docs(agent): #95 both Ghostty fails are Acer pre/post HM, XPS is fine

Bernardo: XPS starts Ghostty. Aspire only fails — first attempt before
home built (OpenGLOutdated), second after home built (quiet GDK_DISABLE
exit). Kitty was the manual working shell used to run both.

agent/BACKLOG.md

@@ -144,12 +144,17 @@ template source. Pass = Chromium launches after installation and HTTP/HTTPS
 mime defaults resolve to `chromium-browser.desktop` (and the intended live-ISO
 browser posture is explicit).
 
-### 95. Live ISO/install: Ghostty does not open — **two distinct failures**
+### 95. Live ISO/install: Ghostty does not open on Acer (old GL + two log shapes)
 
-Two photos, two machines, **two different Ghostty death modes**. Do not
+**Machine:** Acer Aspire M5-481T only. **Dell XPS 9350 starts Ghostty fine**
-collapse them into one "old GPU" fix; a fix for (A) can leave (B) broken.
+(Bernardo clarified 2026-07-15 — do not treat XPS as a Ghostty failure).
 
-**A — Acer Aspire M5-481T: OpenGL too old (explicit)**
+On the Acer, Bernardo installed **kitty** by hand (worked) and launched
+Ghostty from that shell twice — **same box, two lifecycle points, two log
+shapes**:
+
+**1 — Before Home Manager was properly built** (partial first boot after
+#123 bake failure). Explicit GPU floor:
 
 ```
 info(opengl): loaded OpenGL 4.2
@@ -160,11 +165,7 @@ warning(gtk_ghostty_surface): surface failed to initialize
   err=error.SurfaceError
 ```
 
-Ghostty 1.3.1 starts, loads config, hits the GPU floor, prints the reason,
+**2 — After home was built** (full HM generation). Silent-ish exit:
-returns to the shell (~2s). Anything ≤ GL 4.2 (Ivy Bridge era) is dead for
-the default terminal.
-
-**B — Dell XPS 9350 (photo without Acer bezel): silent exit after GDK_DISABLE**
 
 ```
 warning(gtk_ghostty_application): setting GDK_DEBUG=
@@ -172,29 +173,30 @@ warning(gtk_ghostty_application): setting GDK_DISABLE=gles-api,vulkan
 ~ 9s )
 ```
 
-Same Ghostty 1.3.1 / GTK path / config read — but **no** `info(opengl):`
+No `info(opengl):` line, no `OpenGLOutdated`, no `SurfaceError` — process
-line, **no** OpenGLOutdated, **no** SurfaceError. Process ends after ~9s
+ends after ~9s with only the two `GDK_*` warnings. Same Ghostty 1.3.1, same
-and leaves only the two `GDK_*` warnings. XPS Skylake iGPU is not a GL-4.2
+machine; the difference is **session/config state after HM**, not a second
-floor machine; this is a different surface/backend failure (forced desktop
+GPU. Likely still the GL 4.2 floor (or the same surface path) with different
-OpenGL after disabling GLES+Vulkan, Wayland/EGL context, incomplete first
+verbosity once the full Ghostty/GTK/Stylix config is in place — confirm
-session after #123, …). Capture a second run with `GDK_DEBUG` / Ghostty
+before inventing a second root cause. Re-run post-HM with extra logging
-`-e` verbosity or `WAYLAND_DEBUG=1` before guessing the patch.
+(`GDK_DEBUG`, Ghostty verbosity, `WAYLAND_DEBUG=1`) if the quiet path is
+still opaque.
 
-**Shared product impact:** default terminal is Ghostty (`nomarchy.terminal`),
+**Product impact:** default terminal is Ghostty (`nomarchy.terminal`), so on
-so SUPER+Return and the launcher entry are dead when either mode hits.
+this class of GPU SUPER+Return and the launcher entry are dead. Kitty as a
-Bernardo worked around on the Acer by adding `kitty` to `home.packages`.
+manual workaround proved the session/Wayland path is fine.
 
-**Fix direction (likely two slices):**
+**Fix direction:** treat the **GL ≥ 4.3 requirement** as the primary Acer
-1. **(A)** soft-fallback or alternate default when GL < 4.3 (probe, or
+blocker (soft-fallback, alternate default terminal when GL < 4.3, or
-   ship a second terminal and bind/fallback).
+installer/hardware probe that swaps the default). Do not leave SUPER+Return
-2. **(B)** reproduce on XPS-class hardware or QEMU with a capable GL stack;
+bound to a terminal that cannot map a surface. Optionally make the post-HM
-   fix the silent path (must never exit without a user-visible error or a
+failure mode print the OpenGL reason again so the quiet log is not a
-   working window). May interact with #123 if the first session is half-baked.
+support trap.
 
