Repository navigation
Update stacked: 8 packages - #606
Merged
Merged
Conversation
…rin failure No source change. The amd64 tamarin-prover failure (alex version detection) is either a store race under --jobs=$(nproc), which is intermittent, or something deterministic. Re-running the identical tree distinguishes them, which is the one fact pkgs#609 is reasoning without. Expected to fail again — tamarin on main is still broken. What matters is WHETHER it fails the same way.
bryan-minimal
added a commit
that referenced
this pull request
Aug 14, 2026
…aming The previous commit said tamarin "builds on arm64 and fails on amd64", and treated it as the onboarding's amd64-unverified flag finally coming due. That was wrong, and wrong in a way that would mislead the next reader. pkgs#607 and #608 built the SAME tamarin source on the SAME amd64 builder and passed — merged 20:28 and 20:30 on 2026-08-14, hours after #606 failed at 12:33. So amd64 works most of the time. It is not a miscompile, not a missing dependency, and not something #606 introduced: it is a race, and #606 drew the short straw. This strengthens the fix rather than weakening it — removing cabal's own store-provisioned alex removes the racy step — but the diagnosis in the comment should say what is actually true. Worth recording why a flake is the worse shape here: the natural response is to re-run it, which usually works, which is exactly how one survives for weeks without anyone fixing it. If this change does NOT settle it, the next suspect is `--jobs=$(nproc)` itself rather than alex specifically, since the same race can bite any build tool. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
twitchyliquid64
approved these changes
Aug 17, 2026
twitchyliquid64
pushed a commit
to evanspearman/pkgs
that referenced
this pull request
Aug 17, 2026
…ntermittent amd64 failure) (gominimal#609) * tamarin-prover: use the packaged alex/happy instead of cabal's own (amd64) tamarin builds on arm64 and fails on amd64. From pkgs#606's run (2026-08-14 — tamarin is NOT in that PR; the buildbot builds the whole graph, so it failed as collateral): Error: [Cabal-1008] The program 'alex' version >=3.1.4 is required but the version of /build/.cabal-home/store/ghc-9.10.3-inplace/alex-3.5.4.2-.../bin/alex could not be determined. This is NOT a compile error. Cabal BUILT alex successfully ("Completed alex-3.5.4.2 (exe:alex)") and then could not get a version out of it when `language-javascript` configured. The binary exists and will not answer — the shape you get from a partially-installed store entry under `--jobs=$(nproc)`, or from an invocation killed for memory. The arches differ in core count, which fits it failing on one and not the other. Both alex and happy are already PACKAGED for both arches, and tamarin was not using them — cabal was provisioning its own from Hackage. So add them to build_deps and point cabal at them with `--with-alex` / `--with-happy`. That removes the failing step rather than tuning around it, and drops a from-Hackage build out of a pioneering GHC-9.10 port that already carries four source patches. Not a gominimal#606 regression, and not new: the formal-methods onboarding shipped this arm64-proven with amd64 explicitly flagged unverified (this fleet develops on arm64). This is that flag coming due. CANNOT BE REPRODUCED LOCALLY — no amd64 here, and Haskell will not build on macOS at all ("sandbox execution is only supported on Linux"). So the build asserts the packaged tools exist AND can report a version, printing both paths first. If the packaged tools turn out to be the problem rather than the fix, it now fails at that assertion with the evidence in hand, instead of 40 minutes later inside a dependency's configure step with no way to tell which alex cabal used. The next amd64 run is the experiment. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * tamarin-prover: it is INTERMITTENT, not broken — correcting my own framing The previous commit said tamarin "builds on arm64 and fails on amd64", and treated it as the onboarding's amd64-unverified flag finally coming due. That was wrong, and wrong in a way that would mislead the next reader. pkgs#607 and gominimal#608 built the SAME tamarin source on the SAME amd64 builder and passed — merged 20:28 and 20:30 on 2026-08-14, hours after gominimal#606 failed at 12:33. So amd64 works most of the time. It is not a miscompile, not a missing dependency, and not something gominimal#606 introduced: it is a race, and gominimal#606 drew the short straw. This strengthens the fix rather than weakening it — removing cabal's own store-provisioned alex removes the racy step — but the diagnosis in the comment should say what is actually true. Worth recording why a flake is the worse shape here: the natural response is to re-run it, which usually works, which is exactly how one survives for weeks without anyone fixing it. If this change does NOT settle it, the next suspect is `--jobs=$(nproc)` itself rather than alex specifically, since the same race can bite any build tool. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 5 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Update base-soup (8 packages)
Note
These packages declare
replace_on_cyclein their build.ncl, sothey participate in the toolchain rebuild graph and one hash change
cascades through the set. Bundling ensures the cascading rebuild
lands as a single unit, even when only one package is bumping —
avoids back-to-back full rebuilds from singleton PRs.
