Skip to content

Evidence: Incognito project history privacy

Pre-release
Pre-release

Choose a tag to compare

@voltcrash voltcrash released this 02 Nov 16:23
1d8e405

Verification artifacts for the focused Incognito project-history fix. This is an evidence-only prerelease on the contributor fork, not a product release. No assets are committed to the source branch.

Real dev Electron, native preload/bridge, non-headless primary-environment WebContents guests and CDP webview targets were checked. The baseline is upstream ac8dce0; patched captures are from 1d8e405. The branch was subsequently rebased over unrelated upstream changes. A later review fix resolves the project before rejecting the current Incognito visit, preserving previously queued regular visits; a focused fail-before/pass-after regression covers that state-only delta. The captures remain proof of the address-bar/native restart flow, not of the pending-project corner case.

For each variant, the UI opened an Incognito tab and entered a unique local fixture URL, then a Default tab and a distinct regular URL, then a fresh regular empty tab. Baseline Recently used and serialized t3code:browser-history:v1 contain both URLs; patched contains only regular. The primary quit both native Electron browser processes via CDP and observed endpoints stop, then the existing dev launchers relaunched them with the same isolated server homes and separate preserved XDG_CONFIG_HOME profiles. Post-restart checks prove changed browser identity, same t3code-dev://app origin and localStorage sentinel, regular URL persistence, baseline private URL persistence, and patched private URL exclusion.

All four JSON results pass. Both pre/post PNG pairs show the real native app's regular Recently used view: baseline two rows (regular/private), patched one (regular). Screenshots are cropped to the browser panel and show only synthetic history and loopback test ports. The implementation agent inspected all four PNGs and JSON results after the primary executed the integrated pass.

Limits: this native run did not manually exercise configured/discovered-server cards, terminal-link opening, or same-URL private title updates. Store regression tests independently cover private-title filtering at settle and delayed-visit timing, pending-visit exclusion, serialized history and rehydration, and undefined/default/custom persistent profiles. Unchanged entry paths and propagation callers were inspected in source. No mobile or remote/tunnel run is claimed. Video is unnecessary for this static persistence defect.