wslc: add --digests to image list for docker parity

[ggarzia-MSFT]

Summary of the Pull Request

Adds --digests to wslc image list (and its image ls / images spellings), matching docker images --digests. The DIGEST column was previously hardcoded to <none> in json output and absent from the table.

PR Checklist

• Closes: Link to issue #xxx
• Communication: I've discussed this with core contributors already. If work hasn't been agreed, this work might be rejected
• Tests: Added/updated if needed and all pass
• Localization: All end user facing strings can be localized
• Dev docs: Added/updated if needed
• Documentation updated: If checked, please file a pull request on our docs repo and link it here: #xxx

Detailed Description of the Pull Request / Additional comments

WSLCListImagesFlagsDigests was already defined and already honored end to end — WSLCSession::ListImages forwards it to Docker().ListImages(all, digests, filters) and populates WSLCImageInformation::Digest. The CLI never set the flag and never read the field, so ToImageOutput hardcoded Digest to <none>.

• New ArgType::Digests (--digests, no short alias, matching docker where only listing commands' --filter carries -f).
• models::ImageInformation gains a Digest field, and ImageService::List takes a defaulted bool digests that sets the flag and populates it.
• Table output adds a DIGEST column between TAG and IMAGE ID only when --digests is passed, matching docker's defaultImageTableFormatWithDigest.
• The service reports a repo digest (repo@sha256:...); docker's DIGEST column shows only the digest. New DigestFromRepoDigest helper in ImageModel.h performs that reduction and is unit tested directly.
• The digest is only populated when the flag is set. This matters for json: docker gates on needDigest(ctx), which is ctx.Digest || format.Contains("{{.Digest}}"), and --format json satisfies neither. So image list --format json still reports <none>, and only --digests --format json reports a real digest. An earlier revision of this change leaked the digest into plain json output; the tests below lock the gated behavior in.
• Column configs were hoisted into named constexpr values so the 5- and 6-column tables cannot drift.

Validation Steps Performed

All new and existing tests in the touched areas were run against a locally deployed build. 25/25 pass.

Unit tests (WSLCCLIImageDigestUnitTests, new file, 3 tests):

• DigestFromRepoDigest_StripsRepositoryPrefix — plain, fully qualified, and registry-with-port repositories all reduce to the bare sha256: digest. The port case guards the split from being confused by a colon.
• DigestFromRepoDigest_PassesThroughBareValues — empty, already-bare, and <none> values are untouched.
• DigestFromRepoDigest_SplitsOnFirstSeparator — splitting happens once, so a digest is never truncated further.

Unit test (WSLCCLICommandUnitTests):

• ImageListCommand_HasDigestsArgument — both the image list and root-scoped images constructions register --digests as an optional flag with no short alias.

E2E tests (WSLCE2EImageListTests, 22 total, 5 new):

• Digests_AddsColumnBetweenTagAndImageId — asserts DIGEST sits after TAG and before IMAGE ID, and that the default listing does not gain the column.
• Digests_ReportsDigestOnlyWhenRequested — json reports <none> without the flag; anything reported with the flag must be a bare sha256: digest with no repo@ prefix.
• Digests_TableMatchesJson — every digest in json output appears in the table's DIGEST column, so the two renderings cannot drift.
• Digests_QuietStillOutputsIdsOnly — --quiet wins over --digests, emitting bare ids with no header, as docker does.
• Digests_ListedInHelp — --digests and its localized description appear in image list --help.

Note on the test images: they are provisioned via image load from a tarball, so they carry no repo digest and correctly report <none> even with --digests — the same as docker for a tar-loaded image. The digest-reduction logic is therefore covered by the unit tests rather than by asserting a live registry digest in e2e, which would make the suite network dependent.

[Copilot]

Pull request overview

This pull request adds --digests support to wslc image list to match Docker parity, wiring the flag through CLI argument parsing into the image listing service and updating both table and JSON render paths to only surface digests when explicitly requested.

