Commit graph

1544 commits

Author SHA1 Message Date
Nicolas De Loof
fe17e31be0 test(e2e): lifecycle scenarios — cascade, profiles, healthcheck, recreate
Vocabulary grows with Recreated (the mirror of NotRecreated),
ServiceHealthy (healthcheck status observed from inspect) and a Within
action modifier bounding blocking commands like up --wait, replacing
the hand-rolled goroutine+timer of the legacy TestStartInterval.

- cascade: abort-on-container-exit/failure scenarios inline the model;
  exit-code propagation (111) is locked with ExitCode. The one-off
  variant stays legacy — it drives a long-running up in the background.
- profiles: the four scenarios observe which services materialize
  instead of grepping ps output; the dotenv variant declares its .env
  inline, making the activation mechanism visible.
- TestRecreateWithNoDeps now asserts both sides: the service is
  replaced (Recreated) and the healthy dependency is left alone.
- Fixtures start_interval/, profiles/compose.yaml+test-profile.env and
  recreate-no-deps.yaml are removed; cascade/ stays for the legacy
  one-off test.

Signed-off-by: Nicolas De Loof <nicolas.deloof@gmail.com>
2026-08-18 16:49:56 +02:00
Nicolas De Loof
6bc1b36516 test(e2e): expose, container_name, ipc and wait scenarios
- TestExposeRange keeps its regression link (#13378), the invariant
  being simply that up accepts a port range in expose.
- TestUpContainerNameConflict interpolates a project-scoped
  container_name instead of the fixture's global 'test', so parallel
  runs cannot collide on it.
- TestIPC demonstrates the CLI() escape hatch: the external container
  the 'container:' mode points at is created eagerly before the steps,
  so its id is available to the declarative checks. (Runs only in CI:
  ipc container: mode is rejected by Docker Desktop and dind daemons —
  'restricted host mount' — for the legacy test as well.)
- wait scenarios observe which services exited instead of trusting the
  command's silence; TestWaitOnInfinity stays legacy, the DSL has no
  notion of a still-running command. The wait fixture stays for it;
  ipc-test and container_name fixtures are removed.

Signed-off-by: Nicolas De Loof <nicolas.deloof@gmail.com>
2026-08-18 16:49:56 +02:00
Nicolas De Loof
a33cc37ad9 test(e2e): TestRemoveOrphans states its real intent — COMPOSE_REMOVE_ORPHANS from .env
The legacy test read as 'up removes leftover one-offs', but plain up
only warns about them: the removal it asserted came from the fixture's
.env declaring COMPOSE_REMOVE_ORPHANS=true (added by 'restore
setEnvWithDotEnv', d3d378b92) — an invariant the fixture kept invisible.
The scenario declares both files inline as a txtar archive, so the
mechanism under test is in plain sight, and observes the one-off's
removal from container state rather than ps output.

Signed-off-by: Nicolas De Loof <nicolas.deloof@gmail.com>
2026-08-18 16:49:56 +02:00
Nicolas De Loof
a725a8ae04 test(e2e): one-off vocabulary — service checks skip run containers
Snapshots now record whether a container is a one-off (run) container,
plus its start time. Service-facing checks (ServiceState, NotRecreated,
label checks) observe only long-lived containers; three checks join the
vocabulary for the one-off side: OneOffState, OneOffsUntouched (same
containers, same state, same start time — catches an unwanted restart,
which message matching never could) and OneOffsRemoved (errors if there
was nothing to remove, rather than pass vacuously).

TestStartStopWithOneOffs becomes the demonstrator: stop/start/restart
act on services and leave the one-off untouched, down --remove-orphans
takes it away.

Signed-off-by: Nicolas De Loof <nicolas.deloof@gmail.com>
2026-08-18 16:49:56 +02:00
Nicolas De Loof
695a2a759a test(e2e): start/stop scenarios observe container state, not CLI chatter
Migrate the state-friendly start/stop tests to the Scenario DSL:

- TestStartStop: stop halts containers in place, start brings the same
  ones back (NotRecreated); the ls checks match the project name exactly
  in the JSON output, immune to the prefix collisions the other
  start-stop project names would cause under parallel runs.
- TestStartStopWithDependencies: dependency start/stop propagation read
  from container state instead of 'Container X Started' messages.
- TestUpNoDeps replaces the 'Up no-deps links' subtest as its own
  scenario: up --no-deps must not create the linked dependency.
- TestStartSingleServiceAndDependency: create/start walk the dependency
  chain and nothing else — now asserting the unrelated services have no
  container at all, stronger than the legacy message matching.

The one-off and multi-file tests stay legacy: the vocabulary has no
one-off awareness yet and a scenario holds a single compose file.
The links and start-stop-deps fixtures become unused and are removed.

