50 descendants and 50 ancestors of f7b3d64db8e0a0c4
|
2026-08-16
| ||
| 16:00 |
frames streams through the plaintext memo (#84)
frames was the one bulk reader still calling the uncached content(), so a delta base was reconstructed once per referencing delta rather than once per run — superlinear work against a linear stream, on corpora that are ~90% deltas after a fossil import. libfossil (17,449 artifacts): 12.4s -> 2.34s real, 12.0s -> 2.25s user. Output byte-identical, peak RSS down 362MB -> 314MB. memoize_undeltaed stays off: the sweep reads each artifact once, and bases are memoized regardless of their own frame kind, which is the win. closed check-in: 2e3bb0b9db user: dmestas tags: frames-memo | |
| 15:04 | merge record-cards: technote C/T cards and forum N card (#77) check-in: 5bea261864 user: dmestas tags: trunk | |
| 14:46 |
technote and forum verbs write the cards the format allows (#77)
technote add gains --comment (the C card a timeline renders as the note's title, unattachable by any later verb) and repeatable --tag, minting T +sym-NAME * sorted ascending as fossil writes it. forum post and reply gain --mimetype, matching wiki commit's spelling. Empty titles, repeated tags, and whitespace tag names are refused. First CLI-level coverage for either verb: #76 shipped them tested only at the artifact layer. closed check-in: 21f6efdd1c user: dmestas tags: record-cards | |
| 13:52 | merge tag-cards: raw-namespace tags, values, and cancel (#80) check-in: 6b4ac926e6 user: dmestas tags: trunk | |
| 13:26 |
tag: raw-namespace tags, values, and cancel (#80)
tag add could mint only T +sym-NAME, so a verdict like ci-pass became a symbolic check-in name and the second check-in carrying it made that name ambiguous. add and cancel now take fossil's --raw namespace choice, an optional VALUE, and --propagate; list --raw reads raw tags back, since they are deliberately outside the name space. Names under wiki-, tkt-, or event- are refused, as fossil refuses them. The composer's arguments moved into a TagCardSpec, which hides its two precondition asserts from the release-assert scanner (a struct field is not slice-shaped), so its baseline entry goes; the asserts stay. closed check-in: 3fa532af3a user: dmestas tags: tag-cards | |
| 12:49 |
commit --sign: clearsigned manifests, fossil's contract (#72)
The signed text is the stored artifact, so readers strip the wrapper to reach the cards and the signature travels wherever the check-in syncs. Signer invocation matches fossil's clearsign.c: COMMAND OUTFILE INFILE, default 'gpg --clearsign -o ', ssh* basenames take the detached path with assembled armor, 'off' disables. A signer that fails or alters the manifest refuses the commit rather than storing it unsigned. artifact reports signature presence; carry is pinned through pack, verify, and the sync wire. check-in: eb2f1adeda user: dmestas tags: trunk | |
| 05:35 | wiki and ticket leave the top-level command list: the four authoring verbs are machine verbs by the #76 posture (another process is their audience), each still fully documented by its own help check-in: d18a3b5589 user: dmestas tags: trunk | |
| 05:17 | merge worktree-registry: the store-side worktree list and the worktrees verb (#70); verb-set union with forum/technote (#76) check-in: d88bdc5e67 user: dmestas tags: trunk | |
| 05:01 | worktree registry (#70): every open and update stamps the tree's absolute path into the store's .zit-cache/worktrees (advisory tier, best-effort, self-healing — the store-cone discipline); 'zit worktrees' lists the registered trees with branch and check-in, verifies membership by the store file's inode, lazily prunes dead or re-pointed entries, and answers bare via -R closed check-in: 9d079f3009 user: dmestas tags: worktree-registry | |
| 05:01 |
forum and technote authoring verbs (#76)
zit forum post/reply and zit technote add record forum-thread and event artifacts under the #71 machine-verb contract: bare store via -R, hash-only stdout, exact exit codes, unlisted in the human help. A reply resolves its thread root transitively; a technote edit reuses its event id. check-in: 4317c47ed5 user: dmestas tags: trunk | |
| 04:27 |
D cards carry milliseconds (#75)