Warning
2 requested members did NOT ship — each needs separate follow-up.
A member is dropped when it fails to update, has its gs:// mirror
withheld by the scan gate, or is peeled to keep the shipped set a
closed dependency closure. Reasons are abbreviated — the full
diagnostics ride the Slack thread and the run's logs.
llama.cppb10357b10433railway5.35.15.41.0Pkgscan: 1 below-threshold signal across bundle members (all Info severity, risk score 0.1). Demoted in benign contexts (test fixtures, CI workflow setup); expand for details.
nss (upstream release)metadata/recent-bumprecency NOT evaluated — no release date from the version sourceNote
Build risk — 53 dependents across the tree. Package(s) that
build- or runtime-depend on a member of this bundle may need a rebuild,
or could FTBFS on an API/ABI change. Informational (not blocking) — a
heads-up for the reviewer on what this bump can ripple into.
goact,age,bun,caddy,chezmoi,cloud-sql-proxy,cloudflared,cosign(+37 more)harfbuzzchromium-bin,chromium-headless-shell-bin,ffmpeg,libass,pangonsschromium-bin,chromium-headless-shell-binuvdiffoscopeSummary
1.26.51.26.6github:golang/go:tag:operator-pinned14.3.014.3.1github:harfbuzz/harfbuzz:release-asset:operator-pinned3.1263.126.1override:nss:operator-pinned0.14.70.14.8github:ClementTsang/bottom:operator-pinned0.12.30.12.4github:astral-sh/uv:operator-pinned1.6.01.7.0github:golang/vuln:tag:operator-pinned3.256.03.257.0github:pulumi/pulumi:operator-pinned0.16.20.16.3github:astral-sh/ruff:operator-pinnedWarning
1 known vulnerabilities still affect this bundle after update. Run
pkgmgr vulnsfor details.Per-package details
go 1.26.5 → 1.26.65c2c3b16caefa1d9...→708effb774be8237...https://go.dev/dl/go1.26.5.linux-amd64.tar.gz→variants.amd64_urlBSD-3-Clause(source: GitHub + tarball)harfbuzz 14.3.0 → 14.3.116070d77cfc4ba1f...→9dae9538aae2ffdf...gs://minimal-staging-archives/harfbuzz-14.3.0.tar.xz→gs://minimal-staging-archives/harfbuzz-14.3.1.tar.xzMIT-Modern-Variant(source: tarball)nss 3.126 → 3.126.17aa07c758be1453d...→d5fcd5e3ee92b9dd...gs://minimal-staging-archives/nss-3.126.tar.gz→gs://minimal-staging-archives/nss-3.126.1.tar.gzMPL-2.0(source: tarball)bottom 0.14.7 → 0.14.8249fca7809224602...→be10adada9ec1e5d...gs://minimal-staging-archives/ClementTsang/bottom/0.14.7.tar.gz→gs://minimal-staging-archives/ClementTsang/bottom/0.14.8.tar.gzMIT(source: GitHub + tarball)uv 0.12.3 → 0.12.47d95d35a941135b9...→49d16c1451d69309...gs://minimal-staging-archives/uv-0.12.3.tar.gz→gs://minimal-staging-archives/uv-0.12.4.tar.gzMIT OR Apache-2.0(source: GitHub + tarball)govulncheck 1.6.0 → 1.7.070f82d70f3a6757b...→4fb7f0204b7e039f...https://github.com/golang/vuln/archive/refs/tags/v1.6.0.tar.gz→https://github.com/golang/vuln/archive/refs/tags/v1.7.0.tar.gzBSD-3-Clause(source: GitHub + tarball)pulumi 3.256.0 → 3.257.043887337b91f4d61...→e51055e520184b99...gs://minimal-staging-archives/pulumi/pulumi/v3.256.0.tar.gz→gs://minimal-staging-archives/pulumi/pulumi/v3.257.0.tar.gzApache-2.0(source: GitHub + tarball)ruff 0.16.2 → 0.16.3480cd1332b218840...→1eecc23082f91b6b...gs://minimal-staging-archives/astral-sh/ruff/0.16.2.tar.gz→gs://minimal-staging-archives/astral-sh/ruff/0.16.3.tar.gzMIT(source: GitHub)Created by pkgmgr