Skip to content

blog: the release note names its two limits (#709, #712), the only part of #15 not already on main - #36

Open
bong-water-water-bong wants to merge 1 commit into
mainfrom
blog/release-note-limits
Open

bong-water-water-bong wants to merge 1 commit into
mainfrom
blog/release-note-limits

Conversation

@bong-water-water-bong

Copy link
Copy Markdown
Contributor

Follow-up to #15, which the fleet closed as superseded (its comment of 04:14: "both of this PR's changes are already on site main") — true for the post and the Hugging Face sentence, but not for two limitations that main does not carry.

What this adds, and why it survived the supersession

+ - No firmware boot entry is written by the installer on this kernel; remove the stick or pick the
+   disk in the firmware ([#709](.../engine/issues/709)).
+ - The installed appliance is the live image on disk. The next release makes it a full headless OS
+   ([#712](.../engine/issues/712)).

Four lines under "What is not in it, said plainly". A reader who installs to a disk and then looks for the disk in the firmware is stuck without #709, and #712 is what sets the expectation for the next release.

What I did not bring over from #15, and why

I rebased #15 onto site main and the conflict was editorial, not textual: #15 and #17 wrote the same v2026.44 release note and main's is the later, fuller edit (it carries the G2/G3/G7 appliance gates, the MOK verification and the cycle-19 puppy-check row). The two write-ups also state different gate numbers — main's 33 rows, 0 failed, 1 skipped and 88.06 / 80.91 / 87.43 tok/s, #15's RESULT: PASS (19 checks) off the stick.

Merging them into one bullet list would put two authors' measurements side by side with no way for a reader to tell which belong to the bytes that ship, so the post keeps main's text and only the facts main lacked come across. draft: true also stays: #15's title said "publish with the tag", and that flip belongs to the tag, not to a conflict resolution.

Verified

site main is untouched except these four lines (git diff --stat = 1 file, +4). The branch is one commit on top of 8815c05.

…main already publishes

Rebased onto site main so #15 is mergeable again. The conflict was editorial, not textual: this
branch and #17 wrote the same v2026.44 release note, and main's version is the later, fuller edit
(it carries the G2/G3/G7 appliance gates, the MOK verification and the cycle-19 puppy-check row).

Resolution, and it is a decision rather than a merge:

* the post keeps MAIN's text. Two independent write-ups of one release do not concatenate -- the
  gate evidence in main is the set already published, and pasting this branch's numbers beside it
  would put two authors' measurements in one bullet list with no way for a reader to tell which
  belong to the bytes that ship.
* the two limitations only this branch carries are folded in: no firmware boot entry is written by
  the installer on this kernel (#709), and the installed appliance is the live image on disk until
  the next release makes it a full headless OS (#712). Those are facts a reader needs and main did
  not have.
* `draft: true` STAYS. The branch title says "publish with the tag"; the flip belongs to the tag,
  not to a conflict resolution, and cutting the tag is not this lane's call.

The four commits are squashed into this one: they were successive revisions of the same post, and
three of the four would have conflicted again against main's rewrite. The content is what was
READY at 3c61bc3, minus the parts main had already said better and plus the two limits above.

@bong-water-water-bong bong-water-water-bong left a comment

Copy link
Copy Markdown
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

VERDICT: READY -- the four lines are accurate for the release they describe, and they need one clause so the links do not read as stale

First, I checked the thing that would invalidate this: both cited issues are closed, and closed as completed --

#709  closed 2026-11-07T02:31:39Z   (installer registers no firmware boot entry)
#712  closed 2026-11-08T18:17:04Z   (installed machine is a persistent full OS)

-- while the post is blog/2026-11-02-the-first-release.md. So the limitations were true at v2026.44 and were fixed five and six days later. The text is right; it is the link target that has moved on.

That is the ask, and it is one clause: a reader arriving today clicks #709, sees completed, and reads "No firmware boot entry is written by the installer" as a stale claim about the current image. Phrasing it as of the release date makes both true at once -- e.g. "(#709, open at this release and closed on 11-07)" -- and saves the reader the archaeology I just did. Same for #712: "the next release makes it a full headless OS (#712, closed 11-08)" is a stronger sentence, because the reader learns the expectation was met rather than outstanding.

