Study: opening a guide from the Recent guides rail clears the exam selection, leaving Regenerate permanently disabled

[AndresL230]

Found while testing #475 (the #141 reframe), pre-existing on main.

Repro: on /study, open a study guide from the "Recent guides" rail (not via an exam selection). The exam selection state is cleared, and the Regenerate button stays permanently disabled on that path — there is no way to regenerate the guide you're looking at without re-selecting the exam first.

Where: frontend/src/components/screens/Study.tsx — the recent-guides open path clears the selected exam; Regenerate is gated on it.

Small UX fix; noted in PR #475's description. Not addressed there to keep that PR scoped to semester threading.

[AndresL230]

Fixed in #499 (part of #476). Leaving this open for your UI verification.

The issue body misattributes the cause

"the recent-guides open path clears the selected exam"

It doesn't — openRecent sets both ids. The clearing came from the
courseId-keyed exams effect, whose first statement was an unconditional
setExamId(""). That reset can't tell "the user switched course" (selection
now invalid) from "we just opened a specific guide" (selection deliberate and
valid); it only sees that courseId differs from last render, which
openRecent also causes.

Both effects run in the same commit, so the loader still saw the intact pair
and the guide loaded — only the next render lost the exam. That's why the
symptom was "a guide on screen above a dead Regenerate button" rather than
"nothing opens".

It needs a course change to reproduce. A rail entry whose course was
already selected always worked. That's pinned as a control test — it passed
before the fix and still passes, which is what proves the mechanism is the
reset rather than the open path.

What changed

The reset now happens at the two events that mean it: the course picker's
onChange, and a term switch. The term case adjusts state during render (the
StudyModePanel pattern already in the file) instead of in an effect, because
an effect-time reset lands a render late — the loader would commit one read of
the old exam under the new term first. That was the same defect's second
trigger; it reproduced on main and now has a test.

Making Regenerate reachable also exposed a term hazard: it sent the active
selector's term while a recent entry opens under its own term (#475 F1).
Latent before, because the button was dead on the one path where those differ.
Regenerate now replays the term the displayed guide was loaded with.

What to check when you verify

1. /study, click any entry in "Recent guides" → the exam picker should show
   that exam and Regenerate should be clickable.
2. Switch course in the picker → exam selection clears (unchanged behavior).
3. Re-pick the course you're already on → the guide should stay put. (Code
   review caught this one: the first cut of the fix wiped it.)
4. Scope to a term, then open a rail entry from a different term → guide and
   Regenerate work and target the entry's term, but the exam picker still shows
   its placeholder. Known, recorded in a code comment: the options list follows
   the active selector by feat(study): semester-scope the study-tool reads (#141 reframe) #475's design. Squaring it means tracking "the term
   I'm viewing" across list and loads — a feat(study): semester-scope the study-tool reads (#141 reframe) #475-shaped change, not this fix.

Verification

• Study.recentGuides.test.tsx — 8 cases, written red first (5 failed pre-fix
  on the Regenerate/picker assertions and the stale term re-read).
• e2e/study-recent-guides.spec.ts — promoted regression journey. Verified it
  goes red against pre-fix Study.tsx: expect(examPicker).toHaveText(/Midterm Exam/)
  got "Pick an exam…▾".
• Full local cycle twice (after the fix, and after the review fix):
  Playwright 37/37, oracles 0 findings, backend pytest 1529 passed, frontend
  vitest 423/423, tsc + eslint clean.

Follow-up worth filing separately: loadGuide has no stale-response guard, so
two rail entries clicked in quick succession can resolve out of order.
Pre-existing; this fix adds loadedTerm to the state that can go stale.

[AndresL230]

Verified in a real browser on the local stack as the seeded student, and closing.

• Opened the "Midterm Exam" entry from the Recent guides rail → course picker reads CS101, exam picker reads Midterm Exam (pre-fix it read "Pick an exam…"), guide rendered, and Regenerate.disabled === false. That is the issue's exact complaint, gone.
• Re-picked the course already selected → guide stayed put, Regenerate still enabled. This is the regression code review caught in my first cut, so it was worth confirming by hand rather than trusting the unit test alone.
• Switched to a different course (MATH210) → exam selection cleared, guide cleared, Regenerate gone. The behaviour that should reset still resets.

Known limitation, unchanged and documented in a code comment: open a rail entry whose term differs from the active semester selector and the exam picker still shows its placeholder, because the options list follows the selector by #475's design. The guide and Regenerate are correct and aimed at the right term; only the picker label is blank. Squaring that means tracking "the term I'm viewing" across list and loads, which is a #475-shaped change.