Signed-off-by: Nicolas De Loof <nicolas.deloof@gmail.com>
2026-08-18 16:49:56 +02:00
Nicolas De Loof
5baeadf929 test(e2e): no-deps scenarios — the donor container's absence is the observable
Migrate the volumes_from and network_mode:service donor tests to the
Scenario DSL, inlining the compose model and dropping the no-deps
fixtures. Each scenario removes the donor container behind compose's
back and expects up --no-deps to be rejected with the 'container
missing' error — the CLI's reported decision being the observable.

ServiceNotCreated joins the check vocabulary: it asserts a service has
no container at all, here proving the donor really is gone before the
failing step.

Signed-off-by: Nicolas De Loof <nicolas.deloof@gmail.com>
2026-08-18 16:49:56 +02:00
Nicolas De Loof
12288a7fff test(e2e): pause scenarios — paused is a container state, ExitCode locks error paths
Migrate the four pause tests to the Scenario DSL: the invariant is the
container state (paused/running), not an HTTP timeout, so the scenarios
observe state and drop the port-mapping helpers and the pause fixture.

ExitCode joins the check vocabulary for error-path steps, paired with
MayFail. TestPauseServiceDoesNotExist now locks the current behavior —
pause of an unknown service is rejected with 'no such service' — which
resolves the TODO the legacy test carried: it asserted exit 0 while
wishing for an error, and the CLI has since started validating selected
services.

Signed-off-by: Nicolas De Loof <nicolas.deloof@gmail.com>
2026-08-18 16:49:56 +02:00
Nicolas De Loof
b2e3deb18e test(e2e): testdata/<TestName>/ anchors each scenario's project files
Each scenario resolves its project files from testdata/<TestName>/ by
convention: standalone compose files in their native format, directly
runnable with docker compose -f, copied to a temporary directory so the
committed files are never mutated. Subtests map to nested directories.

Ownership is strictly one test per directory: a check fails the suite on
any testdata directory no test owns, which is what keeps this from
degrading into a catch-all fixtures directory — no sharing, no
leftovers after a rename.

The inline Compose() declaration is removed; the demonstrator scenarios
move their models to testdata, including the multi-file build context
of TestUpBuildUnchangedContext (compose.yaml + Dockerfile + marker).

Signed-off-by: Nicolas De Loof <nicolas.deloof@gmail.com>
2026-08-18 16:49:56 +02:00
Nicolas De Loof
fa9b31c2fa test(e2e): checks.go as the reviewed vocabulary, SCENARIO.md codifies the contract
Checks are the shared vocabulary between scenarios: move them to a
dedicated checks.go so the whole vocabulary is read — and reviewed — as
one file, independently of the tests using it.

SCENARIO.md codifies the rules that so far only lived in the PR
description: one intent = one invariant, step names as behavior
sentences, inline compose model, state-based checks first with
OutputContains as an explicit last resort, new checks generic and named
after their observable — plus how to exploit the failure artifacts and
E2E_KEEP_FAILED. AGENTS.md points new e2e work at it.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Signed-off-by: Nicolas De Loof <nicolas.deloof@gmail.com>
2026-08-18 16:49:56 +02:00
Nicolas De Loof
9ff4693312 test(e2e): dump failure artifacts, E2E_KEEP_FAILED keeps the project alive
On failure, a scenario now writes the untruncated diagnostic material to a
stable per-project directory (compose.yaml, each step's full command and
output, project containers, engine events, full container logs, per-step
state snapshots), announced at the top of the failure report. The inline
report stays truncated for readability; the artifacts let a failure be
diagnosed without re-running the scenario.

E2E_KEEP_FAILED=1 skips the teardown of a failed scenario so the live
project state remains inspectable.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Signed-off-by: Nicolas De Loof <nicolas.deloof@gmail.com>
2026-08-18 16:49:56 +02:00
Nicolas De Loof
a443a6589d test(e2e): Eventually check, TestRestart as polling demonstrator
Add Eventually(check, timeout) to the scenario layer: it retries a
state-based check, re-observing the project containers between
attempts. ServiceState(service, state) joins the check library, backed
by a State field in the per-step snapshots.

TestRestart is converted as the demonstrator: it replaces a
WaitForCmdResult polling `ps --format json` for a JSON fragment and a
bare time.Sleep with two declarative expectations, and now also locks
the real invariant — restart must reuse the same container — via
NotRecreated, which the lock-file trick of the original fixture was
proving only indirectly.

Signed-off-by: Nicolas De Loof <nicolas.deloof@gmail.com>
2026-08-18 16:49:56 +02:00
Nicolas De Loof
0632f222f6 test(e2e): declarative Scenario layer, image-identity tests as demonstrator
Introduce a small declarative layer over the e2e framework so a test
reads as its intent: an inline compose.yaml, a sequence of steps
(command + expected observables), and nothing else. The framework owns
project naming (derived from the test name), cleanup, and failure
reporting: on the first unmet expectation the test fails with a step
transcript, the failing command output, the project's container state,
the engine events since the scenario started and the container logs.

