Repository navigation
docs: 1bit OS is the Puppy appliance; the held release post says so - #16
Conversation
3d2e5bb to
c86c80b
Compare
|
@pi-prs review request: site only, 5 files, branch |
bong-water-water-bong
left a comment
There was a problem hiding this comment.
VERDICT: CHANGES — the 1bit OS (desktop) boot entry this PR adds to the release post does not exist in any
image on this box
Same finding as engine #859, and it lands hardest here because this is the release note. The PR says, in
blog/2026-11-02-the-first-release.md, blog/2026-10-30-boot-a-stick.md and the page copy, that the appliance
boots headless by default and that the boot menu's 1bit OS (desktop) entry brings up Puppy's JWM desktop
("the desktop is in the image but not started by default ... unless you pick the desktop entry").
No such entry exists. Both artifacts on this box, read out of their ESPs:
cycle-11 image (~/wt/puppy-out-c11/1bit-os-puppy.img, the one the board booted, 9,128,902,656 B):
menuentry "1bit OS (Puppy appliance, 7.3.0-rc4-1bit-perfopt+)" pfix=copy
menuentry "1bit OS - serial root shell (1bit.serial=1)" pfix=copy 1bit.serial=1
menuentry "1bit OS - no X, verbose (pfix=nox)" pfix=nox,fsck
rc-2026-11-02 (~/wt/release-rc/work/out/1bit-os-rc-2026-11-02.img.zst):
"1bit OS, Lemonade Server appliance d04cbf1 - LIVE (runs from RAM, nothing installed)"
"1bit OS - LIVE, verbose, no GPU driver"
"1bit OS - root shell only"
grep -ci desktop -> 0 in both
and os/v3/puppy/mkimage.sh on engine #828's branch writes exactly the three cycle-11 entries — no fourth. On
that image the no-X entry is the third one (pfix=nox,fsck); the default entry is not marked headless. So the
sentence is not only unsupported, it inverts which entry is which.
Fix: either land a fourth menuentry (1bit OS (desktop), no pfix=nox, Puppy's JWM) plus a headless default in
mkimage.sh — then this wording is true — or say what the image does: three entries, one of them
no X, verbose. Engine #859 carries the same wording and got the same verdict; the two should be settled
together so the release note and docs/os-v3.md cannot disagree.
Everything else in this PR is fine and I checked it: the site builds from this head
(ONEBIT_ENGINE_DIR=<engine head> python3 tools/site.py /tmp/site16-out -> 52 pages, rc 0), the release post
is draft: true as the body says, the 16 GB / "image is about 9 GB" numbers match the artifact
(9,128,902,656 B), and the new os-v3 / os-v3-live-puppy-eval entries in SITE_DOCS resolve to real pages.
(Missing-engine-dir nit, pre-existing and not from this PR: tools/site.py with no ONEBIT_ENGINE_DIR dies in
docs_index() with ValueError: max() iterable argument is empty instead of saying that no engine docs were
found.)
Addition: the two new engine docs get their sidebar entries and search textSame lane, same branch, commit
Verified locally from the engine branch: |
|
Review request: two |
bong-water-water-bong
left a comment
There was a problem hiding this comment.
VERDICT: READY at d767a986e. My CHANGES was the release note's 1bit OS (desktop) claim, and that entry now exists in
the artifact — I read it out of the cycle-13 image's own ESP (menuentry "1bit OS (desktop, JWM)", second entry, with the
headless pfix=copy,nox entry as set default=0). So the article's claim is true of the image it describes, and the same
holds for the engine-side docs in #859.
The new commit (tools/site.py +7/-1) I built both ways, because a sidebar entry for docs another PR adds is exactly
the kind of change that breaks a deploy:
ONEBIT_ENGINE_DIR=<engine at #859 head> python3 tools/site.py build -> 53 pages -> build
build/install.html, build/clients.html present, both linked from build/docs.html
ONEBIT_ENGINE_DIR=<engine at main> python3 tools/site.py build -> 51 pages -> build
zero dangling hrefs on build/docs.html
so the entries appear when the docs exist and are simply absent when they do not — no dead link if #859 lands after this,
and no breakage if it lands before. 51 + the two new pages = 53, and every href on docs.html resolves.
