copy: copy an archive to a new archive name, #2300

[ThomasWaldmann]

Adds a borg copy OLDNAME NEWNAME command, addressing #2300 (the "additional names for an existing archive" request).

Rationale

#2300 asked for archive aliases, sketched as a name -> archive dict in the manifest. In borg2 that design is both impossible and unnecessary:

• Impossible: each archive is a archives/<hex-id> entry in borgstore, keyed by the id of the archive metadata object, and the name lives inside that id-hashed object. Two names cannot point at one archive object.
• Unnecessary: borg2 does not refcount chunks. borg compact only frees chunks that no remaining archive references, so two archives sharing all their chunks via dedup already behave the way the issue wanted -- deleting one leaves the other fully intact, with no dangling-pointer failure mode that a symlink-style alias would have.

So instead of an alias, this creates a real, independent second archive -- which costs almost nothing.

Implementation

Archive.copy() is Archive.rename() without removing the original archive entry: it writes a new archive metadata object (via set_meta("name", ...)) and leaves the old directory entry in place. No file content is read or written, and item_ptrs is carried over unchanged, so the item metadata stream and every content chunk are shared.

The command is modelled on rename_cmd.py and uses @with_archive, so OLDNAME resolution (name if unique, or aid:<hex> prefix) and aid:-aware shell completion come for free.

Copying an archive to its own name is refused: the new metadata would be byte-identical, hence the same archive id, hence it would just overwrite the existing directory entry -- a silent no-op rather than a copy.

Measured behaviour

On a repo with a single 2 MB random-data archive:

step	repo size
after borg create	2036 KB
after borg copy	2044 KB
after borg delete <original> + borg compact	2048 KB

A control run without the copy drops to 24 KB at the last step, so compact really is freeing the data in the absence of a second referrer. After deleting the original, borg check is clean and the copy extracts byte-identical.

Tests

Seven tests in src/borg/testsuite/archiver/copy_cmd_test.py: basic copy, copy by aid:, shared item_ptrs/timestamp, delete-original-then-compact-then-extract, copying into an existing archive series, copy-to-same-name refused, ambiguous OLDNAME refused.

Notes for review

• docs/man/borg-copy.1 is included (following the borg analyze commit); the date-only drift build_man produces in all the other man pages is not.
• The changelog entry went under 2.0.0b24.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 87.31%. Comparing base (0ad66ed) to head (9b89c94).
⚠️ Report is 34 commits behind head on master.
✅ All tests successful. No failed tests found.

Additional details and impacted files

@@            Coverage Diff             @@
##           master   #10221      +/-   ##
==========================================
+ Coverage   87.29%   87.31%   +0.02%     
==========================================
  Files         102      103       +1     
  Lines       18444    18480      +36     
  Branches     2834     2835       +1     
==========================================
+ Hits        16100    16136      +36     
  Misses       1638     1638              
  Partials      706      706              

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.