Checks observe real state (container IDs across steps, labels, image
manifest platform) instead of grepping the progress writer output,
which stays available as an explicit last-resort check.

The e2e tests introduced by #14041 are converted as a demonstrator;
their dedicated fixture files are inlined in the scenarios and removed.

Signed-off-by: Nicolas De Loof <nicolas.deloof@gmail.com>
2026-08-18 16:49:56 +02:00
Nicolas De Loof
02bc297519 refactor: last gocognit candidates — bridge resources and TestViz
LoadAdditionalResources (cognitive complexity 31) moves its
per-service body to loadServiceImageResources, with the local-vs-pull
inspection decision in inspectServiceImage. TestViz (35) factors the
shared graph assertions into assertVizGraphNodes and
assertVizDependencyEdges (t.Helper), merging the allowed/forbidden
edge bookkeeping into one equivalent loop.

This removes the last two of the 19 //nolint:gocognit suppressions
introduced by the gocyclo->gocognit migration: the FIXME debt is fully
paid, the linter now runs suppression-free.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Signed-off-by: Nicolas De Loof <nicolas.deloof@gmail.com>
2026-08-18 13:37:48 +02:00
Nicolas De Loof
c640808709 refactor: extract endpoint and volume translation decisions in create
createEndpointSettings (cognitive complexity 31) moves address parsing
to parseEndpointIPAM and the interface_name/driver_opts merge to
endpointDriverOpts. buildContainerVolumes (34) gets one decision
helper per mount type: bindStringForMount, volumeBindString,
checkImageMountSupported. A dead branch in the bind case (source was
reassigned to the same value) is simplified away — the surrounding
behavior is unchanged.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Signed-off-by: Nicolas De Loof <nicolas.deloof@gmail.com>
2026-08-18 13:37:48 +02:00
Nicolas De Loof
10c058a57b refactor: split publish push path and sensitive-data scanning
publish (cognitive complexity 34) moves its non-dry-run push path to
pushComposeArtifact, with the application image index construction in
pushApplicationIndex. checkForSensitiveData (34) splits per source
kind: scanEnvFiles (per-service env files with the required/missing
rules) and scanFiles (file-backed configs and secrets, deduplicating
two identical loops).

No behavior change: same events, same errors, same findings.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Signed-off-by: Nicolas De Loof <nicolas.deloof@gmail.com>
2026-08-18 13:37:48 +02:00
Nicolas De Loof
1d1b94f0c0 refactor: split doBuildImage context preparation stages
doBuildImage (cognitive complexity 43) mixed builder-capability
guards, context-type resolution (with in-scope defers), tar archiving
and credential conversion. Each becomes a function; the context
resolution returns a classicBuildContext carrying a cleanup so the
defers run at the same point they used to.

No behavior change: same errors, same context handling.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Signed-off-by: Nicolas De Loof <nicolas.deloof@gmail.com>
2026-08-18 13:37:48 +02:00
Nicolas De Loof
33c6435825 refactor: flatten Ps, restart and Logs orchestration
Three command paths with the same shape — a setup phase then per-
container goroutines with inlined bodies:

- Ps (44): the goroutine body becomes containerSummary, with pure
  helpers for publishers, health/exit-code, mounts and networks.
- restart (42): project resolution moves to prepareRestartProject,
  the per-container body to restartContainer. Also fixes a latent
  data race: hooks and restart assigned their error to the function's
  outer err from concurrent goroutines; now local.
- Logs (31): container selection moves to selectLogsContainers, the
  streaming body to logContainer, and the follow-mode listener to
  followStartedContainersLogs.

No behavior change: same events, same warnings, same results.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Signed-off-by: Nicolas De Loof <nicolas.deloof@gmail.com>
2026-08-18 13:37:48 +02:00
Nicolas De Loof
d6e26f427f refactor: split ociRemoteLoader.Load pull and index resolution
Load (cognitive complexity 48) mixed the enable/offline guards with
the artifact pull, the cache check and the image-index indirection.
The pull path moves to pullComposeArtifact and the index-to-manifest
resolution to resolveComposeManifest. Load reads as: guards, cache
lookup, pull on miss.