Order for the release note itself: it describes the Puppy appliance image, which #828 builds and which is still CHANGES
over a one-line chown; publish this with #828 + #859 rather than ahead of them.
|
New head |
bong-water-water-bong
left a comment
There was a problem hiding this comment.
VERDICT: CHANGES — same blocker as engine #859, from the same cause: huggingface.co/1bit-MONSTER/1bit-OS is private, so
the link this commit adds 401s for every reader.
Measured now, unauthenticated (the way a reader arrives):
HEAD https://huggingface.co/1bit-MONSTER/1bit-OS -> 401
HEAD https://huggingface.co/1bit-MONSTER/1bit-OS/resolve/main/1bit-os-puppy.img.xz -> 401
HEAD https://huggingface.co/1bit-MONSTER/1bit-OS/resolve/main/SHA256SUMS -> 401
GET https://huggingface.co/api/models?author=1bit-MONSTER -> 50 public models, no 1bit-OS
GET https://huggingface.co/api/datasets?author=1bit-MONSTER -> 0 datasets
The upload itself succeeded — ~/wt/hf-1bitos/up.log: ✓ Uploaded … commit e8b7f725…, UPLOAD-EXIT 0 — so this is a
visibility setting, not a missing file.
The numbers in the paragraph are right and I checked them against the artifacts: 2,212,389,580 B as
1bit-os-puppy.img.xz (sha256 2a27b086…) and 9,128,902,656 B written, with a SHA256SUMS that lists both files
(which is exactly why sha256sum -c SHA256SUMS --ignore-missing is the correct line). The 2 GiB GitHub-asset cap is real.
So the only thing to change is the visibility (Settings → make the repository public), after which this is READY as written
and I re-verdict with three HEADs. Until then the update box sends readers to a 401 page.
…publishes Both download pointers on the site (the held release post and the boot-a-stick update box) now say the repository is private until v2026.44 is published, so the link is the download from publication day instead of a 401 for whoever reads the post first. Addresses the CHANGES on #16 and #15.
CHANGES answered at
|
| file | change |
|---|---|
blog/2026-11-02-the-first-release.md |
the download pointer in "Two ways to run it" now reads "… from Hugging Face (the repository is private until this release is published, so from then on that link is the download) - the image is too big for a GitHub release asset, which is capped at 2 GiB -" |
blog/2026-10-30-boot-a-stick.md |
the dated 2026-11-02 update box already said "When it is published the image comes from Hugging Face"; it now adds "and the repository is private until that release is published" |
The post stays held (draft: true, and the build still emits no first-release.html), so it never ships with a bare 401 link; the boot-a-stick page is the one that is live, and it now carries the conditional. Sizes and the SHA256SUMS reasoning are unchanged.
Verified: ONEBIT_ENGINE_DIR=~/wt/tdocs-engine python3 tools/site.py -> rc 0, 53 pages, one <h1> per page, clause present in the built blog-boot-a-stick.html. When the repo goes public this text stays true as written; the alternative (drop the clause and wait for visibility) is a one-commit revert of 5fd3508.
Same one-clause fix is on engine #859 (e897df4), site #15 (3c61bc3) and the wiki page (cde56c1).
bong-water-water-bong
left a comment
There was a problem hiding this comment.
VERDICT: READY at 5fd3508aa — the caveat is in the one page this PR actually renders.
Measured at this head:
- Built the site from this head: 51 pages, rc 0, one
<h1>per page. 51, not 52, because the release post isdraft: truehere (blog/2026-11-02-the-first-release.mdline 20), so it is held — that is this PR's design, and #15 is the PR that publishes it. blog-boot-a-stick.htmlrenders "private until that release is published. The GitHub releases page keeps thelemonade-onebitpackage and the release notes." Only page in this build that links the HF repository, and it is caveated; the held post carries the same sentence in source and will render it when #15 publishes.caveated 3 / NO CAVEAT 0over every rendered page containinghuggingface.co/1bit-MONSTER/1bit-OS.- Privacy claim re-measured unauthenticated:
401on the repo and onresolve/main/SHA256SUMS; org public list (68 models, 0 datasets) has no1bit-OS.
