zit 0.14.3

zit 0.14.3 makes import-git spend far less CPU on large histories, fixes its progress total, and makes submanifest show cheap on large stores.

This release is from trunk. Stores, caches and the wire are unchanged from 0.14.2; nothing is rebuilt on upgrade.

import-git

• Deltas against one source first. Each artifact the import stores was delta-encoded against its likely source and up to four neighbouring artifacts, and the smallest delta was kept. Neighbours that share little with the artifact make the delta search scan it byte by byte, and on a large history that was most of the import's CPU. Each artifact is now encoded against its likely source alone, as Fossil's import does, and the neighbours are tried only when that delta is not small enough. On a 37,000-commit history, user CPU fell by 42% and the store grew by 0.13%. Wall time gains less, because other stages now set the pace. pack keeps the full search.
• The progress total counts what is imported. The total counted the commits reachable from the branches and tags but not the pull-request heads the import also takes, so on a repository with many merged pull requests progress ran past 100%. It now counts the same refs the import exports. The later verifying N/M commits lines are a separate stage with their own count.

submanifest show

zit submanifest show read the whole store to answer one setting. It now reads the store's saved log index, as info does, and builds the index once when there is none, so later calls answer at once.

Qualification

The release manifest records native qualification against the exact tagged source archive on macOS ARM64, Linux x86-64 and Linux ARM64, using the pinned Zig 0.17.0 toolchain. Publication is gated on unit and end-to-end tests, Git differentials, deterministic simulation, seeded fuzzing, and Fossil interoperability checks on every distributed target.

For the basis of the conformance claim, see the 0.1.0 release notes.