Changes:

• Added a new --digests flag (no short alias) for image list / image ls / images, and threaded it through ImageService::List.
• Implemented digest formatting by reducing service “repo@sha256:…” values to Docker’s “sha256:…” form via DigestFromRepoDigest, and conditionally added a DIGEST column in table output.
• Added unit + E2E tests to lock the gating behavior (digests only appear when --digests is passed) and validate output parity between JSON and table formats.

Reviewed changes

Copilot reviewed 10 out of 10 changed files in this pull request and generated 1 comment.

Show a summary per file

File	Description
test/windows/wslc/WSLCCLIImageDigestUnitTests.cpp	New unit tests for DigestFromRepoDigest behavior.
test/windows/wslc/WSLCCLICommandUnitTests.cpp	Verifies image list/images register --digests with no short alias.
test/windows/wslc/e2e/WSLCE2EImageListTests.cpp	Adds E2E coverage for DIGEST column placement, gating, help output, and quiet behavior.
src/windows/wslc/tasks/ImageTasks.cpp	Threads --digests into listing and conditionally adds DIGEST column + JSON field population.
src/windows/wslc/services/ImageService.h	Extends ImageService::List signature with a digests parameter (defaulted).
src/windows/wslc/services/ImageService.cpp	Sets the WSLC digest flag and populates ImageInformation::Digest only when requested.
src/windows/wslc/services/ImageModel.h	Adds Digest field to the image model and introduces DigestFromRepoDigest.
src/windows/wslc/commands/ImageListCommand.cpp	Registers the new ArgType::Digests argument for the list command.
src/windows/wslc/arguments/ArgumentDefinitions.h	Defines the Digests argument metadata and localization key mapping.
localization/strings/en-US/Resources.resw	Adds localized description string for --digests.

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[Copilot]

Pull request overview

Copilot reviewed 10 out of 10 changed files in this pull request and generated 1 comment.

Suppressed comments (1)

localization/strings/en-US/Resources.resw:3023

• Most CLI argument descriptions in Resources.resw do not end with a period; this new string does, which makes help output inconsistent.

    <value>Show image digests.</value>

[David Bennett (dkbennett)]

--quiet and --digests are valid options together and Docker CLI allows both. --quiet means only image ids are output, but --digests means that an image with multiple digests can appear multiple times in the output (same is true for tags)

Should add tests here which verify the --quiet and --digests combination (and also images with multiple tags should appear multiple times with --quiet).

This may be a pre-existing issue with quiet that could be addressed here.

[Copilot]

Pull request overview

Copilot reviewed 20 out of 20 changed files in this pull request and generated no new comments.

[Copilot]

Pull request overview

Copilot reviewed 12 out of 12 changed files in this pull request and generated 2 comments.

Suppressed comments (1)

localization/strings/en-US/Resources.resw:3024

• Most CLI argument descriptions in Resources.resw omit trailing punctuation (e.g., "Run container in detached mode"). For consistency, drop the trailing period from this new description.

  <data name="WSLCCLI_DigestsArgDescription" xml:space="preserve">
    <value>Show image digests.</value>
  </data>

[Copilot]

Pull request overview

Copilot reviewed 13 out of 13 changed files in this pull request and generated 1 comment.

Suppressed comments (2)

Previously missed (1) — in code that hasn't changed since the last review.

test/windows/wslc/e2e/WSLCE2EImageListTests.cpp:495

• This assertion is logically correct, but it reads as a double-negative ("IS_FALSE(find != npos)"). Using an explicit npos comparison makes failures easier to interpret and avoids precedence/clarity pitfalls in future edits.

        VERIFY_IS_FALSE(defaultResult.GetStdoutLines()[0].find(L"DIGEST") != std::wstring::npos);

src/windows/wslcsession/WSLCSession.cpp:1958

• When WSLCListImagesFlagsDigests is not set, this branch still emits a digest value (it->second.front()) if RepoDigests are present. Since the flag is meant to gate whether digest info is included at all, it would be safer to always emit an empty digest when digests==false (even if the backend unexpectedly provides RepoDigests).

            else if (!digests)
            {
                rows.push_back({&e, tag, it->second.front()});
            }

[Copilot]

Pull request overview

Copilot reviewed 13 out of 13 changed files in this pull request and generated 1 comment.

