feat(nomadnet): inline page images — parser, WebP decode, /media transport, policy + progress UI, render fixes

[torlando-tech]

Summary

Adds NomadNet inline page-image rendering end to end on the T-Deck, conforming to upstream NomadNet e1e8ab8 (v1.4.3). Two commits:

• 929fa2c — core feature: compact-page image parser, bounded WebP→RGB565 decoder (libwebp 1.3.2 decode-only), sequential /media Resource transport over the retained page Link, and LVGL render path (slot LRU in PSRAM + lv_draw_img_decoded).
• 091c93a — finishing: image-loading policy setting (never/manual/auto/always), manual "Load images" chrome button, transfer progress UI, decoder byte-swap fix, image blit render fix, fork pin bump, and the media interop gate.

What's in the finishing commit

• Settings — new Browser card for the page-image loading policy, persisted in NVS (img_load, default auto), effective on next page load. Manual mode surfaces a "Load images" button (placed just right of home, no title overlap).
• Loader — strictly sequential /media fetch, per-image terminal failure (one bad image never fails the page), progress-aware deadline (base window + RTT slack, refreshed per part).
• Decoder — LV_COLOR_16_SWAP byte-swap on-device for lv_draw_img_decoded buffers; host tests keep mathematical RGB565.
• Render fix — decoded images blit with a uniform two-axis 8.8 zoom (no upscale) sized from the placement box. Two defects fixed: a width-only zoom made height-constrained boxes overflow the content edge (the "third image off the side"), and passing the display size as coords made the LVGL SW path stride a wide buffer as narrow (the "diagonal offset rainbow").
• Progress UI — thin full-width bar under the header (overall page fraction) + status line "Image N of M - % - KB".
• Fork pin — microReticulum 921b3aa → 0487775 (media Resource response fix + jobs()/outbound() self-re-entrancy guard) at all six sites.
• Tests — x86 real-peer flow gains the media scenario (raw file-response delivery, per-part progress; 12/12), reference bumped to NomadNet 1.4.3 (e1e8ab8) / RNS 1.5.4 with hash-pinned files; CI installs matching rns/nomadnet and the flow timeout is raised to 300 s.
• Diagnostics — T:IMG_STATE serial dump (loader entries + decode slots), test-hook gated and stripped from tdeck-release (audit-verified).

Verification

• pio run -e tdeck and tdeck-release SUCCESS; tools/audit_release_build.py passes (pins + stripped test hooks/strings confirmed against the release binary).
• x86 real-peer flow: 12/12 scenarios PASS including the new media (raw file-response, resource_progress=1), manifest + reference hash-pinned.
• pytest tests/build_scripts — 184 passed.
• Physical: flashed to the T-Deck; colors correct, "Load images" button clears the title bar, full-width image renders letterboxed inside the content box, progress bar + status line update live during a LoRa load. (Final image-load visual pass confirmed by Torlando.)

Notes

• The re-entrancy guard fix lives in the pinned fork (torlando-tech/microReticulum 0487775), not upstream, per standing policy.
• Cached-page-with-unloaded-images still shows placeholders without a Load button (pre-existing gap, tracked separately — not part of this PR).

[greptile-apps]

Confidence Score: 5/5

The latest changes appear safe to merge, with no outstanding correctness or repository-rule violations.

Findings

1.  Partial Refresh Drops Images ▶

Fix with agent prompt

### Issue 1
lib/tdeck_ui/UI/LXMF/NomadNetCompactPage.h:undefined-149
When a dynamic partial is refreshed, `assign_replacing_partial()` rebuilds every retained block but does not copy or remap image records. Both images in the fragment and existing images elsewhere on the page therefore lose their `image_index` association and fall back to placeholders after the refresh. The replacement path needs to preserve image records and remap their indices.

---

For each issue above, determine whether it is valid and should be fixed. If so, fix it directly.

Summary

The latest revision fixes bounded image-record reuse during dynamic partial refreshes.

• Blanked records from earlier shrink operations are reused before new records are appended.
• Retained image indices remain stable for decoder and loader state.
• Added regression coverage exercises repeated grow/shrink cycles and verifies all current image blocks resolve to live records.
• The previously unresolved partial-refresh finding is fully addressed by preserving base image records and remapping fragment image records.

Diagram

%%{init: {'theme': 'neutral'}}%%
flowchart TD
    A[Partial refresh] --> B[Classify existing image records]
    B --> C[Retained records<br/>keep stable indices]
    B --> D[Current region records<br/>replace in place]
    B --> E[Blank dead records<br/>reuse for overflow]
    D --> F[Map fragment images]
    E --> F
    F --> G{More fragment images?}
    G -- Yes, capacity available --> H[Append new records]
    G -- No or capacity reached --> I[Use mapped records or text fallback]
    H --> I

Loading

_{Reviews (6) · Last reviewed commit: "fix(nomadnet): reuse blanked image recor..."}

[greptile-apps]

Partial Refresh Drops Images

When a dynamic partial is refreshed, assign_replacing_partial() rebuilds every retained block but does not copy or remap image records. Both images in the fragment and existing images elsewhere on the page therefore lose their image_index association and fall back to placeholders after the refresh. The replacement path needs to preserve image records and remap their indices.

Prompt To Fix With AI

This is a comment left during a code review.
Path: lib/tdeck_ui/UI/LXMF/NomadNetCompactPage.h
Line: 149

Comment:
**Partial Refresh Drops Images**

When a dynamic partial is refreshed, `assign_replacing_partial()` rebuilds every retained block but does not copy or remap image records. Both images in the fragment and existing images elsewhere on the page therefore lose their `image_index` association and fall back to placeholders after the refresh. The replacement path needs to preserve image records and remap their indices.

---

For each issue above, determine whether it is valid and should be fixed. If so, fix it directly.