-**Pass:** SUPER+Return opens a usable terminal on **both** the Acer (A) and
+**Pass:** on the Acer (or a GL 4.2 / softGL stand-in), SUPER+Return opens a
-the XPS (B); modern-GL machines still get themed Ghostty by default; a
+usable terminal without hand-editing `home.packages`; modern-GL machines
-regression that only fixes A or only B is not done. V2 may cover (A) via
+(including the XPS) still get themed Ghostty by default. V2 softGL may cover
-softGL version pinning; (B) needs the XPS or a faithful GL/Wayland guest.
+the floor; V3 on the Acer is the real check.
 
 ### 98. Boreal: button text renders black and is difficult to read
 

agent/HARDWARE-QUEUE.md

@@ -436,13 +436,12 @@ the **T14s** (webcam case).
       **without** running `home-manager switch` by hand. **Fail** = still
       need the recovery one-liner, or GTK/nm-applet still wrong until a
       manual switch (then the bake is still incomplete).
-- [ ] **#95 Ghostty — two machines, two failure modes** — after the fix(es):
+- [ ] **#95 Ghostty on Acer M5-481T only** — XPS is fine (clarified
-      **(A) Acer M5-481T:** SUPER+Return opens a usable terminal (today:
+      2026-07-15). On the Acer after the fix: SUPER+Return opens a usable
-      OpenGL 4.2 → `OpenGLOutdated` / `SurfaceError`). **(B) Dell XPS 9350:**
+      terminal without hand-adding kitty. Today both attempts fail: (1)
-      SUPER+Return opens a usable terminal (today: silent exit after
+      pre-HM → explicit `OpenGL 4.2` / `OpenGLOutdated`; (2) post-HM →
-      `GDK_DISABLE=gles-api,vulkan`, no OpenGL line). A modern-GL box still
+      quiet exit after `GDK_DISABLE=gles-api,vulkan`. Confirm post-fix
-      gets themed Ghostty by default. **Both** A and B required — fixing only
+      both lifecycle points open a window. Modern-GL default remains Ghostty.
-      the GL floor is not done.
 - [ ] **#124 flake pin after main-ISO install** — on a machine installed
       from a **main-built** ISO (post-fix): `~/.nomarchy/flake.nix`
       nomarchy input must not re-resolve to a lagging `v1` that lacks

agent/JOURNAL.md

@@ -31,20 +31,22 @@ Template:
   2. **#124** — seeded flake hardcodes `?ref=v1` while installer writes
      main-era `nomarchy.hardware.*`; v1 lags (~hundreds of commits) →
      `nomarchy.hardware does not exist` on Acer rebuild/pull.
-  3. **#95** — Ghostty fails on **both** machines for **different** reasons
+  3. **#95** — Ghostty **Acer-only** (XPS is fine). Two photos = two
-     (corrected after Bernardo: two photos ≠ one diagnosis). **(A) Acer:**
+     attempts on the Aspire: pre-HM explicit `OpenGLOutdated`, post-HM
-     OpenGL 4.2 → `OpenGLOutdated`. **(B) XPS:** silent exit after
+     quiet exit after `GDK_DISABLE` (same machine; kitty used as the
-     `GDK_DISABLE=gles-api,vulkan` with no OpenGL line — not a GL-floor bug.
+     working shell). Primary blocker is GL ≥ 4.3.
-  HARDWARE-QUEUE V3 rows for all three. Photos stay local (not committed).
+  HARDWARE-QUEUE V3 rows. Photos stay local (not committed).
 - **Verified:** V0 (docs/agent only).
-- **Pending:** implement #123 first (unblocks first-boot polish); then
+- **Pending:** implement #123 first; then #124; #95 independent (Acer GL).
-  #124; #95 is independent and likely **two slices** (A vs B).
 - **Next suggestion:** #123 (installer bake) — highest blast radius.
 
-## 2026-07-15 — Correct #95: two Ghostty photos, two causes
+## 2026-07-15 — #95 timeline: both Ghostty fails are Acer, pre/post HM
-- **Task:** Bernardo noted two Ghostty-fail photos with different reasons.
+- **Task:** Bernardo corrected the Ghostty story again.
-- **Did:** #95 + HARDWARE-QUEUE rewritten: (A) Acer OpenGLOutdated explicit;
+- **Did:** XPS works. Both fail photos are Aspire M5-481T: (1) before home
-  (B) XPS silent exit after GDK_DISABLE — must not be folded into one fix.
+  properly built → OpenGL 4.2/`OpenGLOutdated`; (2) after home built →
+  silent exit after `GDK_DISABLE=gles-api,vulkan`. Kitty was the manual
+  working terminal used to run those experiments. #95 + HARDWARE-QUEUE
+  rewritten; not two machines.
 - **Verified:** V0 (docs only).
 - **Pending:** —
 - **Next suggestion:** still #123.