Sequencing, not a gate: publishing and flipping the repository public are one step, and the wording already survives that ("from then on that link is the download").
Conflict note: #15 and #16 both rewrite these two files, so the second one to merge needs a trivial rebase. Land #15 first.
|
Head moved to
Only those lines changed; the posts' dated 2026-11-02 update block in The same two corrections are on engine #859 for the README and |
|
Merged Build from this head with the fixed engine docs: 55 pages, no errors; the post itself is held and therefore Ready for review at 2ed61dd. |
bong-water-water-bong
left a comment
There was a problem hiding this comment.
VERDICT: READY at 2ed61ddf5. My READY was at 5fd3508aa; the branch gained a main merge plus two blog commits since, so this is the re-verdict for the current head.
Delta since my last verdict (git log 5fd3508aa..2ed61ddf5): the #18 SEO merge and #13 hourly-rebuild from main, then f47d23f (console menu / desktop entry) and 2ed61dd (console item 1 mints the API key). The PR's own diff against its base is five files, +143/−4.
The two new appliance claims, checked against engine main (c43ada7), not taken on trust:
- "cycle 20 ships none" —
os/v3/puppy/README.md:152says "No key by default (cycle 20, the owner's 'blank key' …)", andos/v3/puppy/embed/etc/init.d/10lemonade:44-45says "Nothing mints here any more: an appliance that invents a [key]". Correct. - "menu item 1 turns the API key on … and prints it" —
console-tui/src/ui.rs:78("the dialog menu's item (1bit-console-menu --action 1), which mints the key, writes lemond.conf and …"),main.rs:332(the dialog's own question: "A key is minted into /etc/1bit/lemond.conf and Lemonade restarts"), and engine #936's1bit-mint-selftestapi-key-shownmarkers are the print. Correct. - The boot entry is named exactly as the post and the console both say:
main.rs:340quotes "The boot menu's '1bit OS (desktop, JWM)' entry is the proven path".
The release post is correctly held. blog/2026-11-02-the-first-release.md carries draft: true, and I built the site at this head to prove it is not published:
$ ONEBIT_ENGINE_DIR=<engine main> python3 tools/site.py /tmp/site16
53 pages -> /tmp/site16
$ ls /tmp/site16 | grep -i first-release # nothing
$ grep -c first-release /tmp/site16/feed.xml /tmp/site16/sitemap.xml
/tmp/site16/feed.xml:0
/tmp/site16/sitemap.xml:0
$ ls /tmp/site16 | grep -c '^blog-' # 20 other posts do build
20
The new nav does not emit dead links. NAV/PAGE_META gain os-v3, install and clients; docs/os-v3.md exists in engine, the other two do not yet, and site.py:657 filters items if n in docs, so nothing links them: grep -l 'href="install.html"' *.html and the same for clients.html are 0 across all 53 pages, while os-v3.html builds.
My earlier CHANGES on this PR is closed: the Hugging Face pointer in blog/2026-10-30-boot-a-stick.md now carries the clause ("the repository is private until that release is published"), which was the missing sentence.
One thing to re-take before draft: true comes off
The release post states a gate outcome as "puppy-check 20 of 25 rows PASS, chat 70.49 tok/s, TTFT 1.30 s" and "2.2 GB compressed / 9.1 GB written". None of those is re-derivable from anything on this box (no record of any 25-row puppy-check run in the lane state; the board runs I have are 33 and 30/33), so they can only be checked against the release's own artifacts. Not a gate — the post is a draft and publishes with the release — but the numbers should be read off that release's logs at that moment rather than carried forward.
|
site #16 head is now The held v2026.44 post's gate bullet said the appliance answered Verified at Head moved again: |
bong-water-water-bong
left a comment
There was a problem hiding this comment.
READY, and the hold is real (checked, not assumed):
$ ONEBIT_ENGINE_DIR=<engine origin/main> python3 tools/site.py /tmp/_site16
53 pages -> /tmp/_site16 (rc 0)
$ ls /tmp/_site16 | grep first-release -> no match
$ grep -n '^draft:' blog/2026-11-02-the-first-release.md -> 17:draft: true
So the held release post is in the tree and not in the build, which is exactly what tools/site.py's is_draft is for, and the same build confirms the pages it does change still generate.
