[dependabot]

Updates the requirements on KernelAbstractions and JET to permit the latest version.
Updates KernelAbstractions to 0.10.4

Release notes

Sourced from KernelAbstractions's releases.

v0.10.4

KernelAbstractions v0.10.4

Diff since v0.10.3

Merged pull requests:

• Less compilation for a kernel's first launch (#931) (@​maleadt)
• Replace the benchmarks with a suite for the JuliaGPU benchmark history (#932) (@​maleadt)
• Bump benchmark-action, to flag benchmarks that start allocating (#933) (@​maleadt)
• POCL: cache the device's Float16 and Float64 support (#934) (@​maleadt)
• POCL: copy small arrays from the host in copyto! (#935) (@​maleadt)
• Only test Julia 1.10 and 1.13 on pull requests in CI (#936) (@​maleadt)
• Look up fold_threshold behind an inference barrier, so that loading a backend doesn't invalidate launches (#937) (@​maleadt)

Commits

• 14002d7 Bump version.
• c10ff8c Less compilation for a kernel's first launch (#931)
• 466ad8a Look up fold_threshold behind an inference barrier, so that loading a backend...
• b3daae8 Only test Julia 1.10 and 1.13 on pull requests in CI (#936)
• bf8ebef POCL: copy small arrays from the host in copyto! (#935)
• 21b3454 POCL: cache the device's Float16 and Float64 support (#934)
• d8788f8 Bump benchmark-action, to flag benchmarks that start allocating (#933)
• d2ada37 Replace the benchmarks with a suite for the JuliaGPU benchmark history (#932)
• 4cc35fe Bump version.
• 5dd8a01 Require PoCL's NUMA-aware scheduling, and measure the NUMA example (#928)
• Additional commits viewable in compare view

Updates JET to 0.13.1

Release notes

Sourced from JET's releases.

v0.13.1

JET v0.13.1

Diff since v0.13.0

Changelog

Sourced from JET's changelog.

[0.13.1]

Changed

• Improved the performance of the error analysis of JETAnalyzer, e.g. report_call and report_package, on Julia 1.14.

Fixed

• Fixed false positive reports on Julia 1.14 and later with code coverage enabled, e.g. under Pkg.test(; coverage=true), where JET could reuse the analysis results cached during its precompilation and fail to constant-fold calls:

  const SV = Core.svec(1, 2, 3)
  f() = sin(length(SV) == 3 ? 1.0 : "1")
  @test_call f() # failed, reporting no method for `sin(::String)`

[0.13.0]

Added

• Added support for Julia v1.14. JET now loads full functionality on Julia 1.14 instead of empty stubs.

• Added the allow_unsupported_julia preference, which loads full JET functionality on unsupported Julia versions, e.g. nightly builds, without the development features that JET_DEV_MODE also enables:

  LocalPreferences.toml

  [JET]
  allow_unsupported_julia = true # previously `JET_DEV_MODE = true` was needed

  JET emits a warning when it is loaded on an unsupported Julia version this way, since it may not work correctly there.

• JETInterface now exports has_problems, the predicate that call_test_ex and func_test use to decide whether an analysis result fails a test.

• Calls whose matching methods are ambiguous are now reported as AmbiguousMethodErrorReport instead of MethodErrorReport saying that no matching method is found. For a union-split call, the ambiguous signatures are reported separately from those without any matching method:

  k(::Int, ::Any) = 1
  k(::Any, ::Int) = 2
  m(x::Int) = k(x, x)               # reported: ambiguous `k(::Int, ::Int)`
  n(x::Union{Int,String}) = k(x, x) # reported: ambiguous `k(::Int, ::Int)`
                                    # and no method for `k(::String, ::String)`

... (truncated)

Commits

• See full diff in compare view

Updates KernelAbstractions to 0.10.4

Release notes

Sourced from KernelAbstractions's releases.

v0.10.4

KernelAbstractions v0.10.4

Diff since v0.10.3

Merged pull requests:

• Less compilation for a kernel's first launch (#931) (@​maleadt)
• Replace the benchmarks with a suite for the JuliaGPU benchmark history (#932) (@​maleadt)
• Bump benchmark-action, to flag benchmarks that start allocating (#933) (@​maleadt)
• POCL: cache the device's Float16 and Float64 support (#934) (@​maleadt)
• POCL: copy small arrays from the host in copyto! (#935) (@​maleadt)
• Only test Julia 1.10 and 1.13 on pull requests in CI (#936) (@​maleadt)
• Look up fold_threshold behind an inference barrier, so that loading a backend doesn't invalidate launches (#937) (@​maleadt)

Commits

• 14002d7 Bump version.
• c10ff8c Less compilation for a kernel's first launch (#931)
• 466ad8a Look up fold_threshold behind an inference barrier, so that loading a backend...
• b3daae8 Only test Julia 1.10 and 1.13 on pull requests in CI (#936)
• bf8ebef POCL: copy small arrays from the host in copyto! (#935)
• 21b3454 POCL: cache the device's Float16 and Float64 support (#934)
• d8788f8 Bump benchmark-action, to flag benchmarks that start allocating (#933)
• d2ada37 Replace the benchmarks with a suite for the JuliaGPU benchmark history (#932)
• 4cc35fe Bump version.
• 5dd8a01 Require PoCL's NUMA-aware scheduling, and measure the NUMA example (#928)
• Additional commits viewable in compare view

Updates JET to 0.13.1

Release notes

Sourced from JET's releases.

v0.13.1

JET v0.13.1

Diff since v0.13.0

Changelog

Sourced from JET's changelog.

[0.13.1]

Changed

• Improved the performance of the error analysis of JETAnalyzer, e.g. report_call and report_package, on Julia 1.14.

Fixed

• Fixed false positive reports on Julia 1.14 and later with code coverage enabled, e.g. under Pkg.test(; coverage=true), where JET could reuse the analysis results cached during its precompilation and fail to constant-fold calls:

  const SV = Core.svec(1, 2, 3)
  f() = sin(length(SV) == 3 ? 1.0 : "1")
  @test_call f() # failed, reporting no method for `sin(::String)`

[0.13.0]

Added

• Added support for Julia v1.14. JET now loads full functionality on Julia 1.14 instead of empty stubs.

• Added the allow_unsupported_julia preference, which loads full JET functionality on unsupported Julia versions, e.g. nightly builds, without the development features that JET_DEV_MODE also enables:

  LocalPreferences.toml

  [JET]
  allow_unsupported_julia = true # previously `JET_DEV_MODE = true` was needed

  JET emits a warning when it is loaded on an unsupported Julia version this way, since it may not work correctly there.

• JETInterface now exports has_problems, the predicate that call_test_ex and func_test use to decide whether an analysis result fails a test.

• Calls whose matching methods are ambiguous are now reported as AmbiguousMethodErrorReport instead of MethodErrorReport saying that no matching method is found. For a union-split call, the ambiguous signatures are reported separately from those without any matching method:

  k(::Int, ::Any) = 1
  k(::Any, ::Int) = 2
  m(x::Int) = k(x, x)               # reported: ambiguous `k(::Int, ::Int)`
  n(x::Union{Int,String}) = k(x, x) # reported: ambiguous `k(::Int, ::Int)`
                                    # and no method for `k(::String, ::String)`

... (truncated)

Commits

• See full diff in compare view

[ChrisRackauckas-Claude]

🤖 Automated comment from an AI agent running as @ChrisRackauckas — not written or reviewed by Chris.

Independent review (Codex CLI 0.164.0 / gpt-6.1-sol), session 01a19658-b5c6-7a70-9f01-f7b06b4d8e50, reviewed head 8f1a367

• Verdict: CHANGES
• Risk: low (L5: upper-bound widening only, with no source or numerical changes; L3: test-environment compat).
• Round / slot: 2 / 1
• Blocking priorities: P2

Review written by the reviewer named above in a session other than the PR author's, and copied here by the posting agent without change to its findings. The posting agent did not re-run the reviewer's commands.

Summary: Widens JET and KernelAbstractions compatibility in the package and main test environment. CUDA CI fails during resolution, and passing jobs use older versions, leaving the new compatibility unverified.