No behavior change, including the last-matching-manifest-wins loop
and the silent cache hit on a non-NotExist stat error.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Signed-off-by: Nicolas De Loof <nicolas.deloof@gmail.com>
2026-08-18 13:37:48 +02:00
Nicolas De Loof
68ab763674 refactor: flatten watch trigger preparation and initial sync walk
watch (cognitive complexity 52) inlined three passes over each
service's triggers; they become prepareRebuildTriggers (validation +
always-build marking), watchTriggerPaths (paths to monitor, bind-mount
skips, initial sync) and initialSyncRequested (the DEPRECATED
x-initialSync fallback). initialSyncFiles (34) delegates its WalkDir
closure to initialSyncDirectory, keeping the Info() call order so
error behavior on unreadable entries is unchanged.

No behavior change: same validations, same warnings, same collected
paths and sync mappings.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Signed-off-by: Nicolas De Loof <nicolas.deloof@gmail.com>
2026-08-18 13:37:48 +02:00
Nicolas De Loof
f46228a609 refactor: materialize compose pull scheduling as an imagePuller
pull (cognitive complexity 56) interleaved two scheduling loops
(service images, pre_start hook images) sharing dedup state, failure
slots and the must-build fallback list. That state becomes an explicit
imagePuller; the loops become pullServiceImages and pullHookImages,
the per-service goroutine body becomes runServicePull, and the four
copies of the Skipped event literal collapse into eventSkippedPull.

Also fixes a latent data race: mustBuild was appended from concurrent
pull goroutines without synchronization; it is now guarded by a mutex.

No behavior change otherwise: same skip events, same fail-fast rules,
same error report.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Signed-off-by: Nicolas De Loof <nicolas.deloof@gmail.com>
2026-08-18 13:37:48 +02:00
Nicolas De Loof
0d81e99391 refactor: materialize interactive up as an upSession
Up (cognitive complexity 79) interleaved service creation with an
implicit session object: an errgroup, an error collector, exit status
and termination flag, all shared between six closures. That session is
now an explicit upSession type; each concern becomes a method:
runEventLoop (signals/keyboard/cancellation), stopApplication and
killApplication (the first/second-interrupt actions), stopOnFirstExit
(--abort-on-container-exit cascade), captureExitCodeFrom
(--exit-code-from), followStartedContainers + streamContainerLogs
(attach to (re)started containers). Keyboard menu setup moves to
setupNavigationMenu, closing the keyboard on the desktop-detection
error path exactly where Up's deferred Close used to.

Also fixes a latent data race: the graceful-stop and cascade-stop
goroutines assigned their error to Up's outer err variable (racing
with the main goroutine) before handing it to the collector; the
assignment was redundant and is now local.

No behavior change otherwise: same listeners in the same order, same
cancellation points, same error report.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Signed-off-by: Nicolas De Loof <nicolas.deloof@gmail.com>
2026-08-18 13:37:48 +02:00
Nicolas De Loof
ef54fcaa41 refactor: one handler per container event in monitor.Start, add tests
monitor.Start (cognitive complexity 77) inlined the handling of all
four container actions in the event loop. Each action moves to an
onContainerX handler, with the recurring bits named: watched() (is
this service's container ours to track), notify() (broadcast to
listeners), initialContainers() (seed the tracking set). The loop
drops to complexity 19 and reads as: seed, subscribe, dispatch until
no containers remain. The tracking sets stay explicit parameters so
data flow remains visible.

The monitor had no unit test despite driving up/logs termination; the
refactor is locked by a new suite covering the full lifecycle
(created/recreated/started/restarted/exited, default-name trimming,
exit codes), the restart detection through both engine states
(Restarting, and Running per moby/moby#45538), the already-removed
container on die, service filtering, the no-container fast path,
events stream errors, context cancellation, and exit-code parse
errors.

No behavior change: same events, same ordering, same termination
conditions.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Signed-off-by: Nicolas De Loof <nicolas.deloof@gmail.com>
2026-08-18 13:37:48 +02:00
Nicolas De Loof
bae663129e refactor: one function per depends_on condition in waitDependencies
waitDependencies (cognitive complexity 76) inlined the polling loop
and the per-condition logic for all three depends_on conditions, each
with its own optional-dependency degradation. The polling loop moves
to waitDependency, and each condition check becomes a function
reporting (done, err) — (false, nil) means keep polling. The driver
drops to complexity 11 and its FIXME suppression is gone.

No behavior change: same events, same log messages, same errors.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Signed-off-by: Nicolas De Loof <nicolas.deloof@gmail.com>
2026-08-18 13:37:48 +02:00
Nicolas De Loof
bd7f14c186 fix: bake target names collision for services differing by '.' vs '_'
The dedup loop in bakeTargetNames checked whether the service already
had a name — never true on first pass — instead of whether the name
was already taken, so services like 'a.b' and 'a_b' silently mapped to
the same bake target and one overwrote the other in the bake file.
Allocate names in sorted service order (deterministic) and append '_'
until unique.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Signed-off-by: Nicolas De Loof <nicolas.deloof@gmail.com>
2026-08-18 11:41:00 +02:00
Guillaume Lours
81bb4afed5 fix(events): validate service names before subscribing to events
docker compose events <name> was passing service arguments straight to
the event filter without checking they exist in the project, so a typo
would produce an event stream matching nothing and block forever with
no output and no error.