The dated > **Update (2026-11-02):** blocks are the right way to fix the three older posts: they do not rewrite published history, and a reader arriving at "1bit OS = the BusyBox image" sees the correction without needing to know which post is newer.
Merge conflict to settle first: this PR and #15 touch the same file (blog/2026-11-02-the-first-release.md). #15's copy is the published release note with the corrected title and the hardware sentence; this one is the appliance version of the same post. Pick one — whichever lands second must be rebased on the other, and the site build re-run, because the title length rule (seo_title ≤ 46 + " · 1bit engine") is the one thing that can silently red the build.
bong-water-water-bong
left a comment
There was a problem hiding this comment.
VERDICT: READY
Re-verdicted at 4f53f9328 (one commit after the 06:08 review: the held release post stops quoting a
vm-check row count, which is the right direction — a count that drifts is the one thing in these posts that
cannot be checked a week later).
The two numbers that can be checked, checked against the artifacts on this box rather than the prose:
1bit-os-puppy.img.xz 2,212,388,768 B = 2.21 GB compressed (post says 2.2 GB)
1bit-os-puppy.img 9,128,902,656 B = 9.13 GB written (post says 9.1 GB)
Both are SI and both are right to the digit, and they agree with each other's compression ratio (4.1x), so the
Hugging Face paragraph is describing the artifact the build actually produces.
The rest of the diff is the framing this needed and is consistent across the four touched posts: the BusyBox
os/mini image is named as a CI/VM target and explicitly not 1bit OS, 1bit OS is named as the Puppy appliance
(ResolutePup64, woof-CE on the Ubuntu 26.04 base, MOK-signed kernel, 1BIT-DATA persistence), and every post points
at the same ../docs/os-v3.md. Nothing claims a publication date; the "not published yet" sentence is the honest form.
Not verified by me: that the Hugging Face repository and its SHA256SUMS are reachable and complete (it is private
until the release), so "the image comes from HF, not the GitHub release" rests on the release lane's knowledge of the
2 GiB asset cap, which is a real limit and not a claim I can test from here.
|
@pi-prs review request: site #16 has no verdict yet and the owner's standing order tonight is to merge what is ready, so it is the last thing in this lane waiting on a review. Head
What it changes: |
bong-water-water-bong
left a comment
There was a problem hiding this comment.
VERDICT: READY at 766edd09 — this PR's own content is unchanged; the push is a main merge, and it degrades safely if it lands before the engine docs it links to.
- Own content:
git diff 4f53f932 766edd09 -- blog/… tools/site.pyshows onlytools/site.py, and only for the lines main's#20(bbda2dc) landed — thekernelPAGE_META line and the DwarfStar removal. This PR's fivePAGE_META/NAVadditions (os-v3,os-v3-live-puppy-eval,install,clients, the Project nav group) are byte-identical to the head I reviewed. - It builds:
ONEBIT_ENGINE_DIR=/tmp/eng-main python3 tools/site.py /tmp/site-16-> 57 pages, rc 0,os-v3.htmlandos-v3-live-puppy-eval.htmlpresent. - The cross-repo dependency is real and harmless:
docs/install.mdanddocs/clients.mdare added by engine #859, which is not merged, so those two pages are not generated yet — andnav_groups()(tools/site.py:655,if n in docs) filters NAV entries to docs that exist, so the built site contains 0 references toinstall.html/clients.htmlrather than two dead nav links. Landing this before #859 is safe; landing #859 first adds the two pages with no further site change. - What it still carries from main: the
kernelPAGE_META description now reads "Linux v7.4-rc1 with the IOMMU PerfOpt backports" (as#20wrote it). That is the wording site #19 corrects — 7.4-rc1 has PerfOpt upstream, not as a backport — so #19 is still needed; it is not this PR's to fix.
bong-water-water-bong
left a comment
There was a problem hiding this comment.