Blast radius: Dependency resolution for package consumers, CPU/GPU kernel backends, and JET-based QA; source algorithms and defaults are unchanged.

Evidence: Blind diff assessment was written and hashed before reading author context. Head matches the live PR head. CPU and QA CI passed with KernelAbstractions 0.9.45 and JET 0.12.3 (LTS uses JET 0.9.18); CUDA CI failed before running tests. Recent main CUDA CI passed with the old dependencies. This is a dependency update, not a bug fix; no regression test was added and no local failing-before/passing-after run was performed.

Blocking findings

1. [P2] Newly admitted versions cannot reach CUDA tests — test/Project.toml:19 and Project.toml:37. Confirmed by running the read-only CI-log commands: CUDA run 41111130585 exits 1 during dependency resolution, reporting JET restricted to 0.13 while SciMLTesting permits only 0.9.0–0.12.3, with an underlying empty intersection between KernelAbstractions@0.9.45 and project compatibility 0.10. Expected: the new dependency combination resolves and the CUDA assertions execute. Passing CPU/QA jobs resolve older versions, and the main-branch CUDA run 41196054894 passes with KernelAbstractions 0.9.45, so it does not validate 0.10. Coordinate compatible dependency releases/bounds and obtain a successful GPU run using KernelAbstractions 0.10; also exercise JET 0.13 in its actual QA environment. Until then G2 requires CHANGES (needs-verification).

Non-blocking findings

None.

What I ran

• git fetch origin "+refs/pull/158/head:refs/remotes/pr/158" main; git checkout --detach refs/remotes/pr/158; git rev-parse HEAD origin/main; git fetch origin "+refs/heads/main:refs/remotes/origin/main"; git merge-base origin/main HEAD; git diff origin/main...HEAD (all exit 0, executed in the source checkout). Reviewed head is 8f1a367; current main is 4c6db84; merge-base and PR base metadata are 2bf791b. The entire PR diff is four compat replacements across two files.
• gh pr view 158 --repo SciML/ParallelParticleSwarms.jl --json number,url,state,isDraft,headRefOid,baseRefName,baseRefOid (exit 0: OPEN, not draft, target head unchanged); later read title, body, commits, and statusCheckRollup only after hashing the independent assessment. No linked bug issue or author risk assessment was present.
• git log -1 --no-merges --format=%B (exit 0): no Agent-Harness or Agent-Model trailers.
• gh pr checks 158 --repo SciML/ParallelParticleSwarms.jl (exit 1): CUDA failed; CPU/Core, QA, documentation, downgrade, benchmarks, formatting and spell checks passed; Buildkite build 286 remains pending.
• gh run list --repo SciML/ParallelParticleSwarms.jl --branch main --event push --limit 5 --json databaseId,headSha,status,conclusion,url,name (exit 0): current main GPU Tests succeeds, unlike the PR CUDA resolution failure.
• gh run view 41111130585 --repo SciML/ParallelParticleSwarms.jl --log-failed; gh run view 41111131338 --repo SciML/ParallelParticleSwarms.jl --log; gh run view 41196054894 --repo SciML/ParallelParticleSwarms.jl --log (exit 0 after redirecting the GitHub CLI cache to a writable scratch directory; initial log retrieval exited 1 because its default cache is read-only). Inspected PR CUDA failure, PR CPU/QA logs, and main CUDA success. PR CUDA output ends in resolver failure and exit code 1; CPU/QA and main CUDA output report ParallelParticleSwarms tests passed using the old dependency versions.
• No local Julia process or test suite was run. CI used Julia 1.13.3 for the stable CUDA job; local execution was not needed to establish the resolver failure.

What I did not verify

KernelAbstractions 0.10 CPU/GPU execution and JET 0.13 QA are unverified. No local GPU, AD, MPI, Reactant, full package, or downstream suites were run. There is no local failing-before/passing-after result because this PR is not a bug fix. Buildkite's pending run supplies no completed GPU evidence. The separate QA environment still caps JET at 0.12 in test/qa/Project.toml:10.

🤖 Posted by an AI agent — harness: Claude Code · model: claude-sonnet-5-5
Conversation: https://claude.ai/code/session_01379heiiB35TtywhR7uSU74