Switch to projectOrName (which validates args via ToProject) before
calling backend.Events, consistent with logs, kill, and restart.

Signed-off-by: Guillaume Lours <glours@users.noreply.github.com>
2026-08-18 10:48:59 +02:00
Guillaume Lours
8d763d9bff test(e2e): prepend provider bin dir to PATH in provider tests
When a locally-installed binary with the same name (e.g.
example-provider) sits in an earlier PATH directory than
bin/build/, exec.LookPath resolves the wrong binary.  The
old tests appended the build directory to the end of PATH,
so any pre-existing system binary shadowed the test one.
Prepending instead ensures the freshly built binary always
wins, regardless of what is installed on the developer's
machine.  CI is unaffected since no such system binary
exists there.

Signed-off-by: Guillaume Lours <glours@users.noreply.github.com>
2026-08-18 10:47:48 +02:00
Nicolas De Loof
6d2e1d99dd build: split doBuildBake at its responsibility boundaries
doBuildBake mixed five concerns in a 300-line body (cognitive
complexity 88): progress display setup, translation of the project
into a bake file definition, temp metadata file allocation, bake
command construction, stderr rawjson streaming, and result
collection. Each now lives in its own function; the driver reads as
the sequence of those stages (cognitive complexity 17), and the
FIXME suppression is gone.

The bake variable was named to avoid shadowing the docker/cli
'build' package. No behavior change: prepareBakeBuild emits the same
config, and the stderr loop keeps the decoder-per-line semantics.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Signed-off-by: Nicolas De Loof <nicolas.deloof@gmail.com>
2026-08-18 10:42:37 +02:00
Nicolas De Loof
2340121f0e lint: replace gocyclo with gocognit
Cognitive complexity (gocognit, threshold 30 — the linter default) fits
this codebase better than raw cyclomatic complexity: it barely charges
guard clauses and early returns, and penalizes nesting instead. As a
result 8 of the 21 //nolint:gocyclo suppressions become unnecessary,
while deeply nested functions gocyclo never flagged are now caught.

The 19 functions still above the threshold keep a suppression, each
marked FIXME to complete the migration by restructuring them.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Signed-off-by: Nicolas De Loof <nicolas.deloof@gmail.com>
2026-08-18 10:42:37 +02:00
Nick Sieger
870908cc8f feat(ps): add ENGINE column gated on label presence
Show an ENGINE column in `compose ps` default table output when the
com.docker.compose.engine label is present. Expose {{.Engine}} for custom
--format templates.

Rename api.EngineLabel to api.ContainerEngineLabel to avoid the
name collision with desktop.EngineLabel (com.docker.desktop.address).

Signed-off-by: Nick Sieger <nick@nicksieger.com>
2026-08-17 10:34:46 +02:00
Nicolas De Loof
1e80a0906f chore: inline needless single-use helpers
Inline small functions that were extracted from their single call site
without a real boundary to justify it — no reuse, no dedicated test, no
responsibility of their own — so each caller now tells its whole story
top-down:

- getExecTarget, attachContainer, logContainer: one-line trampolines to
  getSpecifiedContainer / doAttachContainer / doLogContainer, which
  other call sites already use directly
- removeImage: single-statement wrapper, unlike its removeVolume
  sibling which has actual logic
- checkSelectedServices: named like a validation, actually a filter;
  the subtle rule (an unknown service is only an error with an explicit
  compose file) now reads where options.Services is rewritten
- prepareLabels: mutated the map it received while looking pure at the
  call site; the label writes are now visible in getCreateConfigs
- setDefaultTarget: mutation-by-pointer of the loop copy, now visible
  in the loop of injectFileReferences
- buildVolumeOptions/buildTmpfsOptions/buildImageOptions: nil-guard +
  field copies; the buildMountOptions switch now shows side by side
  what each mount type propagates (buildBindOption keeps real logic and
  keeps buildMountOptions under the gocyclo limit)
- displayDryRunBuildEvent: was longer than its only caller
- hasMore: read like a predicate, was a one-line spinner restart
- escapeDollarSign: wrapped a single bytes.ReplaceAll
- extractEnvCLIDefined: replaced by the canonical compose-go helper
  types.NewMappingWithEquals().ToMapping(), as run.go already does
- isPullPolicyValid: rebuilt the valid-values slice on every call; now
  a package-level list checked at the call site
- viewFromStackList: projection now sits next to the render closure
  that consumes exactly its three fields

No behavior change; single gocyclo threshold untouched.