Every artifact-minting path writes the .SSS writer form fossil itself emits, so same-second writes keep a real-time order for latest-wins consumers. SOURCE_DATE_EPOCH pins .000; display surfaces unchanged. check-in: 0ff8e58ce8 user: dmestas tags: trunk | |
| 04:05 | e2e: the import-fossil residue scans iterate a dedicated .iterate handle scoped to the scan — Linux dirRead re-seeks the fd per batch, so iterating a handle owned elsewhere is BADF there (macOS tolerates it) check-in: 68506d0e3e user: dmestas tags: trunk | |
| 02:37 |
sync: per-round credentials over the cookie channel (#74)
An authenticated exchange now carries a fresh credential on every round: body login until the peer's gate is known, then an x-f-l-c cookie whose nonce covers that round's own body. The transport attaches it; serve now verifies it (and refuses tampered, replayed, and duplicate logins) instead of ignoring the header its own announced version solicits. check-in: 1fd84b4327 user: dmestas tags: trunk | |
| 02:21 | zig fmt check-in: f7b57a5869 user: dmestas tags: trunk | |
| 01:51 | merge import-fossil: fossil history import with the compatibility consequence chosen at import time (#61) — --compatible preserves the corpus byte-exact in the origin's sync world; --submanifest re-mints history as sub-manifest trees under a new project code, gated by whole-history replay, proven file-set-equivalent pair by pair, published atomically, with the old->new provenance map shipped as a technote. Merge resolution unions the verb set with wiki/ticket (#71) and completes their top-help listing check-in: 3f1f2a257d user: dmestas tags: trunk | |
|
2026-08-15
| ||
| 23:20 | conversion under the staged-publication discipline plus the provenance technote (#61): the bundle's atomic rename is the only commit point (a failed run leaves a sweepable .zit-tmp-*, never a half-repo), and the old->new map ships as a deterministic event artifact inside the converted repository — sorted pairs, source-tip timestamp, project-code-derived id — so external hash citations resolve by lookup closed check-in: cbbc910841 user: dmestas tags: import-fossil | |
| 23:20 |
wiki and ticket authoring verbs (#71)
zit wiki commit and zit ticket new/set record wiki and ticket-change artifacts on a bare store via -R, under the store write lock. Machine verbs: hash-only stdout, exact exit codes, unlisted in the human help but fully documented via their own --help. check-in: fe8d14d754 user: dmestas tags: trunk | |
| 22:49 | import-fossil --submanifest converts (#61): topo-ordered re-mint of every check-in as an S-tree (text-level card rewrite — F run out, S run in sorted position, B dropped, P/Q/T chased through the old->new map, PGP stripped and counted, Z resealed), control artifacts remapped, blobs and non-referencing chains verbatim, clusters dropped; every (old,new) pair proven file-set-equivalent through the destination flattener before anything survives, counts reconciled against the census, failed runs roll back to nothing. --compatible byte-exactness is a pinned invariant: the frames census of an imported corpus equals the origin's stream, and the import still pulls from it (#73's prerequisite) check-in: 5d61c0da47 user: dmestas tags: import-fossil | |
| 20:52 | fossil_convert: the corpus census and the re-mint cascade rule (#61) — classify every artifact, count the remap surface (check-ins re-mint; wiki/forum/ticket/attachment chains carry verbatim under kept content names), state the casualties (PGP signatures strip, clusters drop), and refuse an incomplete corpus; --submanifest prints the census as the conversion's reconciliation baseline check-in: ce57c94b6d user: dmestas tags: import-fossil | |
| 18:46 | import-fossil verb (#61): the compatibility consequence is chosen at import time — --compatible preserves the corpus verbatim through the clone machinery and stays in the origin's sync world; --submanifest runs the whole-history replay gate (static k sweep at the tip, winner±1 replayed, refuse-if-worse without --force) ahead of the conversion pass, which lands next; fossil support lives in src/fossil/ beside src/git/ check-in: 4c24d037b7 user: dmestas tags: import-fossil | |
| 18:08 | add src/ignore.zig, dropped from the #64 commit check-in: 2d981d8f85 user: dmestas tags: trunk | |
