Skip to content

~95% on SimpleQA (e.g. Qwen3.6-27B on a 3090). Supports all local and cloud LLMs (llama.cpp, Ollama, Google, ...). 10+ search engines - arXiv, PubMed, your private documents. Everything Local & Encrypted.

test(logpanel): exercise negative counter clamp (#5230)

LearningCircuit820fce1e4032
diff --git a/tests/js/components/logpanel.test.js b/tests/js/components/logpanel.test.js
index b2325b917f..03eecf6ea3 100644
--- a/tests/js/components/logpanel.test.js
+++ b/tests/js/components/logpanel.test.js
@@ -2034,11 +2034,11 @@ describe('loadLogsForResearch — counter drift when live entries already exist'
// and the header indicator drift below zero even though the DOM
// ends up back at the cap.
//
- // The structural fix (recompute from the DOM in
- // loadLogsForResearch before exiting the merge path) is what this
- // suite locks in. A second test pins the defensive Math.max(0, ...)
- // floor in updateLogCounter / updateFilterCounters so a future
- // refactor that drops the clamp doesn't surface as "Info -1".
+ // The first test locks in the structural fix (recompute from the
+ // DOM in loadLogsForResearch before exiting the merge path). The
+ // second pins updateFilterCounters' defensive Math.max(0, ...) floor
+ // while verifying that updateLogCounter keeps the indicator and All
+ // badge DOM-derived, so a future refactor cannot surface "Info -1".
//
// setupPanelDom() is intentionally NOT used here: it calls
// logPanel.initialize(rid), which auto-fires a pre-fetch and
@@ -2136,49 +2136,66 @@ describe('loadLogsForResearch — counter drift when live entries already exist'
});
});
- it('clamps the indicator and badges to >= 0 when the underlying state is negative', () => {
+ it('clamps negative category badges while indicator and All follow the rendered DOM', () => {
// Guards the defensive Math.max(0, ...) floor in
- // updateLogCounter / updateFilterCounters. The structural
- // recompute in loadLogsForResearch handles the bulk-merge
- // path; this test pins that a future insertion path that
- // bypasses the recompute (e.g. a regression in
- // addLogEntryToPanel's prune/increment pairing) can't surface
- // as "Info -1" before the recompute is extended to cover it.
- //
- // Seed DOM/state directly so the test stays fast and
- // deterministic — no initialize(), no real fetch, no real
- // addLog round-trip.
+ // updateFilterCounters. The structural recompute in
+ // loadLogsForResearch handles bulk paths; this test pins that
+ // a future live insertion path with stale negative category
+ // state cannot surface a negative badge before the state is
+ // structurally recomputed.
buildPanelDom();
window._logPanelState.counts = emptyCounts();
window._logPanelState.counts.info = -50;
window._logPanelState.counts.warning = -10;
+ window._logPanelState.expanded = true;
+ // Stale indicator text is the one piece of pre-seeded DOM the
+ // code path can read back: updateFilterCounters derives the All
+ // badge from the indicator, so the insert must refresh it from
+ // the rendered DOM before that read. The filter badges are pure
+ // outputs and need no seeding.
document.getElementById('log-indicator').textContent = '-60';
- // Trigger a re-render via the public addLog path, which
- // reaches updateFilterCounters. We have to initialize()
- // first so window._socketAddLogEntry is registered and the
- // addConsoleLog fallback (which logs-and-bails after #5128)
- // doesn't drop the entry. initialize() with a fresh research
- // id is cheap (no DOM changes happen because state is
- // already seeded) and bounded by the fetch mock below.
- globalThis.fetch = vi.fn(() =>
- Promise.resolve({ json: () => Promise.resolve([]) })
- );
- logPanel.initialize('clamp-guard-rid');
-
- // addLog writes through _socketAddLogEntry (set up by
- // initialize) into addLogEntryToPanel(logEntry, true), which
- // calls updateLogCounter(1) and updateFilterCounters(). With
- // counts.info seeded at -50, incrementCounter=true adds +1 to
- // counts.info (now -49) and updateFilterCounters' clamp keeps
- // the badge at "0". Indicator textContent was "-60"; +1 from
- // updateLogCounter clamps to "0".
logPanel.addLog('trigger-render', 'info');
- expect(getIndicatorCount()).toBeGreaterThanOrEqual(0);
- expect(getFilterCount('info')).toBeGreaterThanOrEqual(0);
- expect(getFilterCount('warning')).toBeGreaterThanOrEqual(0);
- expect(getFilterCount('all')).toBeGreaterThanOrEqual(0);
+ expect(
+ document.querySelectorAll('.ldr-console-log-entry')
+ ).toHaveLength(1);
+ expect(window._logPanelState.counts.info).toBe(-49);
+ expect(window._logPanelState.counts.warning).toBe(-10);
+ expect(document.getElementById('log-indicator').textContent).toBe('1');
+ expect(
+ document.querySelector('[data-filter-count="all"]').textContent
+ ).toBe('1');
+ expect(
+ document.querySelector('[data-filter-count="info"]').textContent
+ ).toBe('0');
+ expect(
+ document.querySelector('[data-filter-count="warning"]').textContent
+ ).toBe('0');
+ });
+
+ it('recovers through the exact zero boundary as inserts land on a negative counter', () => {
+ // Companion to the clamp test above, covering the boundary where
+ // the clamp stops being the deciding factor: with info at -1 the
+ // first insert lands exactly on 0 (state and clamped badge agree),
+ // and the second proves the badge resumes tracking the true count
+ // once the state itself turns positive.
+ buildPanelDom();
+ window._logPanelState.counts = emptyCounts();
+ window._logPanelState.counts.info = -1;
+ window._logPanelState.expanded = true;
+
+ logPanel.addLog('first render', 'info');
+ expect(window._logPanelState.counts.info).toBe(0);
+ expect(
+ document.querySelector('[data-filter-count="info"]').textContent
+ ).toBe('0');
+
+ logPanel.addLog('second render', 'info');
+ expect(window._logPanelState.counts.info).toBe(1);
+ expect(
+ document.querySelector('[data-filter-count="info"]').textContent
+ ).toBe('1');
});
it('counts untracked categories (e.g. DEBUG) toward the header indicator and All badge after a bulk load', async () => {