[botandbang]

If making a store was cut off after objects/ and snapshots/ were made but before store.json was written, every later Store::open_or_init refused the folder as not a locallibrary store. The check it runs under the lock accepted only the files locallibrary keeps beside a store, not what a make leaves behind. store copy had the same problem with a destination cut off at the same point.

Added

• Store::left_by_making, which accepts an entry left by a cut-off make: objects/ or snapshots/ while they're still empty real directories, or a temp file.
• A unit test for the reported case, and a Hegel property that draws which leftovers a cut-off make left plus an optional foreign file (in the root, or inside objects/ or snapshots/). It checks the store is made exactly when nothing foreign is there, and that opening it again gives the same store.
• Cases in the offsite copy property for a cut-off copy destination (finished with the source's id) and for one with a file inside objects/ (still refused).
• A changelog fragment.

Changed

• Under the lock, open_or_init accepts those leftovers and finishes the store. A folder holding anything locallibrary didn't write, including a file inside objects/ or snapshots/, is still refused. Before the lock it accepts what it did before.
• offsite::copy carries on into a destination holding only such leftovers instead of refusing it.
• The doc comments on open_or_init and holds_only_ours say what each accepts now.

Removed

• The temp-file check inlined in holds_only_ours; left_by_making covers it.

[github-actions]

Rust coverage

From run 337. Both are held to the thresholds in .cargo/config.toml.

Tests	Lines	Functions	Regions
All	97.45% (87717/90009)	96.52% (14220/14732)	95.7% (193556/202262)
Hegel property tests alone	95.3% (85777/90009)	94.03% (13852/14732)	93.36% (188830/202262)