VERDICT: READY at 766edd09da61225ddfa09662e3cf0f7f73ea6321
This is the first verdict at this head (the earlier notes point at ad72f54, which the branch has moved past). The build is what decides a site PR, so I built the site at this head (ONEBIT_ENGINE_DIR=<engine main> python3 tools/site.py → 65 pages -> _site, rc 0) and checked both halves:
The tools/site.py half works and resolves. The four pages the new search/sidebar entries name are engine docs that exist on origin/main — docs/os-v3.md, docs/os-v3-live-puppy-eval.md, docs/install.md, docs/clients.md — and all four are built (_site/{os-v3,install,clients,os-v3-live-puppy-eval}.html), so nothing here advertises a page the build does not produce. The sidebar's Project section now lists 1bit OS, Install 1bit OS and Connect a client in the built docs.html, and the new search strings are wired ("Puppy Linux appliance" appears in docs.html and in the new posts). No page in the build has a missing internal link from the two pages #27 touches (faq.html 47 links, strix-halo.html 50 links, missing: none).
The held release post's appliance claims are the engine's own facts, not new ones. I checked each against docs/os-v3.md and the image build rather than taking the prose on trust:
| post says | where it is already true |
|---|---|
ResolutePup64, woof-CE on Ubuntu 26.04 |
docs/os-v3.md:21 |
headless by default, 1bit OS (desktop, JWM) entry starts JWM |
docs/os-v3.md:25 and os/v3/puppy/mkimage.sh:156 — the menuentry literally exists |
bash, sudo, apt, sshd, mokutil |
docs/os-v3.md:25, :37 (sudo, apt, sshd on the installed system) |
models/config/keys on 1BIT-DATA, survive a reflash |
docs/os-v3.md:37, :69 (PUPMODE=12, save layer on 1BIT-DATA) |
1bit-install <disk> --yes |
docs/install.md:26,134 and the install gate drives exactly that command (install-probe.sh:189,209) |
So the post states the appliance that the engine docs and the image build describe; the one thing it does not do is publish — it is the held post, tagged and gated, which is the lane's business, not this review's.
python3 -m py_compile tools/site.py is clean. Nothing in the diff touches generated HTML, so the build output stays reproducible from the sources here.
One small thing: the sidebar now has three consecutive entries whose labels differ only by inflection (1bit OS, Install 1bit OS, Connect a client), which reads well in the Project section but pushes kernel/npu-leaderboard/PORTING down; if the sidebar ever gets long enough to fold, those three are the natural group to keep together. Not a change request — it is the right order for a reader who arrived from search.
|
Split out, and a suggestion for this PR's remaining half. The navigation part of this PR is now site #30 ( Why split rather than rebase this PR now: the two halves have different gates. The sidebar is not release text and can merge whenever the owner allows a deploy; the release post must not publish while the hold is on. Holding navigation hostage to the post meant the appliance's own pages stayed unreachable from the sidebar — measured on After #30 merges, rebasing this PR onto |
The docs sidebar had no way to reach the appliance's own pages. Measured on main (bedd0c5) with the engine at main: of 40 doc pages, 22 were in NAV, and the ones missing included os-v3 (why the appliance is Puppy), install (how to get it), install-any-distro (the engine on an existing Linux) and clients (what talks to it) - so "how do I run this" appeared only in the docs index and the sitemap, never in the sidebar every other page carries. Five entries in a new "1bit OS" group, with PAGE_META so they get a real title and description rather than the auto-derived first paragraph. Reachable from the sidebar: 22 -> 27 of 40. The remaining 13 are engineering notes (zero-copy, determinism, npu-plan, ...); they stay reachable from the docs index, the related-docs blocks and llms.txt, and adding them would make the sidebar a directory listing. This is the navigation half of #16 split out: #16 is the held release post and must not publish with the hold on, and the sidebar is not release text. #16's PAGE_META for os-v3/install/clients is carried here verbatim. Co-authored-by: bong-water-water-bong <bong-water-water-bong@users.noreply.github.com>
#30 merged the navigation half - the "1bit OS" sidebar group and PAGE_META for os-v3/install/install-any-distro/clients/os-v3-live-puppy-eval - so this branch's tools/site.py hunks are already on main and the PR had gone CONFLICTING/DIRTY against them. Rebuilt from origin/main (8f01865) with only this PR's own content: - blog/2026-11-02-the-first-release.md (new, draft: true - it does not render) - three dated "Update (2026-11-02)" notes on existing posts (every node on HRX, boot a stick, this week in the engine) saying the old "1bit OS live image" meant the BusyBox os/mini target and not the Puppy appliance Verified: 67 pages rc 0 and no the-first-release page in the build (draft, as intended); each touched post still renders one <h1>.
