
+ needed, we need to find and fix them now while borg is in beta.
|
|
|
+
|
|
|
+ See also some other chunker changes listed below "Other changes".
|
|
|
- serve: add --permissions option as an alternative to BORG_REPO_PERMISSIONS env var
|
|
|
- create: auto-exclude items based on xattrs or NODUMP, see #4972
|
|
|
|
|
|
@@ -165,15 +187,16 @@ New features:
|
|
|
|
|
|
also: create: read stat attrs, xattrs, ACLs early, before file contents.
|
|
|
|
|
|
+Fixes:
|
|
|
+
|
|
|
+- compact: fix cleaning archives directory (catch correct exception, use
|
|
|
+ logger.warning, improve error msg)
|
|
|
+
|
|
|
Other changes:
|
|
|
|
|
|
- support Python 3.14
|
|
|
-- msgpack: allow 1.1.1
|
|
|
-- msgpack version check: ignore "rc" or other version elements
|
|
|
+- msgpack: allow 1.1.1, version check: ignore "rc" or other version elements
|
|
|
- add derive_key to derive new keys from existing key material
|
|
|
-- fish: fix archive name completion
|
|
|
-- refactor: modularize tests
|
|
|
-- refactor: use pathlib.Path
|
|
|
- refactor the chunkers, #8882 #8883:
|
|
|
|
|
|
- transform buzhash chunker C code to Cython
|
|
|
@@ -183,12 +206,15 @@ Other changes:
|
|
|
files and fmaps.
|
|
|
- FileReader uses FileFMAPReader to fill its buffer and offers clients a
|
|
|
`.read(size)` method so they can read pieces of the data.
|
|
|
- - both chunkers now use the FileReader/FileFMAPReader code
|
|
|
+ - all chunkers now use the FileReader/FileFMAPReader code
|
|
|
- split code and test module into packages
|
|
|
-- ChunkerFixed: add fixed chunker tests to selftest
|
|
|
-- ChunkerFixed: do not assert on short header read
|
|
|
-- Chunker: use safe_fadvise
|
|
|
-- ChunkerParams: reject even window size for buzhash, #8868
|
|
|
+- "fixed" chunker: add fixed chunker tests to selftest
|
|
|
+- "fixed" chunker: do not assert on short header read
|
|
|
+- "buzhash*" chunker: use safe_fadvise
|
|
|
+- "buzhash" chunker: reject even window size, #8868
|
|
|
+- fish: fix archive name completion
|
|
|
+- refactor: modularize tests
|
|
|
+- refactor: use pathlib.Path
|
|
|
- tests / CI:
|
|
|
|
|
|
- CI: add bandit, a security-oriented static analysis tool
|
Thomas Waldmann