1. Summary

This round closed all 10 sections/line-groups on #519’s prioritized gap list (found via the automation-status cross-check added the day before, 2026-07-02) — 116 previously-SKIP: not automated test-plan lines now have real subprocess-level e2e/*_test.go coverage. Unlike prior rounds, this was automated test authoring, not a manual click-through of the plan — so this report’s numbers come straight from scripts/e2e-test-report.sh (also added this round, see "New Tooling" below) rather than hand-transcription.

Section Tests Pass Fail Not Automated (this round’s tagging convention)

1

Basic Workflow

6

0

6

2

Validate

25

25

0

3

Forward Sync

23

0

23

4

Reverse Sync + Conflicts

22

7

15

5

Views / Wildcards / Scope / Lifting

28

28

0

6

CLI Add Commands

30

6

24

7

Watch Mode

10

0

10

8

Export

12

0

12

8b

Export-Diagram

10

0

10

8c

Export-Table

6

0

6

8d

Export-Sequence

4

0

4

9

Draw.io File Integrity

5

5

0

10

Security / Injection Testing

9

9

0

11

CLI Flag Interactions

9

9

0

12

Sync State Edge Cases

6

6

0

13

Model File Edge Cases

7

7

0

14

Template Handling

6

6

0

15

Error Output Formatting

8

8

0

16

Analysis & Inspection Commands

15

0

15

17

Evolution & Snapshots

9

0

9

18

Model Tooling & Import

16

0

16

19

Interactive, Overlay & ADR

7

0

7

20

Language Server (LSP)

6

0

6

Total

279

116

163

Overall result: 116 PASS, 0 FAIL, 163 "not automated" by this round’s tagging convention.

Important
 The "Not Automated" column above reflects only lines tagged via this round’s <planID>_…​ subtest naming convention (see internal/e2eplan) or registered in its exception registry — it is not the same as "zero e2e coverage". Sections 3, 7, 8/8b/8c/8d, 16-19 already have substantial coverage from e2e/*_test.go files predating this convention (see the "Automation Status" table in E2E-Test-Plan.adoc, which tracks coverage qualitatively per-file); they just aren’t yet wired into the machine-readable per-line report. Retagging them is future work, not a regression introduced here.

2. Environment

  • Sandboxed dev environment (not the devcontainer); Go 1.19 preinstalled did not satisfy go.mod’s `go 1.25.8 directive (`go1.19’s toolchain parser predates the 3-component version format) — Go 1.26.4 was downloaded and used for this session’s builds/tests instead.

  • drawio / drawio-export / xvfb-run available; xmllint not available (draw.io XML well-formedness checks used internal/drawio.LoadDocument re-parsing instead, consistent with how the CLI itself parses the file).

  • golangci-lint and actionlint were installed ad hoc via go install to verify new code before committing.

3. New Automation Added

One new e2e/*_test.go file per prioritized section (jsonc_comment_preservation_test.go, label_html_handling_test.go, views_wildcards_lifting_test.go, security_injection_test.go, cli_flag_interactions_test.go, error_output_format_test.go, drawio_file_integrity_test.go, model_file_edge_cases_test.go, sync_state_edge_cases_test.go, template_handling_test.go, validate_edge_cases_test.go), plus a runCLISplit helper (stdout/stderr captured separately) added to helpers_test.go for the JSON-error-envelope and stdout-cleanliness assertions in sections 11 and 15.

4. New Tooling

Implements #519’s "per CI statt manuell erzeugen" follow-up:

  • internal/e2eplan — plan-ID → test-name registry, plus a <planID>_…​ subtest naming convention that’s auto-detected without any registry entry.

  • scripts/e2e-test-report-gen + scripts/e2e-test-report.sh (make e2e-test-report) — generates this report’s Summary table (and a full 279-line per-plan-line detail table, omitted here for length — see the e2e-test-plan-report CI artifact or run the script locally) directly from go test ./e2e/…​ -json.

  • Wired into .github/workflows/go.yml as a new e2e-plan-report job: job summary + build artifact + PR comment. Does not auto-commit a dated report file to main — left as a separate, more consequential decision per #519’s own caution about CI commits.

5. Bug Found & Fixed

Issue Description Status

#520

mergeStyles (internal/sync/forward.go) built the merged style string by ranging over a Go map, so two freshly-added same-kind elements in one sync could get differently-ordered (value-equivalent) style attributes — non-reproducible XML output, noisy diffs. Found writing 9.3 (style consistency).

✅ Fixed same branch: overlay key order now tracked in a slice, not derived from map iteration.

6. Test Plan Corrections

Four "Expected" outcomes in E2E-Test-Plan.adoc were stale relative to current behavior and corrected in the same change that closed each section (not treated as product bugs):

Line(s) Correction

5.13, 5.26, 5.27

Described as "Warning, graceful" / "No match, empty page" outcomes; actual behavior since #176’s fix is a blocking model.Validate error (sync exits 1) — softer pre-#176 behavior was still documented.

5.23

Clarified that "Empty page" only holds for a wildcard nonexistent-include pattern (e.g. ["nonexistent."]); a literal nonexistent ID without hits the same blocking error as 5.13/5.26/5.27.

12.5

Described as "Clear error"; internal/sync/state.go:60-65 explicitly treats a zero-byte sync-state file the same as a missing one (empty state, no error) — "e.g. truncated write".

7. Changelog

Version Change

2026-07-03

Round 8 (automated): 116 test-plan lines newly automated across 10 sections/line-groups from #519’s prioritized list, including the previously-optional section 2 (Validate). 1 bug found and fixed (#520). 4 stale test-plan expectations corrected. Added scripts/e2e-test-report.sh tooling and wired it into CI.