Pin the shipped Piccolo to the vetted 1.19 series - #223
Merged
Conversation
`packages/extension/julia/Project.toml` had no [compat] section at all, so the
only thing holding the shipped environment at the vetted Piccolo was the
Manifest. That is enough for a fresh provision — julia_setup.ts copies both
files verbatim and instantiates — but not for anything that re-resolves.
Measured against the registry:
no compat: Pkg.update("Piccolo") -> 1.19.0 becomes 1.20.0
~1.19: Pkg.update("Piccolo") -> holds at 1.19.0
Piccolo 1.20.0 was released today and is deliberately breaking:
`integrator_type = :spline` now errors instead of silently returning the
piecewise-constant integrator, `fidelity(qcp)` became free-phase-aware and
returns a different number for every `free_phase = true` problem, and
`sync_trajectory!` now warns on optimizer-vs-rollout divergence. Every shipped
score and template was vetted against 1.19.0.
So the move to 1.20 is a decision to make deliberately, alongside re-vetting the
templates, rather than something a resolver does on a user's machine mid-session.
`~1.19` admits 1.19.x patches and refuses 1.20.
Verified: instantiate still resolves 1.19.0, `using Piccolo` loads, and the
Manifest is byte-identical afterwards — no manifest churn.
Deliberately not in scope: CairoMakie / JLD2 / TOML carry the same class of
exposure but are third-party, undemonstrated, and a tight bound on Makie in
particular risks resolution failures I cannot test tonight. Worth a follow-up
that bounds them at the minor they are pinned to.
This was referenced Aug 22, 2026
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
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
packages/extension/julia/Project.toml— the environmentjulia_setup.tscopies to~/.amico/juliaand instantiates — has no[compat]section at all. The shippedManifest.tomlpins Piccolo1.19.0by exact tree-sha, which covers a fresh provision, but nothing covers a re-resolve.Piccolo 1.20.0 released today, and it is deliberately breaking.
Measured, not assumed
Against the real registry, on a copy of the shipped
Project.toml+Manifest.tomlpair:Pkg.update("Piccolo")1.19.0→1.20.0Piccolo = "~1.19"1.19.0So one
Pkg.update— or any event that makes Pkg re-resolve — silently moves a user onto a Piccolo nobody vetted the shipped content against.Why 1.20 is not a drop-in
integrator_type = :splinenow errors instead of silently returning the piecewise-constantBilinearIntegrator.fidelity(qcp)became free-phase-aware, so it returns a different number for everyfree_phase = trueproblem.sync_trajectory!now emits a divergence@warnby default, which lands in run-dir output that the extension parses and users read.Every shipped score and template — including the vetted P(MIS) = 0.9947 figure on the Pasqal golden path — was measured against 1.19.0. Moving minors is a decision to make deliberately, alongside re-vetting, not something a resolver does on someone's machine mid-session.
~1.19admits 1.19.x patches (non-breaking under SemVer) and refuses 1.20.Verification
Pkg.instantiate()still resolves Piccolo1.19.0.using Piccololoads.Manifest.tomlis byte-identical afterwards — the bound needs no manifest churn, so this PR is one file.Pkg.update("Piccolo")holds at1.19.0.Note on convention
packages/schema/julia/Project.tomlalready carries a[compat](JSONSchema = "1",julia = "1.10"). The extension env — the one users actually run solves in — was the outlier with none.Deliberately out of scope
CairoMakie/JLD2/TOMLcarry the same class of exposure, but they are third-party, the risk is undemonstrated, and a tight bound on Makie in particular risks resolution failures I could not test tonight. Worth a follow-up bounding them at the minor they are already pinned to.Raising this to
~1.20should happen in the same PR that re-vets the templates against the new minor.