public/obsidian-livesync
1
0
Fork 0
mirror of https://github.com/vrtmrz/obsidian-livesync.git synced 2026-07-23 04:52:58 +00:00
Code Issues Packages Projects Releases Wiki Activity

Retire obsolete browser Harness

Browse Source
This commit is contained in:
vorotamoroz
2026-07-20 13:22:22 +00:00
parent 80d1416a81
commit 4cbe5679e1
38 changed files with 45 additions and 2689 deletions
Download Patch File Download Diff File Expand all files Collapse all files
docs/adr/2026_06_real_obsidian_e2e.md
+11 -10
View File
@@ -2,15 +2,15 @@
## Status
Proposed / Spike Implemented
Accepted / Implemented
## Release
Not yet. Planned after the serviceFeature refactoring branch is reviewed.
Not yet. Included in the 1.0.0 release preparation.
## Context
The current end-to-end tests run through Vitest browser mode and a mocked Obsidian environment in `test/harness`. This has been useful for exercising synchronisation flows without launching Obsidian, but it is no longer a reliable final signal for plug-in behaviour.
When this decision was proposed, the end-to-end tests ran through Vitest browser mode and a mocked Obsidian environment in `test/harness`. This was useful for exercising synchronisation flows without launching Obsidian, but it was no longer a reliable final signal for plug-in behaviour.
The main issues are:
@@ -31,7 +31,7 @@ The long-term test pyramid should be:
2. Integration tests for CouchDB, Object Storage, P2P services, database operations, and replication protocols.
3. Real Obsidian E2E tests for boot-up sequence, vault reflection, command registration, settings dialogues, restart scheduling, and user-visible workflows.
The existing `test/harness` should be demoted to a transitional compatibility layer. It may remain temporarily while the real Obsidian runner reaches parity for critical flows, but new high-level E2E coverage should target the real runner.
Retain the existing browser Harness only as a transitional compatibility layer while the real Obsidian runner reaches parity for critical flows, then remove it. New high-level E2E coverage targets the real runner.
## Non-Goals
@@ -155,7 +155,7 @@ Current verification:
Known limits:
- The smoke runner currently proves only one-vault launch and plug-in load readiness. Broader workflows are covered by separate real Obsidian scripts, including CouchDB upload, startup scan, two-vault note synchronisation, Hidden File Sync, Customisation Sync, and setting Markdown export.
- Cross-platform support is still discovery-level. The working path has been validated on Linux ARM64. macOS and Windows should be validated in their own environments as follow-up work.
- The working path has been validated on Linux ARM64 and on macOS with Obsidian 1.12.7. Windows remains unverified.
- CI wiring is intentionally not implemented. The runner depends on a licensed desktop application and is treated as a local verification tool.
### Phase 2: First Real Workflow
@@ -199,9 +199,10 @@ Current implementation status:
Current implementation status:
- `test/harness` is now documented as a transitional compatibility layer.
- New broad E2E work should target `test/e2e-obsidian/` when real Obsidian behaviour is the risk being tested.
- The next high-value scenarios are RedFlag variants and Fast Setup (Simple Fetch) variants, not a line-by-line migration of `test/suite`.
- The mocked Vitest browser suites, their P2P runner, their root-level relay helpers, and the manual `harness-ci` workflow have been removed after maintained suites covered the critical flows.
- CouchDB and Object Storage combinations, with and without encryption, are owned by the CLI two-Vault matrix. P2P validation is owned by the CLI Compose E2E suite. Plug-in lifecycle and vault integration are owned by real Obsidian E2E.
- The Obsidian compatibility implementation still needed by the Webapp has moved to `src/apps/webapp/obsidianMock.ts`; it is not a retained browser E2E Harness.
- Remaining high-value scenarios, including RedFlag and Fast Setup (Simple Fetch) variants, should be added according to their owning integration boundary rather than copied line by line from the retired suite.
## Local Verification Strategy
@@ -240,12 +241,12 @@ Real Obsidian E2E is a local verification layer. It should not be wired into the
4. Add `test:e2e:obsidian:smoke` for one-vault plug-in load.
5. Document required local environment variables, especially `OBSIDIAN_BINARY`.
6. Port one CouchDB-backed workflow after the smoke test is stable.
7. Mark `test/harness` as transitional and block new broad E2E work from targeting it.
7. Retire the browser Harness after critical workflows have replacement coverage.
8. Add the local suite script for broader local verification.
## Consequences
- Real Obsidian E2E becomes the source of truth for plug-in lifecycle and vault integration.
- Unit and integration tests remain the primary fast feedback loops.
- The old browser harness can be deleted once the new runner covers the critical workflows.
- The old browser Harness has been deleted now that the replacement coverage owns its critical workflows.
- The project will gain slower but higher-confidence tests for the behaviours most likely to differ between mocks and Obsidian itself.