Suppressed comments (3)

Previously missed (2) — in code that hasn't changed since the last review.

test/windows/wslc/WSLCCLITableOutputUnitTests.cpp:274

• In this test, namePos/statusPos are used for arithmetic without first verifying the substrings were found. If either find() returns npos, the expected-position math can overflow and the assertion may become misleading.

        const auto namePos = dataLine.find(L"container-a");
        const auto statusPos = dataLine.find(L"running");
        const auto expected = namePos + wcslen(L"container-a") + TableOutput<3>::DefaultColumnPadding + wcslen(L"DIGEST") +
                              TableOutput<3>::DefaultColumnPadding;
        VERIFY_ARE_EQUAL(expected, statusPos);

test/windows/wslc/WSLCCLITableOutputUnitTests.cpp:306

• This assertion uses namePos in arithmetic without checking that "container-a" (and "running") were actually found. If either find() returns npos, the computed offset can overflow and hide formatting regressions.

        const auto& dataLine = cap.lines()[1];
        const auto namePos = dataLine.find(L"container-a");
        VERIFY_ARE_EQUAL(namePos + wcslen(L"container-a") + TableOutput<3>::DefaultColumnPadding, dataLine.find(L"running"));

localization/strings/en-US/Resources.resw:3024

• This new argument description includes a trailing period, which is inconsistent with nearby argument descriptions (e.g., "Run container in detached mode") and will show up in --help output.

  <data name="WSLCCLI_DigestsArgDescription" xml:space="preserve">
    <value>Show image digests.</value>
  </data>

[Copilot]

Pull request overview

Copilot reviewed 13 out of 13 changed files in this pull request and generated 1 comment.

Suppressed comments (1)

test/windows/wslc/e2e/WSLCE2EImageListTests.cpp:549

• This assertion only checks whether the digest text appears anywhere in the table output. That can pass even if the value appears in a different column (or in multiple places), so it doesn't actually verify the DIGEST column as the comment claims. Consider extracting the DIGEST column slice using the header offsets and comparing against that field.

        // Every digest reported by json output must appear in the table's DIGEST column, so the two
        // renderings cannot drift.
        for (const auto& image : ParseNdjsonOutputAs<ImageOutputInformation>(jsonResult))
        {
            const auto digest = wsl::shared::string::MultiByteToWide(image.Digest);
            VERIFY_IS_TRUE(
                tableResult.StdoutContainsSubstring(digest),
                WEX::Common::String().Format(L"'%ls' was missing from the table DIGEST column", digest.c_str()));
        }

[Copilot]

🟡 Changes recommended

Four unresolved moderate issues remain around digest association, bare digest handling, and deterministic coverage.

Get a fresh assessment by requesting another Copilot review.

Review details

Suppressed comments (3)

src/windows/wslcsession/WSLCSession.cpp:1933

• The repository Name preserves the daemon's spelling, so this key does not match equivalent references such as a tag ubuntu:latest and a digest docker.io/library/ubuntu@sha256:... (or index.docker.io/...). That leaves the tag row with <none> and emits a separate digest-only row instead of associating the digest with the tag; use the normalized repository identity consistently for this map and the tag lookup/set (for example, Repository.GetCanonical()).

            digestsByRepo[reference->Repository.Name].push_back(repoDigest);

src/windows/wslcsession/WSLCSession.cpp:1962

• The added end-to-end cases use tar-loaded images, whose RepoDigests are empty, so this per-digest row path is never exercised; the unit tests only cover stripping a repository prefix. A regression in daemon digest parsing, tag/repository association, or multiple-digest expansion would therefore pass. Add deterministic coverage with a fixture or mock response containing at least one valid repo digest (and preferably two).