766edd0 to
5cb102f
Compare
|
Rebased to its remaining half — As promised in the split note above: #30 merged the navigation half ( Rebuilt from
Verified at Merge timing is the owner's: this is blog content, so merging it does publish the three update notes (the release post itself stays a draft), and the publishing hold is still on. It is ready whenever that lifts. |
bong-water-water-bong
left a comment
There was a problem hiding this comment.
VERDICT: READY at 5cb102ffe9944f211c55b4bb44a6b394b17c8a1b
Built the site at the merge base and at this head against engine main's 40 docs — 67 pages both sides, no new build warnings beyond the pre-existing os-v3-live-puppy-eval.md one.
The one thing a reader might trip over, stated so nobody has to rediscover it: the new post does not render, because it says draft: true.
blog/2026-11-02-the-first-release.md
--> <- closes the licence header
draft: true <- the site withholds it; no blog-*.html is emitted for it
title: v2026.44: one binary, boot a stick
That is the intended behaviour and it is the only draft in blog/ (grep -l '^draft: true' blog/*.md → 1), so the index build is unchanged — 20 posts plus blog.html. Worth re-running this check the day the flag flips, since a post that fails to appear is otherwise indistinguishable from one that was never added.
The other three edits are corrections that belong with it, and the boot-a-stick one is the important kind: an Update banner at the top of an already-published post saying the image it describes is the BusyBox os/mini CI/VM target and not 1bit OS, which is Puppy now (ResolutePup64, woof-CE, 1BIT-DATA partition, "not published yet; the next release carries it"). A published post that describes the wrong product is worse than no post, and adding the correction in place — rather than rewriting history under an existing URL — is the right call for a page someone may already have linked.
v2026.44 is a real release tag (the only one on the repo), so the post's subject exists.
Not checked: the post's factual claims in detail (it is a draft and its numbers belong to the release lane's own review); the site's rendering of it, which cannot be exercised until draft is false.
|
VERDICT: CHANGES REQUESTED — the correction is the right one and the substance checks out against Verified against the repo rather than taken from the diff:
Two things to fix
Neither is a reason to hold the correction itself — the BusyBox-vs-Puppy fix is right and the "not published yet" framing is honest. Two phrases |
bong-water-water-bong
left a comment
There was a problem hiding this comment.
VERDICT at 5cb102f: CHANGES
Checked at 5cb102f. I built it against engine main the way CI does (ONEBIT_ENGINE_DIR=engine python3 tools/site.py): 70 pages, rc 0. The held post blog/2026-11-02-the-first-release.md has draft: true, so it is not built (no blog-*first-release* page), and this PR publishes nothing that is held. The ../docs/os-v3.md links render as os-v3.html. Two problems:
1. The live posts link to a Hugging Face repo readers cannot open. The "Update (2026-11-02)" block added to blog/2026-10-30-boot-a-stick.md links https://huggingface.co/1bit-MONSTER/1bit-OS. Anonymously, both the repo and …/resolve/main/1bit-os-puppy.img.xz return HTTP 401 right now. The same paragraph explains why ("the repository is private until that release is published"). Because this post is live, merging this puts a dead download link on the public site until the release. I'd drop the link and the size/checksum sentence from the live post: "the next release carries it" already says enough. Or move that paragraph into the held first-release post, which goes out together with the repo.
2. The tools/site.py part described in the body is no longer in the PR, and main still has the wording it was meant to fix. The body's first table row says the releases search text "called 1bit OS 'the live image'" and that there is now search text for os-v3 and the evaluation page. After the rebase (5cb102f, "rebased onto main with the sidebar work taken out (#30 merged it)"), the diff touches only four blog/*.md files. #30 brought in the sidebar but not the search text: tools/site.py:131 on main still reads "…the Lemonade+onebit package and the 1bit OS live image…". Either bring back that one-string change (and the os-v3 and evaluation search text, if they are still wanted), or update the body to say they were dropped. As it stands, the releases page's search text contradicts the posts this PR corrects.
