Files
nixpkgs/.github/workflows/eval.yml
Wolfgang Walther b523f257ac workflows/eval: test all available versions
With this change, we start running Eval on all available Lix and Nix
versions. Because this requires a lot of resources, this complete test
is only run when `ci/pinned.json` is updated.

The resulting outpaths are checked for consistency with the target
branch. A difference will cause the `report` job to fail, thus blocking
the merge, ensuring Eval consistency for Nixpkgs across different
versions.

This implements a kind of "ratchet style" check: Since we originally
confirmed that the versions currently in Nixpkgs at the time of this
commit match Eval behavior of Nix 2.3, we can ensure consistency with
Nix 2.3 down the road, even without testing for it explicitly.

There had been one regression in Eval consistency for Nix between 2.18
and 2.24 - two tests in `tests.devShellTools` produce different results
between Lix 2.91+ (which was forked from Nix 2.18) and Nix 2.24+. I
assume it's unlikely that such a change would be "fixed" by now, thus I
added an exception for these.

As a bonus, we also present the total time in seconds it takes for Eval
to complete for every tested version in a summary table. This allows us
to easily see performance improvements for Eval due to version updates.
At this stage, this time only includes the "outpaths" step of Eval, but
not the generation of attrpaths beforehand.
2025-08-12 10:14:01 +02:00