                for (const auto& repoDigest : it->second)
                {
                    rows.push_back({&e, tag, repoDigest});
                }

test/windows/wslc/e2e/WSLCE2EImageListTests.cpp:558

• These E2E images are explicitly tar-loaded and have no repository digests, so this test never exercises the new RepoDigests loop that emits one row per digest. It only checks properties of the existing <none> rows; add deterministic service-level/fixture coverage with multiple repo digests before treating this as coverage for the new expansion behavior.

        auto result = RunWslc(L"image list --digests --format json");
        result.Verify({.Stderr = L"", .ExitCode = 0});

        const auto digestRows = ParseNdjsonOutputAs<ImageOutputInformation>(result);

• Files reviewed: 16/16 changed files
• Comments generated: 1
• Review effort level: Lite

[Copilot]

🔵 Needs a closer look

Repository-name canonicalization can misassociate digests, and deterministic coverage for real and multiple repository digests is still missing.

Review details

Suppressed comments (3)

src/windows/wslc/services/ImageService.cpp:334

• The new end-to-end cases use tar-loaded images, so image.Digest remains empty and this branch is never exercised with a real repo digest; the helper unit tests only validate string stripping. That leaves the new service/session wiring and row grouping unverified. Add deterministic coverage using the existing local-registry setup or a fixture that supplies at least one non-empty RepoDigests value, including the tagged-row mapping.

        if (digests)
        {
            info.Digest = DigestFromRepoDigest(image.Digest);
        }

src/windows/wslcsession/WSLCSession.cpp:1950

• This lookup uses Repository.Name, which preserves the daemon's original spelling, so equivalent references such as ubuntu:latest and docker.io/library/ubuntu@sha256:... are treated as different repositories. In that case the digest is not attached to the tagged row and an extra digest-only row is emitted; key the grouping and tagged-repository set by Repository.GetCanonical() while retaining the original spelling for display.

            auto repoName = reference->Repository.Name;
            const auto it = digestsByRepo.find(repoName);
            taggedRepos.emplace(std::move(repoName));

src/windows/wslcsession/WSLCSession.cpp:1962

• The new per-digest expansion is not exercised by the added E2E tests: their tar-loaded images explicitly have no RepoDigests, so this loop is never non-empty; the existing native test only checks that a digest can exist, not that multiple rows are emitted. Add a deterministic fixture with at least two RepoDigests for one tagged repository to validate this loop and the tag/digest grouping.

                for (const auto& repoDigest : it->second)
                {
                    rows.push_back({&e, tag, repoDigest});
                }

• Files reviewed: 16/16 changed files
• Comments generated: 0 new
• Review effort level: Lite

[Copilot]

🟡 Changes recommended

Moderate issues remain around digest-row coverage and matching, default listing semantics, and dangling-image test-key uniqueness.

Get a fresh assessment by requesting another Copilot review.

Review details

Suppressed comments (3)

src/windows/wslcsession/WSLCSession.cpp:1981

• Repository.Name intentionally preserves the original spelling rather than the normalized identity. Keying this map by Name can fail to associate valid equivalent references such as ubuntu:latest and docker.io/library/ubuntu@sha256:..., producing an undigested tag row plus an extra digest-only row. Use a canonical key for matching while retaining the original repository name for display.

            digestsByRepo[reference->Repository.Name].push_back(repoDigest);

src/windows/wslcsession/WSLCSession.cpp:2025

• When RepoTags is empty but the daemon still returns RepoDigests (for example, a pulled image after its last tag was removed), the normal listing now emits repoName with a <none> tag instead of the existing <none>:<none> dangling-image row. This changes default image-list semantics even though --digests was not requested; skip these digest-only rows when digests is false and let the fallback below create the unnamed row.