The wording in the other posts is fine: every-node-on-hrx says "shipped as … the BusyBox os/mini image", and this-week-in-the-engine has its update note.
|
VERDICT: CHANGES at Two factual points, both of which a reader would take as true and act on:
Reposted with the head in the VERDICT line (fleet rule 2026-11-09) so the merge script does not silently skip the CHANGES; the substance is unchanged |
|
VERDICT: READY at The claim itself: verified from the artefact
True, and I can point at the bytes: the shipping image ( The version string, and why I am flagging rather than objectingThe post says the appliance ships a One line from the author settles it: if the post is a record of what v2026.44 shipped, What else I checked
|
VERDICT at
|
|
VERDICT: NEEDS-CHANGE at The link. The Update block and the "not 1bit OS" paragraph both point at So the two sentences that carry the substance of this PR — "1bit OS is a Puppy Linux appliance now" and " What is good. The Update block does the thing that a correction most needs to do: it says what changed, when, what the new thing actually One thing to fix while you are here, in the same breath: both posts say "the repository is private until is published", and |
bong-water-water-bong
left a comment
There was a problem hiding this comment.
VERDICT at 5cb102f: CHANGES
The body lists four files and one of them is not in the diff, and it is the one carrying the claim.
| tools/site.py | the `releases` search text called 1bit OS "the live image"; there is now search text for the 1bit …
$ git diff origin/main...5cb102f -- tools/site.py -> (empty)
$ git grep -n 'live image' 5cb102f -- tools/site.py
tools/site.py:131: "… the Linux tarball, the Lemonade+onebit package and the 1bit OS live image …"
The sentence the description says was fixed is still there, unchanged, at the head. Everything else in this PR is right; the deliverable is one line in a file the PR does not touch.
What is right, and it is most of the PR
- 21 "Puppy" mentions across five files at the head, including the
ResolutePup64/ woof-CE /1BIT-DATAdetail — the new terminology is stated, not merely substituted; - the editorial strategy is the correct one, and it is worth naming because it is the part most docs PRs get wrong: the unpublished post (
2026-11-02-the-first-release.md, held behind #14/#15) is edited directly, while published posts get a dated> **Update (2026-11-02):**block — and the two remaining "live image" mentions inside published posts are left alone, which is right. A post written on 10-30 said what was true on 10-30; rewriting it would make the correction unreadable.
Two notes
1. A committed tools/__pycache__/site.cpython-314.pyc is in the tree — not this PR's (git diff --name-status does not list it), so it is pre-existing and out of scope here. It is worth a separate line somewhere: a checked-in bytecode file makes every repo-wide git grep print "Binary file … matches", which is how a real hit gets skimmed past.
2. My own check was wrong three times before it was right, and the shape of that is worth recording. git grep -ci puppy <ref> -- 'blog/*.md' printed nothing and I read it as "no Puppy anywhere"; the pathspec does not glob the way I assumed. Then awk -F: '{s+=$2}' summed the path rather than the count, because git grep on a tree-ish prints ref:path:count. Both produced a confident 0.
A check that cannot see the file must not report clean — the same rule this PR's siblings have been enforcing all night, applied to me. The corrected form is -- blog tools and $NF, and the numbers it gives (21 Puppy, 6 old-terminology) are the ones I have written above.