| 18:07 |
ignore mechanism: ignore-glob setting, .zit-settings/ignore-glob, .zitignore (#64)
extras, addremove, clean, and add's directory expansion share one exclusion point; explicitly named files still add. Fossil glob dialect for the setting and versioned file, gitignore dialect for a root .zitignore, unioned when both exist. check-in: 5c68eb4688 user: dmestas tags: trunk | |
| 17:39 | the submanifest -R pin follows the analyze-mode grant check-in: dab7ec41f3 user: dmestas tags: import-fossil | |
| 17:19 | submanifest analyze: -R over a bare store (tip file set, no checkout) and --replay-all for whole-history replay (#61 step 2) — the conversion gate's evidence pass and before/after table in one read-only command; settings actions refuse -R in-verb check-in: 70b7e2f06e user: dmestas tags: import-fossil | |
| 16:37 | extract the staged-publication pipeline into src/publication.zig (#61 step 1): PublicationBundle's one-way stage transaction, PublicationHooks, generation names, and the atomic-rename extern move verbatim; git_import_cli re-exports the same names, so import-git behavior, tests, and callers are unchanged — history converters share one publication discipline check-in: 7eff935454 user: dmestas tags: import-fossil | |
| 16:24 | gate: check zig fmt check-in: ad900494fb user: dmestas tags: trunk | |
| 16:16 | zig fmt check-in: 54cdfbbfaa user: dmestas tags: trunk | |
| 16:08 | mirror-export: prune closed-branch refs on every export check-in: b0fde46c32 user: dmestas tags: trunk | |
| 15:58 | merge worktrees: detached checkouts, store lock, open --store re-point (#59) check-in: bcbf29d017 user: dmestas tags: trunk | |
| 15:30 | merge diff-repo-override: bare-store diff --from/--to via -R (#69) check-in: 8066a7c21d user: dmestas tags: trunk | |
| 15:16 | diff --from/--to honors -R (#69): the check-in pair touches no working tree, so a bare store serves it; the worktree comparison refuses -R in-verb with the works-in-a-checkout diagnostic closed check-in: 4002513eeb user: dmestas tags: diff-repo-override | |
| 15:06 |
open --store re-points an already-open tree (#59)
A moved store strands its satellites, and a stranded tree can hold uncommitted work. Re-open repairs the pointer in place, gated on the new store containing the tree's current check-in. closed check-in: 7ac1ec2255 user: dmestas tags: worktrees | |
| 13:39 |
worktrees: detached checkouts of one store (#59)
Every store writer serializes on an exclusive flock held open-to-commit (released before autosync and the drift advisory). A checkout's store_location section is resolved in one place; open --store PATH creates a detached working tree, and probeRoot accepts a checkout whose pointer resolves. Store-scoped verbs (pack, shun, serve) follow the pointer end to end. check-in: 94d4659a53 user: dmestas tags: worktrees | |
| 13:04 | merge help-topics: help topics, single-source option fragments, exit-status blocks everywhere, mechanical usage-line audit (#68) check-in: 9a5ae9774b user: dmestas tags: trunk | |
| 12:50 | help topics (#68): 'zit help names/sparse/repo/exit-codes'; verbs point at the names topic instead of restating the resolution litany (timestamp forms were documented in one help of six); unknown help topics refuse with exit 2; the -R/--no-cache prose and the conventional exit-status block become single spliced consts (28 helps gain the block); a unit test asserts every options-table flag reaches its usage line — it caught commit --no-sync, merge --no-sync, and the cone flags on open/update/clone closed check-in: 21fa16cc4b user: dmestas tags: help-topics | |
| 12:10 |
import-git: O(changes) per-commit cost and live progress (#65)
Tree edits adopt untouched sibling leaves, the tree witness digest is structural over the root S-run (flat digests unchanged), and log reads go through a primed plaintext cache. Ingest reports strided progress with pack/verify phase markers. check-in: db4b2573e7 user: dmestas tags: trunk | |
| 04:01 | merge read-verbs: frames incremental artifact stream and annotate VERSION/--full (#66, #67) check-in: b41272c88c user: dmestas tags: trunk | |