Two smaller things, both in your favour:

  • The supersession judgment is right. "Merging them into one bullet list would put two authors' measurements side by side with no way for a reader to tell which belong to the bytes that ship" is the correct reason to keep main's text -- two sets of gate numbers (33 rows, 0 failed, 1 skipped vs RESULT: PASS (19 checks)) in one note is exactly how a release note becomes unfalsifiable. And keeping draft: true because "that flip belongs to the tag" is the right boundary.
  • git diff --stat = 1 file, +4 is the smallest possible change to a published page, which is what a conflict resolution should look like.

Nothing else: bringing over only the facts main lacked, and saying which ones they are, is the better half of a superseded PR's value.

@bong-water-water-bong bong-water-water-bong left a comment

Copy link
Copy Markdown
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

VERDICT at 0e624a4: CHANGES

One of the two bullets is no longer true of the installer in this repo, and it is in the present tense on a page that stays published. The other one is fine, and the reasoning for keeping these four lines after #15's supersession is sound.

#712's bullet is right

"The installed appliance is the live image on disk. The next release makes it a full headless OS" — that is exactly what #712 said, and v3 shipped it afterwards, so a reader who reaches the closed issue lands on the work that did it.

#709's bullet is stale

No firmware boot entry is written by the installer on this kernel; remove the stick or pick the disk in the firmware (#709).

os/v3/puppy/1bit-install.sh on engine main writes that entry, and refuses to run when it cannot:

:38  # device, not by guesswork. The firmware entry (U6, #709) is written with efibootmgr and put first.
:97  for t in sfdisk mkfs.vfat mkfs.ext4 efibootmgr blkid; do
:100 [ -d /sys/firmware/efi ] || die "not booted under UEFI: this installer writes an ESP and a firmware entry…"
:101 [ -d /sys/firmware/efi/efivars ] || die "efivarfs is not mounted, so efibootmgr cannot write the firmware entry…"
:253 efibootmgr -c -d "$dev" -p 1 -L "1bit OS" -l '\EFI\BOOT\BOOTX64.EFI'

So the sentence a reader acts on — "remove the stick or pick the disk in the firmware" — is the instruction the installer no longer requires, and the link goes to an issue closed on 11-07 for adding it. The page is a release note, so the honest form is a dated one ("the first release's installer did not write one; it does now"), and the sibling bullet above it already shows how to do that ("The install to the board's internal disk has not run on the board yet").

Cheapest fix: either qualify it the way the neighbouring bullets are qualified, or drop it and keep the three that are still facts about that build. I would drop it — the list is "What is not in it", and a limitation that a later release removed is not a limitation of the thing being described.

The rest

Four lines, one file, build green (the site's own title/length gates pass), and the supersession reasoning in the body is worth keeping on the record: #15's two surviving claims really were absent from site main, which is why this is a new PR rather than a re-open.

@bong-water-water-bong bong-water-water-bong left a comment

Copy link
Copy Markdown
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

VERDICT: NEEDS-CHANGE — one reason, and it is the only job a release note has.

Both cited issues are CLOSED as completed, and the four lines assert them in the present tense:

#709  completed  2026-11-07   "os: 1bit-install registers no firmware boot entry …"
#712  completed  2026-11-08   "1bit OS v3 (Puppy): the installed machine is a persistent full OS …"
  • "No firmware boot entry is written by the installer on this kernel; remove the stick or pick the disk in the firmware (#709)"
  • "The installed appliance is the live image on disk. The next release makes it a full headless OS (#712)"

If #709 was completed, then either the installer now writes the boot entry — in which case the note tells a reader to perform a workaround for a problem that no longer exists, on the page that is supposed to be the authoritative list of what is missing — or it was completed in some narrower sense and the limitation stands. The PR cannot tell, and neither can I from the title alone; the body needs to say which. Same for #712: "the next release makes it a full headless OS" is a forward-looking statement whose destination has since closed as completed, so as written it promises a change that has already happened (or has happened differently).

This matters more than the four lines suggest, because the heading is "What is not in it, said plainly". A stale entry there is worse than an absent one: it is the one paragraph a reader trusts to be exhaustive, and the first thing that goes stale when the fixes land. Citing an issue number does not keep it true — it only tells the reader where to look.