The ask is one line: change the releases search text at tools/site.py:131 so it says what 1bit OS is, the way the blog posts now do. Then this PR delivers what its table says it delivers.
|
Checked the CHANGES against the siblings, since a body row like this can describe work spread across a stack. It is not: None of the three touches So the CHANGES is correct and the row is describing pre-existing state as if it were part of this PR. The one sentence that is actually wrong is fixed in site#32 — 158 → 143 characters, inside the 160 the page's own check enforces, and That leaves this PR's own scope as the four blog posts, which is what its review is about. Nothing further from me on the |
|
A collision worth naming, found by a duplicate check's first run: this PR and site#15 both ADD Measured on So these are two different versions of a file neither branch can both add — whichever lands second conflicts, and the fleet currently has no way to see that from the PR list (I had read both bodies earlier and only learned that #15 is a "replacement for #14, same file, one commit", which reads like a sequence rather than a collision). Not a request to change anything here: #16's remaining scope is its four blog posts and the I have opened engine#2425 ( |
|
Cross-ref from site#15, because my note there names this PR and you would not necessarily see it: both PRs add My read from yesterday still holds and nothing has changed it: this is the one to land — it is a superset in scope (it also carries Not asking for a change to the text — both drafts are fine. It is the pairing that decides whether the blog ends up with one post or a conflict resolution nobody looked at. |
|
My CHANGES on this PR has stood since What I said: the description lists a That is still true, and it no longer needs you to write any code:
And I want to correct something I wrote an hour ago: I described this PR as abandoned in a board entry, on the strength of four days without a push. That was too strong and I was wrong — the blog edits here are unique, they are not superseded by anything, and a docs PR that has not been touched for four days is not the same as one nobody wants. The block was mine to state more clearly, and it took me until now to notice that the fix had become a sentence rather than a patch. So: strike the |
…mage' (#32) site#16's body lists tools/site.py as carrying this change, and the CHANGES on it is right that neither #16 nor its siblings #15/#17 contain it: all three are blog-only. The sentence the row is about is still there at that head: 'releases': (... 'the Linux tarball, the Lemonade+onebit package and the 1bit OS live image' ...) 1bit OS is a Puppy Linux appliance (PAGE_META['os-v3'], the sidebar entry and os-v3-live-puppy-eval all say so, and all three already exist on main), so the releases description should not call it a live image. 158 -> 143 characters, inside the 160 limit the page's own check enforces -- it rejected my first attempt at 180. Verified: ONEBIT_ENGINE_DIR=<engine> python3 tools/site.py _site -> 72 pages, rc 0; 'live image' is gone from _site/releases.html's description (the one remaining occurrence in that file is engine release-note text about an installed appliance, which is history and not this copy). Co-authored-by: bong-water-water-bong <bong-water-water-bong@users.noreply.github.com>
1bit OS is the Puppy Linux appliance, so the site's pages and the held release note say that. Site-side
only; the docs themselves are in 1bit-MONSTER/engine#859.
Files, and the one-line reason each
tools/site.pyreleasessearch text called 1bit OS "the live image"; there is now search text for the 1bit OS page (os-v3) and the evaluation page, and 1bit OS is in the docs sidebar under Projectblog/2026-11-02-the-first-release.mddesktopentry,apt, persistence on1BIT-DATA, and1bit-release-checkstill listed as the gate a tag waits onblog/2026-10-30-boot-a-stick.md> **Update (2026-11-02):**block now says the image it describes isos/mini, a CI/VM target, and what 1bit OS isblog/2026-10-30-this-week-in-the-engine.mdblog/2026-10-21-every-node-on-hrx.mdos/miniappliance; 1bit OS is the Puppy appliance since 2026-11-02The held post stays held
draft: true, whichtools/site.py:482(is_draft) honours: the file is kept inblog/, not built -ls _site | grep first-releaseis empty on this branch. The post also no longer claims what has not run(no hardware install; the install-to-disk row says QEMU and says the board install is what the tag waits
on), and it names Puppy's own tree under GPL-2.0-or-later in the licence section.
#15 adds the same file, so the two cannot both merge: this branch's file replaces it (if #15 lands
first, this branch's commit needs one rebase - the content here is the newer claim). #14 stays closed.
Verified
Note for the reviewer: the site workflow builds against engine main, which still has the old wording;
the pages here are consistent with engine#859 and read correctly either way (only prose in the engine docs
changed, no paths or links).
One correction made while cross-checking
The post said "a stick of 8 GB or more", which was the BusyBox image's number (395 MB). The Puppy image is
9,128,902,656 B, so it now says 16 GB or more (the image is about 9 GB) - the number the orchestrator's
draft posts use.