            if (!digests)
            {
                rows.push_back({&e, repoName, std::string{}});

test/windows/wslc/e2e/WSLCE2EImageListTests.cpp:570

• <none> is not a unique repository/tag key: the service emits <none>:<none> for every dangling image (WSLCTests.cpp:1167-1171). If the test environment contains two dangling images, the first emplace ID is reused and this test fails even though the digest rows are valid; skip placeholder rows or include the image ID in the uniqueness key.

            const auto it = idByRepoTag.emplace(std::make_pair(image.Repository, image.Tag), image.ID).first;
            VERIFY_ARE_EQUAL(it->second, image.ID, L"Rows sharing a repository and tag must report the same image ID");

            VERIFY_IS_TRUE(
                seen.emplace(image.Repository, image.Tag, image.Digest).second,

• Files reviewed: 15/15 changed files
• Comments generated: 1
• Review effort level: Lite

[Copilot]

🔵 Needs a closer look

Three moderate issues remain regarding repository grouping, default digest-only rows, and non-empty digest test coverage.

Review details

Suppressed comments (3)

src/windows/wslcsession/WSLCSession.cpp:1981

• This groups by Repository.Name, which deliberately preserves the input spelling rather than the normalized repository identity. If Docker returns a short RepoTag such as ubuntu:latest and a fully qualified RepoDigest such as docker.io/library/ubuntu@sha256:..., the lookup at line 1997 misses it: the tagged row gets <none> and a separate digest-only row is emitted. Key the tag/digest groups by Repository.GetCanonical() while retaining an original name separately for display.

            digestsByRepo[reference->Repository.Name].push_back(repoDigest);

src/windows/wslcsession/WSLCSession.cpp:2025

• When digests is false, this branch still emits a row for every repository found only in RepoDigests, using a bare repository name. The previous behavior emitted <none>:<none> whenever RepoTags was empty, and the CLI relies on that sentinel, so a digest-only image can change from a dangling row to repo/<none> in the default listing. Skip digest-only repositories when !digests and let the fallback below create the unnamed row.

            if (!digests)
            {
                rows.push_back({&e, repoName, std::string{}});

test/windows/wslc/e2e/WSLCE2EImageListTests.cpp:567

• All of these e2e cases use tar-loaded images with no RepoDigests, so none exercises the non-empty digest path: service field population, repo@sha256 reduction, or per-digest row expansion. The tests would still pass if those new branches were broken; add a deterministic local-registry fixture or a direct service/session fixture containing a real digest.

            // The fixtures carry no repo digest, so every row reports the placeholder.
            VERIFY_ARE_EQUAL(std::string{c_none}, image.Digest);

• Files reviewed: 15/15 changed files
• Comments generated: 0 new
• Review effort level: Lite

[Copilot]

🔵 Needs a closer look

One or more issues must be addressed before approval.

Review details

Suppressed comments (3)

src/windows/wslcsession/WSLCSession.cpp:1981

• These lookup keys use Repository.Name, which preserves the daemon's original token. If RepoTags contains ubuntu:latest while RepoDigests contains docker.io/library/ubuntu@sha256:..., the lookup at lines 1996-1997 misses, so --digests prints <none> for the tagged row and an extra digest-only row. Key both collections by a normalized repository name while retaining the appropriate familiar/original name for display, and cover mixed qualification with a test.

            digestsByRepo[reference->Repository.Name].push_back(repoDigest);

src/windows/wslcsession/WSLCSession.cpp:2009

• This expands one tagged repository/tag into one output row per RepoDigests entry. ImageService::List and both formatters consume each WSLCImageInformation as a row, so an image carrying two digests for the same repository will duplicate the REPOSITORY/TAG/IMAGE ID rows. Docker's images --digests keeps one row per tag and puts the digest in that row; preserve the one-row-per-tag behavior and select the digest to display instead of expanding the row set.