Signed-off-by: Nicolas De Loof <nicolas.deloof@gmail.com>
2026-08-14 22:49:53 +02:00
Nicolas De Loof
df094b036c fix(images): tolerate containers whose image record is gone
A running container may reference an image record that no longer
exists: under the containerd image store, `up --build` with identical
content moves the tag to a new index digest (provenance attestation
churn) and the daemon drops the old index the container was created
from — without compose recreating the container, by design (#13636).
`docker rmi -f` of a running container's image produces the same state.

`compose images` used to fail the whole listing on the resulting
NotFound. Degrade to what the container itself knows (image ID, and
repository/tag when the reference is not a raw ID) instead of failing.

Fixes #14014

Signed-off-by: Nicolas De Loof <nicolas.deloof@gmail.com>
2026-08-14 15:20:37 +02:00
Nicolas De Loof
bbcd8108b2 test: detect containerd image store directly in multi-arch error test
The 'builder does not support multi-arch' subtest expects the docker
driver to reject multi-platform builds, and skipped itself by checking
that buildx lists both linux/amd64 and linux/arm64. That heuristic
misses containerd-store hosts without binfmt emulation (plain CI
runners): only native platforms are listed even though the driver
happily cross-builds, so the build succeeds and the test fails with
'ExitCode was 0 expected 1'. Skip on the containerd image store
directly, detected from the daemon's DriverStatus.

Signed-off-by: Nicolas De Loof <nicolas.deloof@gmail.com>
2026-08-13 12:03:29 +02:00
Nicolas De Loof
5ec38bbd5c test: image volume from an already-local image
Regression test for #14005, pulled-image scenario: with the source image
of a type=image volume already present in the local store, compose used
to rewrite the mount source to a digest the daemon can't resolve as a
mount source under the containerd image store (No such image).
TestImageVolume only exercises this path when a previous test happens to
have left the image locally; pre-pulling makes it deterministic. Also
asserts a second unchanged up doesn't recreate the service.

The test fails on main under the containerd image store and passes with
the fix from #14011.

Signed-off-by: Nicolas De Loof <nicolas.deloof@gmail.com>
2026-08-13 12:03:29 +02:00
Nicolas De Loof
7ff6748564 test: local multi-platform image must satisfy the missing pull policy
Regression test for #14007: with the containerd image store, a local
multi-platform image holding the requested non-native variant must be
used as-is by the default missing pull policy — compose used to inspect
the image without a platform, compare the host variant's platform fields
to the requested one, conclude the image is missing and try to pull the
(unpublished) tag.

The test fails on main and passes with #14011, whose platform check
resolves the requested platform against the locally available manifests.

Signed-off-by: Nicolas De Loof <nicolas.deloof@gmail.com>
2026-08-13 12:03:29 +02:00
Nicolas De Loof
b32c5b9a94 docs+test: disambiguate content digests from distribution digests
Compose manipulates two digest kinds that must never be conflated:
the platform-specific content digest (localContentDigest) is ONLY an
identity to compare a running container with a fresh build/pull, while
pinning image references in a reproducible compose model (publish /
config --resolve-image-digests) must keep the registry descriptor
digest — the multi-platform index — through ImageDigestResolver, or
the published file would be bound to the platform of whoever resolved
it. Document the distinction on both producers and lock it with a test
that fails if the resolver ever goes through a local image inspect.

Signed-off-by: Nicolas De Loof <nicolas.deloof@gmail.com>
2026-08-13 12:03:29 +02:00
Nicolas De Loof
a2e7fe6590 test: cover image-identity corner cases end to end
- dry-run up with a missing image must not hit the real daemon
- create twice with a non-native DOCKER_DEFAULT_PLATFORM is idempotent
- two services sharing an image with mixed platform pinning get their
  own platform's digest label, idempotently, whatever pull ordering
- explicit pull refreshes ahead of a daily refresh window

Signed-off-by: Nicolas De Loof <nicolas.deloof@gmail.com>
2026-08-13 12:03:29 +02:00
Nicolas De Loof
64cbcb2276 fix: address image-identity review findings
- pullRequiredImages resolves each distinct pulled image once, after
  all pulls completed, for the host default platform — the exact way
  getLocalImagesDigests resolves already-local images. Resolving from
  each pull's goroutine, for the pulled platform, let the recorded
  digest depend on pull completion order when several services pull
  the same tag for different platforms (last-writer-wins), and made
  the first up after a pull recreate containers. Platform-pinned
  services keep getting their own platform's digest from
  serviceImageDigest. In dry-run nothing was actually pulled, so the
  local inspect is skipped — it reached the real daemon through a
  DryRunClient dispatch that no longer matched and failed with 'No
  such image'.
- explicit 'compose pull' treats daily/weekly/every_N windows as due:
  it is the only way to force a refresh ahead of the window.
- 'compose pull' no longer silently skips services declaring both
  provider: and image:.
- pre_start hook images go through the same mustPull interpreter on the
  up path as on the pull path, so refresh windows apply consistently.
- matchLocalManifest falls back to the inspect's flat platform fields
  when the lone available manifest carries no ImageData (locally built
  images), instead of reporting the platform unsatisfied and pulling a
  possibly local-only image.
- scale resolves DOCKER_DEFAULT_PLATFORM without validating
  build.platforms: a conflict on a service that isn't being built must
  not abort the command. run needs nothing: createOptions.Apply
  resolves and validates platforms on its path already.
- resolveImageVolumes documents the accepted one-time-recreate and
  retag-race tradeoffs.

