chore(no_run): drop 3 dead NEEDS_FIX markers + 1 inert SLOW entry in HowTo* repos

[Jammy2211]

Overview

Follow-up from PyAutoLabs/autogalaxy_workspace#143, which found that HowToGalaxy's config/build/no_run.yaml carried an entry copy-pasted wholesale from the parent workspace repo when the tutorials were split out. Auditing every entry in both HowTo repos against the real matcher (PyAutoHands/autobuild/build_util.py::should_skip) shows the copy-paste is systemic:

Repo	Entries	Match ≥1 file	Match zero files
HowToGalaxy	18	1 (tutorial_searches)	17
HowToLens	32	2 (tutorial_searches, tutorial_5_borders)	30

This issue takes a deliberately narrow slice: the entries that actively mislead. The bulk cruft is reported but left alone.

The two problems being fixed

1. Dead NEEDS_FIX markers (3). These advertise open bugs that do not exist and inflate Heart's stale-parked-script count (currently 58):

• HowToGalaxy: ellipse/modeling, guides/advanced/over_sampling
• HowToLens: group/slam

None matches any file — HowToGalaxy holds only chapter_*/ + simulators/, HowToLens only chapter_*/ + simulator/, and patterns containing / are substring-matched against the file path.

2. A broken pattern whose stated reason rests on a cap that does not exist (1).
howtolens/chapter_4_pixelizations/tutorial_10_brightness_adaption # SLOW 2026-04-10 - pixelization tutorial exceeds 60s test timeout

This one is not dead cruft — the file genuinely exists at scripts/chapter_4_pixelizations/tutorial_10_brightness_adaption.py. The stale pre-split howtolens/ prefix means the substring never matches, so the entry has not applied since the repo split and the tutorial has been running in CI all along.

The obvious move — repair the prefix — is wrong:

• The 60s cap does not exist. PyAutoHands/autobuild/run_all.py sets DEFAULT_TIMEOUT_SECS = 300.
• Measured: 233.7s, exit 0 under the HowToLens smoke profile. It fits the real cap.

So repairing the prefix would newly park a script that passes. Deleting the entry is behaviour-preserving (it is inert today) and matches reality. Deleted rather than repaired.

Plan

• Delete the 3 dead NEEDS_FIX entries.
• Delete the 1 broken, inert SLOW entry rather than repairing it, per the measurement above.
• Verify no live entry is lost: after the change HowToGalaxy still resolves tutorial_searches, HowToLens still resolves tutorial_searches + tutorial_5_borders — byte-identical live sets.
• No script changes, so no notebook regeneration.

Detailed implementation plan

Affected Repositories

• HowToGalaxy (primary)
• HowToLens

Branch Survey

Repository	Current Branch	Dirty?
./HowToGalaxy	main	clean
./HowToLens	main	clean

Branch: feature/howto-no-run-stale-entries
Worktree: ~/Code/PyAutoLabs-wt/howto-no-run-stale-entries/

Implementation Steps

1. HowToGalaxy/config/build/no_run.yaml — delete the ellipse/modeling and guides/advanced/over_sampling lines.
2. HowToLens/config/build/no_run.yaml — delete the group/slam and howtolens/chapter_4_pixelizations/tutorial_10_brightness_adaption lines.
3. Re-parse both YAMLs and re-run the should_skip audit, asserting the live-entry set is unchanged and no NEEDS_FIX string remains.

Verification performed

• should_skip audit of all 50 entries across both repos against every tracked scripts/**.py.
• tutorial_10_brightness_adaption.py timed end-to-end: 233.7s, exit 0, zero tracebacks, under smoke profile (PYAUTO_TEST_MODE=2 + skips).
• Post-change: HowToGalaxy 16 entries / live ['tutorial_searches']; HowToLens 30 entries / live ['tutorial_searches', 'tutorial_5_borders'].

Out of scope — reported, not fixed

• ~44 remaining zero-match entries across both repos. Not blindly deleted: several are stem patterns (fits_make, png_make, data_fitting, deflections) that would begin applying if a script with that stem is ever added, so they may be prospective guards rather than cruft. Needs a deliberate call.
• tutorial_6_model_fit (HowToLens, # InversionException due to test mode) — no file of that stem has ever existed in HowToLens history, and the intended target is ambiguous (tutorial_8_model_fit.py exists in chapter_4). Left for a human rather than guessed at.

Risk note

tutorial_10_brightness_adaption consumes 233.7s of the 300s budget (22% headroom) on a laptop. If CI is slower it could time out. Should that happen, the correct fix is a correctly-spelled SLOW entry (chapter_4_pixelizations/tutorial_10_brightness_adaption), not the broken one being removed here.

Refs PyAutoLabs/autogalaxy_workspace#143

Original prompt

Filed as PyAutoMind/draft/bug/howto/howto_no_run_stale_and_broken_entries.md from the follow-up identified while closing #143.

[Jammy2211]

Shipped and merged

• chore(no_run): drop 2 dead NEEDS_FIX markers #35 MERGED — dropped ellipse/modeling + guides/advanced/over_sampling.
• chore(no_run): drop dead group/slam marker and inert tutorial_10 SLOW entry HowToLens#44 MERGED — dropped group/slam + the inert tutorial_10_brightness_adaption SLOW entry.

Live-entry sets verified byte-identical before and after: HowToGalaxy resolves tutorial_searches; HowToLens resolves tutorial_searches + tutorial_5_borders. No NEEDS_FIX string remains in either file. Config-only, no notebook regeneration.

Heart gate: YELLOW score 52, red_reasons: [], reason set unchanged from the human acknowledgement given earlier in the session.

Net effect across this follow-up and #143: every remaining 2026-04-10 NEEDS_FIX marker in the HowTo* repos is now gone, and none of them turned out to be a real bug — 3 were copy-paste entries matching zero files, and the 4th (HowToFit tutorial_5_results_and_samples) is being cleared separately by PyAutoFit#1412 + HowToFit#24.

Closing. The two out-of-scope items below remain open questions, deliberately not guessed at:

1. ~44 zero-match entries still in these two lists. Several are stem patterns (fits_make, png_make, data_fitting, deflections) that would begin applying if such a script were added, so they may be prospective guards rather than cruft.
2. tutorial_6_model_fit (HowToLens) — no file of that stem has ever existed in the repo's history; tutorial_8_model_fit.py exists in chapter_4 and may or may not be the intended target.