                for (const auto& repoDigest : it->second)
                {
                    rows.push_back({&e, tag, repoDigest});

test/windows/wslc/e2e/WSLCE2EImageListTests.cpp:540

• This assertion searches the entire table output, but every fixture digest is <none> and that placeholder can already occur in REPOSITORY or TAG. The test can therefore pass even if the DIGEST cells are missing or empty, leaving the real sha256 table/json consistency path untested; assert the value within the DIGEST column (or add a pulled-image table case).

            const auto digest = wsl::shared::string::MultiByteToWide(image.Digest);
            VERIFY_IS_TRUE(
                tableResult.StdoutContainsSubstring(digest),

• Files reviewed: 15/15 changed files
• Comments generated: 0 new
• Review effort level: Lite

[Copilot]

🟡 Changes recommended

A critical row-cardinality regression remains, and multi-digest E2E coverage is incomplete.

Get a fresh assessment by requesting another Copilot review.

Review details

Suppressed comments (1)

test/windows/wslc/e2e/WSLCE2EImageListTests.cpp:554

• This test claims to cover a tag with several digests, but it uses the tar-loaded fixtures and then requires every Digest to be <none> at lines 566-567. It therefore never exercises the multi-digest path added in WSLCSession.cpp:2007-2010; an implementation that ignored or collapsed additional RepoDigests would still pass. Add a repository/tag state with at least two real repo digests and assert the expected rows.

        // A tag carrying several digests is emitted once per digest, so
        // rows stay unique per repository, tag and digest while the image ID repeats across them.
        std::map<std::pair<std::string, std::string>, std::string> idByRepoTag;

• Files reviewed: 15/15 changed files
• Comments generated: 1
• Review effort level: Lite

[Copilot]

🔵 Needs a closer look

Three moderate digest-grouping issues remain unresolved, along with one test-coverage nit.

Review details

Suppressed comments (4)

src/windows/wslc/tasks/ImageTasks.cpp:260

• The real-digest registry test validates only JSON; the table tests use tar-loaded images whose DIGEST is always <none>. A regression in the non-empty digest table cell would therefore pass. Extend the registry scenario to run image list --digests and assert each pulled digest appears in the DIGEST column.

            table.WriteRow({
                MultiByteToWide(entry.Repository),
                MultiByteToWide(entry.Tag),
                digests ? MultiByteToWide(entry.Digest) : std::wstring{},
                MultiByteToWide(entry.ID),

src/windows/wslcsession/WSLCSession.cpp:2020

• This digest-only repository loop runs even when digests is false: the !digests branch appends a row for every RepoDigests entry. For a tagless image whose daemon response includes repo digests, the default listing therefore changes from the previous <none>:<none> row to repository-only rows, even though digest references were not requested. Only enumerate these repositories when digests is true so the fallback below preserves the no-flag behavior.

        for (const auto& [repoName, repoDigests] : digestsByRepo)
        {
            if (taggedRepos.contains(repoName))
            {
                continue;

src/windows/wslcsession/WSLCSession.cpp:1981

• This correlation is keyed by Repository.Name, but that field intentionally preserves the original spelling while RepositoryReference::Parse normalizes equivalent names into Server/Path (for example, ubuntu and docker.io/library/ubuntu). If the daemon uses those valid spellings across RepoTags and RepoDigests, the lookup misses, leaving the tagged row at <none> and emitting an extra digest-only row. Use a canonical Server/Path key for matching while retaining the original name for display.

            digestsByRepo[reference->Repository.Name].push_back(repoDigest);

src/windows/wslcsession/WSLCSession.cpp:2010

• The new multi-digest expansion path is not exercised by the added E2E coverage: Digests_RepeatsIdOncePerDigest uses tar-loaded images and asserts every digest is <none>, while the registry test creates only one digest per repository. Add a fixture with at least two repo digests (or a digest-only repository) and assert the resulting row/quiet multiplicity so this loop cannot regress silently.

                for (const auto& repoDigest : it->second)
                {
                    rows.push_back({&e, tag, repoDigest});
                }

• Files reviewed: 15/15 changed files
• Comments generated: 0 new
• Review effort level: Lite