| 02:52 | read verbs for external consumers (#66, #67): frames streams whole raw artifacts since an opaque cursor in persisted order, with LogKey prefix validation (appends keep a cursor alive, a pack/shun rewrite fails it with exit 4; exit 3 = current) and per-artifact hash verification; annotate gains a VERSION operand resolved like cat's (bare-repo -R works) and --full for full artifact IDs closed check-in: 68eba6cd49 user: dmestas tags: read-verbs | |
| 01:03 | merge sparse-cone: sparse checkouts and sparse stores per draft-fossil-sparse-cone-00 — worktree cones with exclusions, cone-scoped clone/open/update, cone/cone-not wire pragmas with graceful degradation, demand fetch, pack --drop-elided, cross-cone record merges, and the sparse-origin clone refusal check-in: 2ad75a9bea user: dmestas tags: trunk | |
| 00:39 | sparse origin refuses clones it does not cover (spec 5.3): dense and uncovered-cone clone requests answer one error card and no content; the client treats a progress-free error round-0 reply as fatal (spec amended — the sync 11.1 discard tolerance is for cosmetic cards on progressing replies) and quotes the peer's refusal closed check-in: 6e064b7301 user: dmestas tags: sparse-cone | |
|
2026-08-14
| ||
| 23:23 | cone exclusions (spec 3.2): --cone-not/'!' cone-file lines subtract a subtree, longest-match decides coverage (spec reworded from ordered last-match — order-independent, keeps canonical serialization and digests), dead rules and contradictions refused; serialized cone v2 (0x00 lead-in + flag byte, v1 bytes unchanged without exclusions); cone-not wire pragma with over-send degradation check-in: 60274756a0 user: dmestas tags: sparse-cone | |
| 21:37 | clone/open/update: --cone-file=FILE reads cone members from a file (one per line, # comments); clone help gains the missing --cone entry check-in: 9191baea4b user: dmestas tags: sparse-cone | |
| 21:16 | cross-cone merge applies to the record (spec 4.4): out-of-cone paths resolve at hash level with no blob reads, theirs-only changes ride a merge_overlay container section (v4) in the pending-merge splice, commit composes and clears it one-shot, and only both-changed conflicts refuse — naming the path and the widen remedy check-in: de5e6d9dc0 user: dmestas tags: sparse-cone | |
| 20:17 | merge trunk check-in: 8b3d577ec1 user: dmestas tags: sparse-cone | |
| 20:02 | zit cone verb: show/add/remove drive the update --cone path, so add fetches and remove dematerializes; member cap to 255 (the count byte's range, spec 3.2 bounds sentence added); spec gains exclusions (cone-not pragma, last-match-wins) and the patterns-deliberately-absent rationale check-in: 2cd32f8a38 user: dmestas tags: sparse-cone | |
| 19:07 | cone lifecycle completes: update --cone/--dense demand-fetch through the saved remote with one bounded retry; store policy follows the worktree cone only on already-sparse stores (a dense store's --cone can never arm reclaim); pack --drop-elided reclaims the elide set composed with the shun filter (spec 5.4, the only deletion) check-in: f79b3e469a user: dmestas tags: sparse-cone | |
| 18:42 | pull scoping on the wire: pull/sync clients carry the store cone as pragmas, and the server filters its igot advertisement so out-of-cone holdings never mint requester phantoms — the transfer narrows at its source check-in: 9abd665d70 user: dmestas tags: sparse-cone | |
| 18:27 | open on a sparse store demand-fetches (spec 6.4 policy form): the requested materialization becomes the store cone, an ordinary pull aligns to it, and one bounded re-attempt resolves fresh; without a remote the refusal names the class and remedies check-in: 658ae6bf26 user: dmestas tags: sparse-cone | |
| 18:03 | add sparse.zig: the shared cone classification the previous check-in imports check-in: f031a5283c user: dmestas tags: sparse-cone | |
| 18:02 | cone pragma on the wire (spec 6.1/6.2): client requests scoping via fossil's silently-ignored pragma contract; the server filters clone batches from a digest-keyed classification memo dropped with its snapshot; a cone-aware pair transfers spine + in-cone content only check-in: a52eb95fc0 user: dmestas tags: sparse-cone | |
| 17:34 | sparse-cone spec: the cone request rides pragma cards — fossil answers unknown bare cards with a protocol error (xfer.c) but specifies unknown pragmas as silently ignored, so degradation becomes a documented property rather than assumed tolerance check-in: f7b3d64db8 user: dmestas tags: sparse-cone | |