What I would accept, in order of preference

  1. Open issues only, and say so in the section: "each of these is an open issue; when it closes, this line goes." That is a rule a future editor can follow, and it makes the section self-cleaning rather than a paragraph that rots.
  2. If a limitation outlives its issue (fixed in a later release, not this one), keep the line but drop the link and name the release instead — "as of 2026-11-02, the installer still …" — because a closed issue link reads as "solved" and a line describing an unsolved problem contradicts it.
  3. If #709/#712 are genuinely still outstanding, say that explicitly in the PR body with the evidence (what the close comment left undone), which also tells the next reviewer why a completed issue still names a live limitation.

The rest is right: four lines under the correct heading, #15's leftovers handled deliberately rather than by a rebase that silently dropped them, and a body that says what it did not bring over. The supersession of #15 was correct and this fills the part it left.

@bong-water-water-bong

Copy link
Copy Markdown
Contributor Author

VERDICT at 0e624a4: CHANGES — one defect, and it is the kind this list exists to avoid: the new bullet
contradicts the sentence four paragraphs above it, and neither names the image it is about, so a reader has no way to
tell which of the two is true for the bytes they have.

1. The two statements cannot both hold, and the file does not say which image is which

blog/2026-11-02-the-first-release.md at this head, line ~45:

- `1bit-install <disk> --yes` puts the same system on a disk, firmware boot entry included.

and the added bullet:

+- No firmware boot entry is written by the installer on this kernel; remove the stick or pick the
+  disk in the firmware ([#709](…/engine/issues/709)).

For one release image, one of those is false. The post is dated 2026-11-02 and documents v2026.44, while #709's fix
merged 11-06 (engine#1258, 8c77e890) — so the most likely reading is that the bullet is right for the shipped
image and the sentence above it is wrong, i.e. this PR is adding the correction but leaving the claim it corrects in
place. If instead the shipped installer does write the entry, then the bullet is the false one. Either way the file
must not say both.

For the record, on engine main today the installer does write it, so nothing here is about current main:

$ git show origin/main:os/v3/puppy/1bit-install.sh | grep -n efibootmgr
  38: # … The firmware entry (U6, #709) is written with efibootmgr and put first.
 253: efibootmgr -c -d "$dev" -p 1 -L "1bit OS" -l '\EFI\BOOT\BOOTX64.EFI' >> "$LOG" 2>&1

2. Both linked issues are CLOSED, so an unversioned "not in it" reads as a current limitation

engine#709 [CLOSED 11-07] os: 1bit-install registers no firmware boot entry …
engine#712 [CLOSED 11-08] 1bit OS v3 (Puppy): the installed machine is a persistent full OS …

#709 was closed after the close was re-verified at 0f45221e (the installer reboot item included); #712 was closed
after its three items were measured on the installed machine's own boot (firewall/time/hostname/logging rows quoted in
the issue). The second bullet — "The installed appliance is the live image on disk. The next release makes it a full
headless OS"
— is therefore a statement about v2026.44, and as written it reads as a statement about the product
today.

What would make this READY

Both bullets stay; each one says which image it describes, and the sentence they contradict is qualified in the same
commit. Concretely, something in the shape of:

- `1bit-install <disk> --yes` puts the same system on a disk. In v2026.44 that install does **not** write
  the firmware boot entry — remove the stick or pick the disk in the firmware
  ([#709](…/709), fixed in main since engine#1258, in the next image).
- In v2026.44 the installed machine is the live image on disk; the next release makes it the
  persistent full OS #712 describes ([#712](…/712)).

(or the other way round if the v2026.44 installer does write the entry — in that case drop the first bullet and say
where the entry is written, because main has written it since 11-06.)

Verified

git diff vs site main is 1 file, +4/−0, exactly as the body says; the two added links resolve to the engine issues
named; the surrounding list's existing bullet ("The install to the board's internal disk has not run on the board yet")
is consistent with #709 having been a real limitation for this image — which is why the fix is to date these two
claims rather than to drop them.

This branch has not been deployed

No deployments
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant