[mango (mangomac)]

Fixes #1836.

The diagnostic feature ignored structure.array changes, so reordering notebook cells did not trigger a new pull. With interFileDependencies enabled it now pulls after a structure change.

Tests: a new client-node-tests case moves a cell and asserts a pull happens. I also set interFileDependencies: true in fullNotebookServer. Before the change the suite was 214 pass / 1 fail, after it is 215 / 0. Only the client package was tested; the browser and other packages were not run.

Note: this pulls all matching cells and could be narrowed. It does not change the 500ms delay discussed in #1837.

[mango (mangomac)]

@microsoft-github-policy-service agree

[Copilot]

🟡 Changes recommended

Structural refreshes duplicate newly opened cell requests, and the regression test can pass on unrelated pulls.

2 open issues

• Verify reorder diagnostics target the moved cell after synchronization · New
• Avoid duplicate diagnostic pulls for newly opened or changed cells · New

What changed in this PR

Refreshes notebook diagnostics after cell reordering to address #1836, without changing the scheduler delay discussed in #1837.

Changes:

• Pulls matching cells after structural changes when inter-file dependencies are enabled.
• Enables that capability in the test server and adds a reorder regression test.

Verification here was limited to source inspection; no tests were run.

File	Description
client/​src/​common/​diagnostic.ts	Adds diagnostic pulls after notebook structure changes.
client-node-tests/​src/​servers/​fullNotebookServer.ts	Enables inter-file diagnostic dependencies.
client-node-tests/​src/​integration.test.ts	Adds a cell-reorder regression test.

_{Review effort: Balanced}

[Dirk Bäumer (dbaeumer)]

mango (@mangomac) see copilot review

[mango (mangomac)]

Pushed ff03c87 to address the review notes.

• diagnostic.ts: on cell insert or replace, the same cell could be pulled twice (once from the opened-cells loop, once from the structure loop). The structure loop now skips URIs that were already pulled in that pass.
• integration.test.ts: the reorder test now settles first, resets the recorded pulls, moves the cell, and polls (bounded, 3s) until the moved cell's URI shows up as pulled, instead of asserting that any pull happened after a fixed delay.

Caveat: I compiled the two files, but the VS Code integration tests need the electron host and I could not run them locally, so that part is untested. CI will be the first real run.

[Copilot]

🟢 Approval recommended

The focused fix preserves capability gating and existing request patterns, with regression coverage and no identified blocking issues.

No open issues 🎉

2 resolved since last review

• Avoid duplicate diagnostic pulls for newly opened or changed cells
• Verify reorder diagnostics target the moved cell after synchronization

_{Review effort: Balanced}

[Copilot]

🟢 Approval recommended

The focused change follows existing diagnostic-pull handling, includes regression coverage, and has no identified blocking issues.

No open issues 🎉

_{Review effort: Balanced}