Troubleshooting and support
Start with the first failed layer, not the last screen you saw. Preserve unsaved text and the exact error before refreshing, signing out or repeating a request. A later warning can occur after a change has already taken effect.
Choose the failing layer
| Symptom | Open |
|---|---|
| Build, routes, assets or search | Build and reader issues |
| Sign-in, files, scope or accounts | Editor and account issues |
| Publish stage or stale remote site | Publish and deployment issues |
| Product boundary or bug report | FAQ and support |
These pages distinguish exact server messages from symptoms and parameterized message patterns. Inspect the real status/body and saved state. Do not weaken Host, Origin, CSRF, permissions or file validation merely to make an error disappear.
Establish what changed
| State | Evidence |
|---|---|
| Browser buffer | Current text/recovery state |
| Saved source | Actual server file/settings |
| Local artifact | Build report and generated files |
| Remote artifact | Host deployment and responses |
| Reader cache | Actual loaded URLs/versions |
A saved file can coexist with an old build. A correct local build can coexist with an old remote deployment. A switch warning can coexist with new local output. History can fail to write after the action, so an absent event is not proof the action failed. Check current state before another destructive or duplicate operation.
Collect useful diagnostics
Record the pinned revision, actual Node/runtime, command or request, first relevant error, relative source path or reader route, expected result and actual result. For visual issues include theme, viewport and a safe screenshot. For search include the actual query and target; for publication include stage and reported live state.
Use a small synthetic reproduction, not private source or accounts. Remove passwords, cookies, tokens, private user files, real identities and unrelated paths. Inspect screenshot pixels and log text before sharing. The FAQ and support page gives a report table and safe handling guidance.
Verify the repair
Rebuild or restart only the affected layer after preserving recovery evidence. Check the original failing task, then home, a deep page, a missing route, assets and search when reader delivery changed. Test server authorization with a representative synthetic role when writing scope changed. Inspect real deployed responses separately from local preview.
If the source is correct but the domain, TLS, proxy or host is not, investigate that external layer. MarkStack’s local tests do not certify your infrastructure. Validation limits makes those remaining checks explicit.