it's not recommended to suppress warnings or errors,
but the user may decide this on his own.
note: --warning is not given to borg serve so a <= 1.0.0 borg
will still work as server. it is not needed as it is the default.
(Remote)Repository.close() is not a public API anymore, but a private
API. It shall not be used from within other classes than Repository
or it's tests. The proper way is to use a context manager now. However,
for RPC/Remote compatibility with Borg 1.0 it is kept and unchanged.
Repositories are not opened by __init__ now anymore, it is done
by binding it to a context manager. (This SHOULD be compatible both ways
with remote, since opening the repo is handled by a RepositoryServer method)
Decorators @with_repository() and @with_archive simplify
context manager handling and remove unnecessary indentation.
Backported to 1.0-maint
[1]
This worked incidentally because OSes tend to return at least one page
worth of data when EOF is not reached. Increasing WINDOW_SIZE beyond
the page size might have lead to data loss.
[2]
If read() of the passed Python object returned something not-bytes,
PyBytes_Size returns -1 (ssize_t) which becomes a very larger number for
memcpy()s size_t.
Previously, on 'borg diff', the output always had first the modifications, then
additions, and finally removals. Output may be easier to follow if the various
kinds of changes are interleaved. This commit is a simple solution that first
collects the output lines and sorts them by file path before printing. This new
behavior is optional and disabled by default. It can be enabled with '--sort'
command line option.
This option will be especially useful after the planned multi-threading changes
arrive. Multi-threading may shuffle the archive order of files making diff
output hard to follow without sorting.
Resolves#797.
New repository config item, repository.append_only, causes Borg to never
delete or append to existing data files. Hints and indices are handled
as before, old ones are deleted, because they can get quite large, but
are automatically reconstructed: no need to keep them.
When append_only is activated a file /path/to/repo/transactions/<NUMBER>
will be created for every commit.
Deleting all segments <NUMBER+1> and higher will rollback to that commit.
Note that this only influences Borg behaviour. Since repository config
can't be altered remotely (except for repository.key) this can't be
disabled when accessed remotely over SSH with "borg serve" as the
forced command.
This is only a feature to support the use case, and does not replace
appropriate file system permissions or monitoring.
Resolves#809
Makes it easy to see paths excluded by --exclude* options for testing of
regexes, and for ongoing monitoring that files desired for backup aren't
getting excluded accidentally.
AESKeyBase.decrypt makes less copies of the data, specifically data[41:],
which copies the payload, is gone. This is possible since this commit
makes our lil' crypto API compatible with simple buffers.
In Cython there is a syntax for creating memoryviews from buffer-enabled
objects, however, it doesn't support read-only buffers. ro_buffer gets
this job done, though, and also does the proper type checking
(by PyBUF_SIMPLE).
Note: msgpack doesn't support memoryviews. Otherwise we could avoid
copying the en/de-crypted output with the final "return out[:ctl]"
This commit leads to a solid 10-15 % increase in performance of hash-lists
and extract.
The hmac_sha256() function releases the GIL.
The main design goals of the new format:
- One file takes exactly one line of output
- The format is easy to read with typical, long list of changes
- Metadata changes are visible and easy to examine
- Unuseful information is not shown
Resolves#757.
Marian Beermann