Signed-off-by: Nicolas De Loof <nicolas.deloof@gmail.com>
2026-08-13 12:03:29 +02:00
Guillaume Lours
d5e275dfb8 fix(pull): interpret pull_policy like up does
compose pull switched on the raw pull_policy string: daily/weekly/every_N
never matched a case and fell through to an unconditional re-pull, and
the hook-image loop was a second interpreter that ignored the refresh
window entirely. Delegate the decision to the exact interpreter the up
path uses (mustPull), with hook images routed through the same decision
(build mapped to missing — a hook image can't be built as a fallback).

Two deliberate differences with up are kept and documented in
shouldPullImage: a service without an explicit pull_policy is always
refreshed (skipping it would turn an explicit compose pull into a no-op
once images exist), and a present latest tag is still refreshed under
missing/if_not_present — the tag is expected to move, and triggering the
pull lets the daemon negotiate with the registry, a manifest check with
no download when the local image is already current.

User-visible change (changelog): compose pull now honors
daily/weekly/every_N refresh windows instead of always re-pulling.

Signed-off-by: Guillaume Lours <glours@users.noreply.github.com>
2026-08-13 12:03:29 +02:00
Guillaume Lours
a9d054fb17 fix(build): canonical content-digest producer, single image-label writer
Image identities recorded for staleness detection were produced by
several independent paths yielding different digest kinds for the same
image: the platform check compared flat inspect fields while the digest
picked a manifest with the host matcher (never the service's pinned
platform), a wrong-platform summary just discarded still leaked its
digest into the label, bake substituted digests host-side in batch, and
the classic builder recorded the raw build-stream ID as-is. Any of those
mismatches makes the next up see a phantom image change and recreate
containers.

Converge every producer on one selection (matchLocalManifest /
localContentDigest): the shared parallel inspect feeds both the digest
and the platform check, platform-pinned services resolve THEIR platform's
manifest in-process (no extra API call), and both builders route through
canonicalBuiltDigest. Registry-only builds (push-only, multi-platform
without load) keep the builder-reported digest — volatile but honest, an
actual rebuild is still detected, where a stable placeholder would hide
real image changes. ensureImagesExists' final loop becomes the label's
single writer so the pinned resolution can't be overwritten, superseded
only by pull/build results already platform-resolved by their producers —
and when a pull or build refreshed the shared entry mid-run (a digest
resolved for whichever service triggered it), a service pinned on another
platform re-resolves its own with one extra inspect, in that case only.

With every producer converged, TestUpIdempotentContainerdStore is
un-skipped here.

Signed-off-by: Guillaume Lours <glours@users.noreply.github.com>
2026-08-13 12:03:29 +02:00
Max Malm
599e12216d Use content digest for pulled service images
pullServiceImage returned the pulled image's raw inspect ID, while
getImageSummaries resolves already-local images through contentDigest
(the platform image-manifest digest). Both values feed the
com.docker.compose.image label that mustRecreate compares to detect
image changes, so the two paths disagreeing made the first 'up' after
the pulling 'up' see a phantom image change and recreate every
container once, with no change anywhere. Under the containerd image
store a tag@digest reference triggers this: the raw inspect ID is the
index digest, while contentDigest picks the platform manifest digest.

Resolve the pulled image through the same manifests-aware inspect and
contentDigest call getImageSummaries uses, so both sides of the
staleness comparison speak the same scheme.

Verified against a fresh docker:dind (29.7.0, containerd store) with a
tag@digest service: unpatched v5.4.0 recreates the container on the
second 'up'; with this fix the container survives repeated 'up' runs.
Existing behavior is preserved for engines without manifest support
(contentDigest falls back to the plain ID).

(Squashed with the follow-up lint cleanup from the same PR.)