| 17:33 | merge trunk: autosync + commit fork guard; update grows --no-sync beside --cone/--dense check-in: c50dcc978a user: dmestas tags: sparse-cone | |
| 17:18 | sparse store round trip: verify summarizes elision distinctly from damage (streamed off the existing plaintext pass); commit omits R only when an elided carried blob makes the aggregate uncomputable, and the minted check-in converges to dense peers check-in: 6eecca3136 user: dmestas tags: sparse-cone | |
| 16:57 | commit: fork guard — parent must still be its branch's leaf (--allow-fork to fork); autosync — settings verb, pull before commit/update/merge, push after commit, warn-and-continue (#57) check-in: 7a4dff5d5f user: dmestas tags: trunk | |
| 16:34 | sync on a sparse store: elided phantoms are never gimme'd (Eligibility.elided, distinct from shun) and volunteered content filters at every receive persist via post-run reclassification, so new in-cone commits arrive while declined content stays declined check-in: 4fb1a29e3e user: dmestas tags: sparse-cone | |
| 16:10 | clone --cone: local filtering at persist (spec 6.3) — structural artifacts and in-cone blobs keep, exclusively-out-of-cone blobs elide, count reported; open's preflight classifies absence three ways (foreign / elided / transit) instead of reading a sparse store as foreign check-in: 5a74b96333 user: dmestas tags: sparse-cone | |
| 15:55 | store cone: .zit-cache/store-cone member with loss-heals-to-dense semantics; derived caches proven cone-independent (spec 7 corrected: no cone term in cache keys) check-in: 7c734e7763 user: dmestas tags: sparse-cone | |
| 15:40 | merge trunk: import-git drops std.c (Ambient environ, Io dir fsync); export-git DEST preflight names its own git dir check-in: 7c1fb6e58d user: dmestas tags: sparse-cone | |
| 15:37 | linux: iterate/fsync need real fds — reopen O_PATH dir handles with iterate rights (7 sites + test helpers); runScripted captures into caller-owned storage, ending a use-after-return the Linux stack exposed check-in: a60df1f752 user: dmestas tags: trunk | |
| 15:17 | sparse cone: tree-dialect keystone on an S-card repo, narrowing dematerializes cleanly; README RFC row; open help teaches --cone check-in: 18ad2eea27 user: dmestas tags: sparse-cone | |
| 14:42 | sparse commit synthesis: snapshot inside the cone, parent cards carried verbatim outside; R stays log-sourced; keystone e2e pins reader-indistinguishability check-in: e7c880ac83 user: dmestas tags: sparse-cone | |
| 14:37 | import-git: drop std.c — child environ threads from Init via Ambient, dir fsync goes through Io; Linux builds no longer demand libc check-in: 951b654c5b user: dmestas tags: trunk | |
| 14:33 | sparse worktree verbs: open/update --cone and --dense; add/mv refuse outside the cone; extras/clean/addremove/status cone-scoped; R verified over a dense probe (spec 4.3 verification clause added) check-in: c59a64678f user: dmestas tags: sparse-cone | |
| 14:12 | sparse cone at the seam: MergeWalk filter, cone-aware requireClean, SparseOpt on applyCheckin (record+cone one splice), cross-cone merge refuses fail-closed check-in: 259219f90e user: dmestas tags: sparse-cone | |
| 13:50 | container v3: cone + store_location tags; v2 folds forward, version negative space pinned check-in: 49815f9ac6 user: dmestas tags: sparse-cone | |
| 13:43 | export-git: DEST preflight must name DEST's own git dir — upward discovery adopted the enclosing repo and aimed fast-import at its refs check-in: f451567485 user: dmestas tags: trunk | |
| 13:09 | sparse-cone: RFC draft + cone module (spec 3): sorted-minimal members, covers/coversRange, canonical serialize with fail-closed parse, SHA3 digest check-in: 0fade1c499 user: dmestas tags: sparse-cone | |
| 12:21 | merge sub-manifest: delta-manifest dialect, R cards, diff hunks, server identity, stash, open/log resolution, clone remote+progress, timestamp names; e2e expectations updated check-in: 8a45a021f9 user: dmestas tags: trunk | |
| 10:11 | clone: save the URL as the remote (canonical, password-free, --once opts out); live per-round progress on a TTY stderr (#55 #56) closed check-in: 69c14cd747 user: dmestas tags: sub-manifest | |
