8 Commits

Include renames

Author	SHA1	Message	Date
archvillainette	978e2913f6	fix(nwsync): pin emitter version, ship the shim, harden manifest reads ... Review follow-ups on the same work: - emitter_version is its own sidecar field, not the build revision. Keying the merge check on created_with invalidated every published index on every unrelated crucible commit, forcing a re-emit of the whole corpus. - flake.nix builds cmd/crucible-nwsync, so Nix consumers get the binary the docs promise. - readManifest uses io.ReadFull: bytes.Reader.Read can return a short count with no error, so a truncated manifest parsed into zero-padded garbage. - emit refuses a .mod outright and names why, rather than failing on an unknown extension. - SOURCE_DATE_EPOCH pins the sidecar timestamp; the manifest was already deterministic. - Entry.identity replaces the resref/restype key struct duplicated in emit and assemble. Refs #53.	2026-07-29 12:17:08 +02:00
archvillainette	0de1c1e280	feat(nwsync): emit blobs and per-artifact NSYM, assemble merged manifests ... Adds `crucible nwsync emit` and `crucible nwsync assemble`, the split replacement for upstream nwn_nwsync_write, which cannot be used because it wants every hak, the TLK and the module present in one run on one disk. emit explodes one artifact — a .hak/.erf or a loose file such as the TLK — into NWSync blobs (NWCompressedBuffer framing, magic NSYC, zstd) plus a NSYM v3 manifest describing only that artifact, with the same .json sidecar upstream writes. Blob names are the sha1 of the uncompressed bytes, restypes nss/ndb/gic are always skipped, an unresolvable restype is a hard error, a resource over 15 MB fails closed, and no latest or .origin file is ever written. assemble merges the per-artifact manifests into one, reading no bulk data. The merge rule is resref shadowing, not concatenation: a resref present in more than one artifact resolves to the earliest artifact in --order, the way the game resolves it. It refuses to merge across mismatched emitter versions, and takes --group-id from the caller (1 current, 2 testing; 0 is absent). Refs #53.	2026-07-29 12:12:22 +02:00
archvillainette	56c67132f3	claude: fold in old sow-tools	2026-06-13 09:49:29 +02:00
archvillainette	5a101e94c3	gitea.westgate -> git.westgate	2026-06-11 22:47:22 +02:00
archvillainette	97f8427c4b	Phase 5: scaffold Crucible (sow-tools) — dispatcher + builder shims, container, PR-first CI, fail-closed (D11, D19).	2026-06-11 20:36:13 +02:00
archvillainette	000d31ad24	YAML Configuration Refactor	2026-05-07 13:44:46 +02:00
archvillainette	dc93feb176	Topdata Migration (#2) ... Native topdata migration summary - Replaced the legacy 2dabuilder runtime path with the native topdata builder. - Native build, validate, compare, and convert flows now cover the canonical topdata pipeline. - compare-topdata is now a native self-check; legacy reference-builder runtime usage was removed. - Parts generation follows the native asset-scan contract and uses sow-assets via project asset resolution. - Generated feat families, class-feat injects, masterfeat/successor expansion, and dataset-derived feat generation were brought to parity and regression-covered. - Remaining mirrored datasets were migrated into native-owned canonical data while preserving lock IDs. - normalize-topdata bridge behavior and migration-era ownership markers were removed. - Documentation was rewritten around the native workflow and active contracts were cleaned up. - Final hardening pass removed stale validator assumptions and cleaned ignored/generated artifacts for PR readiness. Reviewed-on: https://gitea.westgate.pw/ShadowsOverWestgate/sow-tools/pulls/2 Co-authored-by: vickydotbat <vickydotbat@tutamail.com> Co-committed-by: vickydotbat <vickydotbat@tutamail.com>	2026-04-09 07:14:01 +00:00
archvillainette	b7f188779a	Initial sow-tools import	2026-04-02 17:19:23 +02:00