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Signed-off-by: Max Malm <benjick@dumfan.net>
Signed-off-by: Guillaume Lours <glours@users.noreply.github.com>
2026-08-13 12:03:29 +02:00
Guillaume Lours
d98c41ba07 ci(e2e): run e2e against the containerd image store
The e2e suite only ran on graphdriver daemons, where the different kinds
of image digests coincide — the blind spot that let #13636, #13998 and
#14005 through. Add one matrix entry enabling the containerd image store,
plus TestUpIdempotentContainerdStore: two consecutive `up` runs with no
change must not recreate any container. The test is red on this
configuration (the com.docker.compose.image label is written from the
index digest on the pulling run, then compared against the per-platform
manifest digest on the next run) and skipped until the next commit
resolves the pull-path digest.

Signed-off-by: Guillaume Lours <glours@users.noreply.github.com>
2026-08-13 12:03:29 +02:00
Ricardo Branco
1615bf352f fix(build): resolve image volumes to a mountable name, not a manifest digest
With the containerd image store, ImageSummary.ID holds the digest of the
platform-specific manifest so ServiceHash stays stable across attested
rebuilds (see contentDigest). resolveImageVolumes reused that same value
as the `type: image` mount Source, but the daemon only resolves a mount
Source by name/tag or top-level image ID, not by manifest digest — so
`compose up` failed with "No such image" whenever the volume's source
image was already present locally (always for a built image; on a
second run for a pulled one).

Keep Source as the resolved image name, and track the digest separately
via a new com.docker.compose.image-volume-digest label so mustRecreate
can still detect a rebuilt/updated source image independently of Source.

Fixes #14005

Signed-off-by: Ricardo Branco <rbranco@suse.de>
Signed-off-by: Guillaume Lours <glours@users.noreply.github.com>
2026-08-13 12:03:29 +02:00
Branislav Osif
fc860cbd1f fix: ignore one-off container events in up monitor
`up --abort-on-container-exit` tore down the project when a one-off
container created by `docker compose run` exited. The container list
built in monitor.Start already excludes one-off containers via
oneOffFilter(false), but the event subscription did not, and the only
guard on incoming events is the service label, which a one-off
container inherits from the service it was run from.

Regression since v2.39.0 (last unaffected release is v2.38.2), and a
recurrence of docker/compose-cli#1955, fixed there by
docker/compose-cli#1987.

Signed-off-by: Branislav Osif <brano@osif.digital>
2026-08-12 16:02:25 +02:00
Endika Iglesias
832673c8e4 test(watch): cover the permission branches as root
CI runs the tests as root, where the unreadable-directory test is always
skipped. Extract the WalkDir callbacks and inject the watch registration
so a synthetic permission error can drive those branches instead.

Signed-off-by: Endika Iglesias <endika2@gmail.com>
2026-08-11 14:10:40 +02:00
Endika Iglesias
fb543842d8 fix(watch): skip unreadable directories instead of failing the watch
Signed-off-by: Endika Iglesias <endika2@gmail.com>
2026-08-11 14:10:40 +02:00
Guillaume Lours
7bc2630b65 fix(config): resolve service environment when computing --hash
Containers are created from a project whose service environment is
resolved (env_file merged into environment), but since v2.22.0 `config
--hash` skipped that resolution, so hashes diverged from the
com.docker.compose.config-hash label for services using env_file.

Resolve the environment of the hashed services only, so a broken
env_file or platforms on an unrelated service still doesn't prevent
hashing, and honor --no-env-resolution as an escape hatch.

Fixes #14001

Signed-off-by: Guillaume Lours <glours@users.noreply.github.com>
2026-08-11 08:37:13 +02:00
Guillaume Lours
5573bd2601 fix(watch): stop pruning every dangling image of the project
pruneDanglingImagesOnRebuild probed the name-keyed built-images map with
an image ID, so the spare-check never matched and every dangling image
carrying the project label was removed on each rebuild with --prune —
not just the superseded ones. Match dangling IDs against the map's
values (the freshly built image IDs) instead.

Signed-off-by: Guillaume Lours <glours@users.noreply.github.com>
2026-08-11 08:05:14 +02:00
Ricardo Branco
668d81d4ec test: Set stop_signal to SIGTERM in nginx-based services
The official nginx images set STOPSIGNAL to SIGQUIT which dumps core.
Set it to SIGTERM to avoid dumping core on e2e tests when containers
running "sleep infinity" are stopped.

Signed-off-by: Ricardo Branco <rbranco@suse.de>
2026-08-10 12:42:02 +02:00
Eric Wyles
d95ff5b3d5 Reuse bridge image version in e2e tests
Signed-off-by: Eric Wyles <23637493+ericwyles@users.noreply.github.com>
2026-08-06 16:00:08 +02:00
Eric Wyles
1b76c732f2 Update pkg/bridge/convert.go
Co-authored-by: Guillaume Lours <705411+glours@users.noreply.github.com>
Signed-off-by: Eric Wyles <23637493+ericwyles@users.noreply.github.com>
2026-08-06 16:00:08 +02:00