| 10:11 | resolve: §8.3 portable T/Z timestamp form and the NAME:TIMESTAMP compound (#60) check-in: f2691f0a23 user: dmestas tags: sub-manifest | |
| 01:57 | open: optional CHECKIN — resolve like info and materialize it directly; no-arg tip behavior unchanged (#54) check-in: 8573f4f57d user: dmestas tags: sub-manifest | |
| 01:33 | log: CHECKIN resolves like info — branch, tag, prefix, reserved forms; full hex IDs keep the stateless fast path check-in: b77d2898ad user: dmestas tags: sub-manifest | |
| 00:11 | stash: save/list/show/apply/pop/drop over a durable .zit-stash sidecar — fail-closed apply (clean targets, same check-in), corruption refuses loudly; the no-stash reflex retires; README rows updated check-in: ba3d1e705b user: dmestas tags: sub-manifest | |
|
2026-08-13
| ||
| 22:12 | commit: keep the delta-manifest dialect — once a repo stores a B-card manifest (durable seen_delta config, ingest-detected), commits emit deltas against the parent's baseline with delete markers, byte-cost guarded; commit also resolves a delta parent's true effective set check-in: 60b3557200 user: dmestas tags: sub-manifest | |
| 20:14 | diff: textual hunks by default — patience alignment, git-compatible unified output (patch -p1 round-trips), -c N context, --brief keeps name-status check-in: ca0f3df379 user: dmestas tags: sub-manifest | |
| 20:13 | sync §9.1: real per-instance servercode — derived from the checkout's absolute path (deriveServerCode, domain-separated from the project code), emitted by serve and all four client verbs; placeholder retired to test fixtures check-in: 59e8660999 user: dmestas tags: sub-manifest | |
| 19:22 | README: five-RFC table, capability overview, and the compatibility framing — what Fossil has, what zit has, and the one break (sub-manifests) check-in: 0f8b901650 user: dmestas tags: sub-manifest | |
| 19:22 | commit: emit the R aggregate checksum on every check-in; update/open verify it fail-closed (Fossil's default repo-cksum posture); diff: rev-to-rev name-status via --from/--to check-in: ad8012ec78 user: dmestas tags: sub-manifest | |
| 18:43 | docs: carry the three errata the code implements ahead of the RFC text (repo-state §5.4 cancel-halt, artifact-format §4.1.2 sterilizer, sync §3.3 content-received OPTION B) — provenance for the divergence until the canon applies them check-in: 062e9a5512 user: dmestas tags: sub-manifest | |
| 18:31 | repostate: cancel-tag propagation halts at a rowless node, matching fossil tag.c (#16); erratum filed for the divergent §5.4 draft text check-in: e286beebd3 user: dmestas tags: sub-manifest | |
| 16:09 | export-git: mirror canonical history into git check-in: ae7c26a1a3 user: dmestas tags: sub-manifest | |
| 16:09 | bench: qualify incremental git export scaling — memo parse bound pinned; althttpd/wapp export clean (fsck 0), phantom-parent lineage refuses naming both hashes check-in: 75fb61025d user: dmestas tags: sub-manifest | |
| 15:18 | e2e: verify git export commit by commit check-in: 08ca21c63c user: dmestas tags: sub-manifest | |
| 15:14 | cli: export git mirrors incrementally check-in: 5b2c6139d5 user: dmestas tags: sub-manifest | |
| 15:03 | export-git: publish mirrors without partial state check-in: 41dfef63b2 user: dmestas tags: sub-manifest | |
| 14:32 | export-git: translate check-ins refs and failures precisely check-in: 74c5c8aba0 user: dmestas tags: sub-manifest | |
| 13:52 | export-git: plan topologically and enumerate through the seam check-in: 491cbc50c5 user: dmestas tags: sub-manifest | |
| 13:35 | export-git: persist a disposable local mark map check-in: 17c02d8950 user: dmestas tags: sub-manifest | |
| 13:23 | export-git: serialize bounded fast-import streams (the writer files the previous commit's message described) check-in: bb1cc4a01b user: dmestas tags: sub-manifest | |
| 13:23 | export-git: serialize bounded fast-import streams check-in: 2450e4f142 user: dmestas tags: sub-manifest | |
| 13:08 | git: gather every interop root into src/git/ — cli.zig joins the zit module (nine importers rerouted), moved roots take parent deps through the module; ~986 duplicate test executions eliminated check-in: 49c08bb191 user: dmestas tags: sub-manifest | |