392 lines
15 KiB
YAML

name: Eval
on:
workflow_call:
inputs:
mergedSha:
required: true
type: string
targetSha:
type: string
systems:
required: true
type: string
testVersions:
required: false
default: false
type: boolean
secrets:
OWNER_APP_PRIVATE_KEY:
required: false
permissions: {}
defaults:
run:
shell: bash
jobs:
versions:
if: inputs.testVersions
runs-on: ubuntu-24.04-arm
outputs:
versions: ${{ steps.versions.outputs.versions }}
steps:
- uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8 # v5.0.0
with:
path: trusted
sparse-checkout: |
ci/supportedVersions.nix
- name: Check out the PR at the test merge commit
uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8 # v5.0.0
with:
ref: ${{ inputs.mergedSha }}
path: untrusted
sparse-checkout: |
ci/pinned.json
- name: Install Nix
uses: cachix/install-nix-action@fc6e360bedc9ee72d75e701397f0bb30dce77568 # v31
- name: Load supported versions
id: versions
run: |
echo "versions=$(trusted/ci/supportedVersions.nix --arg pinnedJson untrusted/ci/pinned.json)" >> "$GITHUB_OUTPUT"
eval:
runs-on: ubuntu-24.04-arm
needs: versions
if: ${{ !cancelled() }}
strategy:
fail-fast: false
matrix:
system: ${{ fromJSON(inputs.systems) }}
version:
- "" # Default Eval triggering rebuild labels and such.
- ${{ fromJSON(needs.versions.outputs.versions || '[]') }} # Only for ci/pinned.json updates.
# Failures for versioned Evals will be collected in a separate job below
# to not interrupt main Eval's compare step.
continue-on-error: ${{ matrix.version != '' }}
name: ${{ matrix.system }}${{ matrix.version && format(' @ {0}', matrix.version) || '' }}
outputs:
targetRunId: ${{ steps.targetRunId.outputs.targetRunId }}
timeout-minutes: 15
steps:
# This is not supposed to be used and just acts as a fallback.
# Without swap, when Eval runs OOM, it will fail badly with a
# job that is sometimes not interruptible anymore.
# If Eval starts swapping, decrease chunkSize to keep it fast.
- name: Enable swap
run: |
sudo fallocate -l 10G /swap
sudo chmod 600 /swap
sudo mkswap /swap
sudo swapon /swap
- uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8 # v5.0.0
with:
sparse-checkout: .github/actions
- name: Check out the PR at the test merge commit
uses: ./.github/actions/get-merge-commit
with:
mergedSha: ${{ inputs.mergedSha }}
merged-as-untrusted: true
pinnedFrom: untrusted
- name: Install Nix
uses: cachix/install-nix-action@fc6e360bedc9ee72d75e701397f0bb30dce77568 # v31
- name: Evaluate the ${{ matrix.system }} output paths for all derivation attributes
env:
MATRIX_SYSTEM: ${{ matrix.system }}
MATRIX_VERSION: ${{ matrix.version || 'nixVersions.latest' }}
run: |
nix-build untrusted/ci --arg nixpkgs ./pinned -A eval.singleSystem \
--argstr evalSystem "$MATRIX_SYSTEM" \
--arg chunkSize 8000 \
--argstr nixPath "$MATRIX_VERSION" \
--out-link merged
# If it uses too much memory, slightly decrease chunkSize
- name: Upload the output paths and eval stats
uses: actions/upload-artifact@ea165f8d65b6e75b540449e92b4886f43607fa02 # v4.6.2
with:
name: ${{ matrix.version && format('{0}-', matrix.version) || '' }}merged-${{ matrix.system }}
path: merged/*
- name: Log current API rate limits
env:
GH_TOKEN: ${{ github.token }}
run: gh api /rate_limit | jq
- name: Get target run id
if: inputs.targetSha
id: targetRunId
uses: actions/github-script@60a0d83039c74a4aee543508d2ffcb1c3799cdea # v7.0.1
env:
MATRIX_SYSTEM: ${{ matrix.system }}
TARGET_SHA: ${{ inputs.targetSha }}
with:
script: |
const system = process.env.MATRIX_SYSTEM
const targetSha = process.env.TARGET_SHA
let run_id
try {
run_id = (await github.rest.actions.listWorkflowRuns({
...context.repo,
workflow_id: 'push.yml',
event: 'push',
head_sha: targetSha
})).data.workflow_runs[0].id
} catch {
throw new Error(`Could not find a push.yml workflow run for ${targetSha}.`)
}
// Waiting 120 * 5 sec = 10 min. max.
// Eval takes max 5-6 minutes, normally.
for (let i = 0; i < 120; i++) {
const result = await github.rest.actions.listWorkflowRunArtifacts({
...context.repo,
run_id,
name: `merged-${system}`
})
if (result.data.total_count > 0) {
core.setOutput('targetRunId', run_id)
return
}
await new Promise(resolve => setTimeout(resolve, 5000))
}
// No artifact found at this stage. This usually means that Eval failed on the target branch.
// This should only happen when Eval is broken on the target branch and this PR fixes it.
// Continue without targetRunId to skip the remaining steps, but pass the job.
- name: Log current API rate limits
env:
GH_TOKEN: ${{ github.token }}
run: gh api /rate_limit | jq
- uses: actions/download-artifact@634f93cb2916e3fdff6788551b99b062d0335ce0 # v5.0.0
if: steps.targetRunId.outputs.targetRunId
with:
run-id: ${{ steps.targetRunId.outputs.targetRunId }}
name: merged-${{ matrix.system }}
path: target
github-token: ${{ github.token }}
merge-multiple: true
- name: Compare outpaths against the target branch
if: steps.targetRunId.outputs.targetRunId
env:
MATRIX_SYSTEM: ${{ matrix.system }}
run: |
nix-build untrusted/ci --arg nixpkgs ./pinned -A eval.diff \
--arg beforeDir ./target \
--arg afterDir "$(readlink ./merged)" \
--argstr evalSystem "$MATRIX_SYSTEM" \
--out-link diff
- name: Upload outpaths diff and stats
if: steps.targetRunId.outputs.targetRunId
uses: actions/upload-artifact@ea165f8d65b6e75b540449e92b4886f43607fa02 # v4.6.2
with:
name: ${{ matrix.version && format('{0}-', matrix.version) || '' }}diff-${{ matrix.system }}
path: diff/*
compare:
runs-on: ubuntu-24.04-arm
needs: [eval]
if: needs.eval.outputs.targetRunId
permissions:
statuses: write
