8000 testing pkgmgr by jtnkminimal · Pull Request #1 · gominimal/pkgs · GitHub
[go: up one dir, main page]

Skip to content

testing pkgmgr - #1

Merged
twitchyliquid64 merged 2 commits into
mainfrom
update/pkgmgr-test
Jan 16, 2026
Merged

twitchyliquid64 merged 2 commits into
mainfrom
update/pkgmgr-test

Conversation

@jtnkminimal
Copy link
Copy Markdown
Contributor

No description provided.

@twitchyliquid64
twitchyliquid64 merged commit 8bf4b98 into main Jan 16, 2026
@twitchyliquid64
twitchyliquid64 deleted the update/pkgmgr-test branch January 16, 2026 18:29
bryan-minimal added a commit that referenced this pull request Aug 4, 2026
First package retired off the foreign prebuilts. m4 was chosen first because it is free
under both cut models, its prebuilt is foreign and unattested (zero .intoto envelopes in
gs://minimal-staging-archives/prebuilts/**, and that bucket appears nowhere in
trust-config), and it has the highest fan-in of the cheap targets (34).

DELETE, not de-cycle: breaker_hydrate.rs:211-217 is the entire gate and it never tests for
a cycle — it is replace_always. While the block exists the from-source recipe never runs,
so breaking cycles in the graph removes zero substitutions.

Measured basis for doing it this way (rd83, 2026-08-04): building 'base' on arm against a
trust-config whose gcc/glibc/binutils bootstrap_artifacts anchors pointed at our sealed
bedrock ladder produced AB_BUILD_OK with our shas mentioned ZERO times, while 58
hydrate/twin lines fired from prebuilts/. Trust-config anchors cannot reach the base soup;
per-package build.ncl surgery is the only lever.

UNVERIFIED at commit time: whether m4 builds from source cleanly without its breaker. That
is the next measurement, not an assumption.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
bryan-minimal added a commit that referenced this pull request Aug 4, 2026
The first attempt used brace-matching from the 'replace_on_cycle =' offset and consumed one
brace too many, leaving a malformed record — Nickel decode failed at graph load (which is
at least a fail-shut error, not a silent mis-parse). The block is lines 37-63; cutting that
range leaves outputs closing cleanly and attrs following directly. Verified: zero
replace_on_cycle and zero prebuilts/ references remain.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
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.

3 participants

0