| 11:41 | submanifest: replay history to pick k (#41) — analyze replays last 200 commits through the real builder+delta encoder (both cost axes, bootstrap split, baseline-charged materialization); enable pins the replayed argmin, snapshot objective as fallback. Advisory only. check-in: b9ceda433f user: dmestas tags: sub-manifest | |
| 03:06 | clone: stop inflating served/received delta containers for their header — prefix-inflate 16B, base frame read header-only; server CPU per 1M-artifact clone 300s+ -> 22s, clone verifies 0 faults check-in: f5eafc6a2f user: dmestas tags: sub-manifest | |
| 01:25 | serve: notice a --force shun/pack generation swap — stat the path beside the held handle, reopen between rounds; shunned bytes stop serving with no restart. Closes #17 check-in: faa8befc3b user: dmestas tags: sub-manifest | |
| 00:16 | serve: scale the write deadline with the reply — 4MB clone batches at the flat 5s cut healthy slow peers mid-write (#22 shape); slow-drain e2e proves it check-in: fa230db9e8 user: dmestas tags: sub-manifest | |
|
2026-08-12
| ||
| 23:21 | timeline: order tied runs by generation, not per-pair ancestry walks — 305k-commit CVS import renders in 5s vs CPU-hours; B1 descendant-first preserved exactly check-in: d823a2811d user: dmestas tags: sub-manifest | |
| 21:47 | clone: byte-budgeted batches (4MB) — 159 rounds vs ~1000 for a 1M-artifact store; warm clones plant the checkpoint (first command 0.06s); print transfer/derive phases. Closes #53 check-in: c32414008b user: dmestas tags: sub-manifest | |
| 19:56 | checkpoint: accept delta manifests; anchor at newest check-in (mainline fold collapsed on real repos); sync: index the pending-delta queue (O(1) hasTarget, O(deps) completeFor) — zit now clones 2GB pkgsrc-2012 from real fossil in 10min, resumes opens in 9s vs 4min derive check-in: c70f48f825 user: dmestas tags: sub-manifest | |
| 09:04 |
checkpoint: retire the tree shape — sidecar only, structurally
The tree shape minted permanent SUBMANIFEST artifacts for benefits already had elsewhere: the rebuild win lives in the prefix record, verify's §6.2 oracle runs buildTree in memory, and the subtree-sharing bytes re-materialize at the first real S-emitting commit at identical cost. Minting early only created unreferenced permanent artifacts and the allow-tree/demotion/gate apparatus needed to keep automation from doing it by accident — all deleted with it. build no longer opens the log for writing or reads the emission settings at all; sidecar layout v3 (old sidecars miss and rebuild in seconds). Proven at scale: SHA-256 of pkgsrc's 593MB store is byte-identical across a full build; cold 4:12, rebuild 3.4s. check-in: 2a8fdcff60 user: dmestas tags: sub-manifest | |
|
2026-08-11
| ||
| 22:45 |
checkpoint: one-pass build, incremental advance, opportunistic refresh
Build paid a bypass derive, a whole-prefix delta scan, and a capture re-derive; a view-free captureRecords (byte-identical to the deriving walk, pinned at every boundary) now does capture and delta detection in one pass, and an existing sidecar's body plus its delta-free attestation reduce a rebuild to the unscanned tail. pkgsrc: cold build 17:58 -> 4:21; rebuild or anchor advance -> 5.1s. Repository.open's healing paths keep an existing checkpoint fresh: witness shape forced (the automatic path can never mint an artifact, whatever §8.1 says — a tree pin advances by demotion and is re-pinned deliberately), gated on 4MB growth AND anchor progress so blob-only growth never churns the sidecar. No checkpoint, no auto-creation: building once is the opt-in. check-in: adae602908 user: dmestas tags: sub-manifest | |
| 20:04 |
repostate: carry the C-card comment in the crosslink record
An unlimited timeline re-read and re-parsed every manifest just to print its comment first-line — 163k inflate+parse cycles on pkgsrc. The comment now rides the crosslink record into the derived state (commentOf; image backing falls back to the manifest read), so the render loop is a lookup per row: full timeline 12.3s -> 1.06s. Record-format consumers bumped so old bodies miss instead of misparsing (repocache v3, checkpoint v2 — one re-derive per repo, then warm). zit log's pinned B-log-1 walk untouched. check-in: 57d8de4120 user: dmestas tags: sub-manifest | |