timeout-minutes: 5
steps:
- uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8 # v5.0.0
with:
sparse-checkout: .github/actions
- name: Check out the PR at the target commit
uses: ./.github/actions/get-merge-commit
with:
targetSha: ${{ inputs.targetSha }}
target-as-trusted: true
pinnedFrom: trusted
- name: Download output paths and eval stats for all systems
uses: actions/download-artifact@634f93cb2916e3fdff6788551b99b062d0335ce0 # v5.0.0
with:
pattern: diff-*
path: diff
merge-multiple: true
- name: Install Nix
uses: cachix/install-nix-action@fc6e360bedc9ee72d75e701397f0bb30dce77568 # v31
- name: Combine all output paths and eval stats
run: |
nix-build trusted/ci --arg nixpkgs ./pinned -A eval.combine \
--arg diffDir ./diff \
--out-link combined
- name: Compare against the target branch
env:
AUTHOR_ID: ${{ github.event.pull_request.user.id }}
run: |
git -C trusted fetch --depth 1 origin ${{ inputs.mergedSha }}
git -C trusted diff --name-only ${{ inputs.mergedSha }} \
| jq --raw-input --slurp 'split("\n")[:-1]' > touched-files.json
# Use the target branch to get accurate maintainer info
nix-build trusted/ci --arg nixpkgs ./pinned -A eval.compare \
--arg combinedDir "$(realpath ./combined)" \
--arg touchedFilesJson ./touched-files.json \
--argstr githubAuthorId "$AUTHOR_ID" \
--out-link comparison
cat comparison/step-summary.md >> "$GITHUB_STEP_SUMMARY"
- name: Upload the comparison results
uses: actions/upload-artifact@ea165f8d65b6e75b540449e92b4886f43607fa02 # v4.6.2
with:
name: comparison
path: comparison/*
- name: Add eval summary to commit statuses
if: ${{ github.event_name == 'pull_request_target' }}
uses: actions/github-script@60a0d83039c74a4aee543508d2ffcb1c3799cdea # v7.0.1
with:
script: |
const { readFile } = require('node:fs/promises')
const changed = JSON.parse(await readFile('comparison/changed-paths.json', 'utf-8'))
const description =
'Package: ' + [
`added ${changed.attrdiff.added.length}`,
`removed ${changed.attrdiff.removed.length}`,
`changed ${changed.attrdiff.changed.length}`
].join(', ') +
' — Rebuild: ' + [
`linux ${changed.rebuildCountByKernel.linux}`,
`darwin ${changed.rebuildCountByKernel.darwin}`
].join(', ')
const { serverUrl, repo, runId, payload } = context
const target_url =
`${serverUrl}/${repo.owner}/${repo.repo}/actions/runs/${runId}?pr=${payload.pull_request.number}`
await github.rest.repos.createCommitStatus({
...repo,
sha: payload.pull_request.head.sha,
context: 'Eval Summary',
state: 'success',
description,
target_url
})
# Creates a matrix of Eval performance for various versions and systems.
report:
runs-on: ubuntu-24.04-arm
needs: [versions, eval]
steps:
- name: Download output paths and eval stats for all versions
uses: actions/download-artifact@634f93cb2916e3fdff6788551b99b062d0335ce0 # v5.0.0
with:
pattern: "*-diff-*"
path: versions
- name: Add version comparison table to job summary
uses: actions/github-script@60a0d83039c74a4aee543508d2ffcb1c3799cdea # v7.0.1
env:
SYSTEMS: ${{ inputs.systems }}
VERSIONS: ${{ needs.versions.outputs.versions }}
with:
script: |
const { readFileSync } = require('node:fs')
const path = require('node:path')
const systems = JSON.parse(process.env.SYSTEMS)
const versions = JSON.parse(process.env.VERSIONS)
core.summary.addHeading('Lix/Nix version comparison')
core.summary.addTable(
[].concat(
[
[{ data: 'Version', header: true }].concat(
systems.map((system) => ({ data: system, header: true })),
),
],
versions.map((version) =>
[{ data: version }].concat(
systems.map((system) => {
try {
const artifact = path.join('versions', `${version}-diff-${system}`)
const time = Math.round(
parseFloat(
readFileSync(
path.join(artifact, 'after', system, 'total-time'),
'utf-8',
),
),
)
const diff = JSON.parse(
readFileSync(path.join(artifact, system, 'diff.json'), 'utf-8'),
)
const attrs = [].concat(
diff.added,
diff.removed,
diff.changed,
diff.rebuilds
).filter(attr =>
// Exceptions related to dev shells, which changed at some time between 2.18 and 2.24.
!attr.startsWith('tests.devShellTools.nixos.') &&
!attr.startsWith('tests.devShellTools.unstructuredDerivationInputEnv.')
)
if (attrs.length > 0) {
core.setFailed(
`${version} on ${system} has changed outpaths!\nNote: Please make sure to update ci/pinned.json separately from changes to other packages.`,
)
return { data: ':x:' }
}
return { data: time }
} catch {
core.warning(`${version} on ${system} did not produce artifact.`)
return { data: ':warning:' }
}
}),
),
),
),
)
core.summary.addRaw(
'\n*Evaluation time in seconds without downloading dependencies.*',
true,
)
core.summary.addRaw('\n*:warning: Job did not report a result.*', true)
core.summary.addRaw(
'\n*:x: Job produced different outpaths than the target branch.*',
true,
)
core.summary.write()
misc:
if: ${{ github.event_name != 'push' }}
runs-on: ubuntu-24.04-arm
timeout-minutes: 10
steps:
- uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8 # v5.0.0
with:
sparse-checkout: .github/actions
- name: Check if the PR can be merged and checkout the merge commit
uses: ./.github/actions/get-merge-commit
with:
merged-as-untrusted: true
- name: Install Nix
uses: cachix/install-nix-action@fc6e360bedc9ee72d75e701397f0bb30dce77568 # v31
- name: Ensure flake outputs on all systems still evaluate
run: nix flake check --all-systems --no-build ./untrusted
- name: Query nixpkgs with aliases enabled to check for basic syntax errors
run: |
time nix-env -I ./untrusted -f ./untrusted -qa '*' --option restrict-eval true --option allow-import-from-derivation false >/dev/null