fix: address unresolved PR #860 review feedback

CHANGELOG.md

@@ -186,6 +186,48 @@ The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.1.0/).
 
 ### Fixed
 
+- **Token balance refresh status**: requesting a refresh while token balances
+  are already updating now shows a brief informational note instead of a red
+  error banner that must be dismissed.
+
+- **Topping up an identity from more than one funding wallet no longer gets
+  stuck loading**: requesting a new deposit address at the same time as
+  refreshing your asset-lock transactions could silently drop the refresh,
+  leaving those wallets stuck showing "Loading" for the rest of the session.
+  Both requests are now sent together.
+
+- **A single damaged transaction record no longer hides an entire wallet**:
+  previously, one unreadable entry in a wallet's transaction history could
+  make the wallet and its balance disappear from every screen until the
+  underlying data was manually repaired. The app now skips the damaged entry
+  and keeps the wallet visible, with the rest of its history intact.
+
+- **Scheduled DPNS vote sweeps no longer get stuck after an unexpected error**:
+  an internal failure while casting due votes could permanently block future
+  vote sweeps for that network until the app was restarted. It now recovers
+  on its own.
+
+- **Send confirmations now show what the recipient actually receives**: when
+  the network fee is deducted from the amount you entered, the confirmation
+  dialog now says so and shows the reduced amount the recipient will get,
+  instead of implying they receive the full entered amount.
+
+- **A missed deposit no longer leaves the funding screen waiting forever**:
+  if the app missed the notification that your deposit arrived (for example
+  because you were on another screen), it now also checks your wallet
+  balance directly, so the funding step advances even if the one-time
+  notification was missed.
+
+- **Deposit-address screens no longer strand you after an address error**: if
+  generating a new receive address failed, the screen used to reset to a bare
+  view with no way to retry. It now keeps the retry button available.
+
+- **A network preference that can't be restored no longer defaults you to
+  Mainnet**: if the app can't confirm your previous network selection during
+  an upgrade, it now asks you to choose a network explicitly instead of
+  silently starting the session on Mainnet — important if you were previously
+  using Testnet.
+
 - **Submitted Platform actions are no longer reported as rejected when only
   confirmation failed**: if a state transition was broadcast but its result
   could not be confirmed, the app now tells you to check whether it completed

docs/gui-testing/README.md

@@ -25,12 +25,106 @@ network timing, or a flow `kittest` can't simulate.
 
 ## How to run a scenario
 
-The mechanical launch recipe (X display setup, accessibility tree, gotchas
-like stderr redirection) lives in the global `desktop-gui` Claude Code skill
-(`~/.claude/skills/desktop-gui/SKILL.md`) — read that first, it's not repeated
-here. This directory covers what's specific to *this project*: which
-scenarios exist, what credentials they need, and the safety rules for running
-them against a live network.
+Use a graphical session whose `DISPLAY` already points at the desktop you want
+to observe. Do not assume a display number: local desktops, SSH forwarding,
+CI, and headless X servers all use different values. Verify the selected
+display before launching:
+
+```bash
+: "${DISPLAY:?Set DISPLAY to the desktop used for GUI testing}"
+xdpyinfo >/dev/null
+```
+
+On a minimal Ubuntu host, egui/eframe also needs an X keyboard library and a
+wgpu backend. Install missing packages only after the launch log identifies
+the corresponding failure:
+
+```bash
+sudo apt-get install -y libxkbcommon-x11-0 mesa-vulkan-drivers xdotool
+```
+
+Build in the checkout under test, then ask Cargo for the effective target
+directory. This honors `CARGO_TARGET_DIR` and any `target-dir` configured in
+Cargo's configuration files.
+
+```bash
+cargo build
+TARGET_DIR=$(cargo metadata --format-version 1 --no-deps | \
+  python3 -c 'import json, sys; print(json.load(sys.stdin)["target_directory"])')
+BIN="$TARGET_DIR/debug/dash-evo-tool"
+test -x "$BIN"
+```
+
+Prepare isolated state and launch the binary as a detached process. The app
+redirects some diagnostics into its data directory, so inspect both the launch
+log and `det-stderr.log` / `det.log` after a crash or panic.
+
+```bash
+DATADIR=$(mktemp -d)
+cp .env.example "$DATADIR/.env"
+LOG="$DATADIR/gui-test-launch.log"
+
+pgrep -af dash-evo-tool
+DASH_EVO_DATA_DIR="$DATADIR" nohup "$BIN" >"$LOG" 2>&1 &
+APP_PID=$!
+
+pgrep -af "$BIN"
+WID=$(xdotool search --pid "$APP_PID" | head -1)
+xdotool getwindowgeometry "$WID"
+xdotool windowsize "$WID" 1260 780
+xdotool windowactivate "$WID"
+```
+
+### Accessibility tree
+
+Dash Evo Tool publishes its AccessKit tree over AT-SPI2 when accessibility is
+enabled. A headless Ubuntu session needs the accessibility bus and Python
+bindings:
+
+```bash
+sudo apt-get install -y at-spi2-core python3-pyatspi
+export DBUS_SESSION_BUS_ADDRESS="unix:path=/run/user/$(id -u)/bus"
+
+dbus-send --session --print-reply \
+  --dest=org.a11y.Bus /org/a11y/bus org.a11y.Bus.GetAddress
+gdbus call --session --dest org.a11y.Bus --object-path /org/a11y/bus \
+  --method org.freedesktop.DBus.Properties.Set \
+  org.a11y.Status ScreenReaderEnabled '<true>'
+gdbus call --session --dest org.a11y.Bus --object-path /org/a11y/bus \
+  --method org.freedesktop.DBus.Properties.Set \
+  org.a11y.Status IsEnabled '<true>'
+```
+
+Add `DASH_EVO_TOOL_ACCESSIBILITY=1` to the launch command, focus the window
+with `xdotool windowactivate "$WID"`, and inspect the tree with any AT-SPI
+client. This self-contained Python example prints roles and labels:
+
+```bash
+python3 - <<'PY'
+import pyatspi
+
+
+def walk(node, depth=0):
+    print(f"{'  ' * depth}{node.getRoleName()}: {node.name}")
+    for child in node:
+        walk(child, depth + 1)
+
+
+for app in pyatspi.Registry.getDesktop(0):
+    if app.name == "dash-evo-tool":
+        walk(app)
+PY
+```
+
+An empty application list normally means the AT-SPI status flags are still
+disabled or the app window is not focused. The tree can lag during screen
+transitions and omits purely decorative visuals, so use it for semantic labels
+and structure while using screenshots for pixels, colors, and final visual
+confirmation. Capture durable screenshots with `scrot -o <path>.png`.
+
+The complete recipe is versioned here because an installed `desktop-gui`
+automation skill is not available to every contributor. When present, that
+skill may still provide convenient screenshot and input tooling.
 
 ## Non-negotiable safety rules
 
@@ -138,8 +232,8 @@ one scenario silently invalidating another:
 ## Known UI/environment quirks
 
 - **Default window is small (800×600) and clips controls** (sidebar items,
-  settings sections below the fold). Resize immediately after launch — see the
-  `desktop-gui` skill's launch recipe. Some settings sections are collapsible
+  settings sections below the fold). Resize immediately after launch with the
+  `xdotool windowsize` command above. Some settings sections are collapsible
   *and* below the fold even after resizing: expect to expand a section, then
   scroll, before a control becomes visible — don't conclude a control doesn't
   exist from the first screenshot after expanding.
@@ -151,12 +245,11 @@ one scenario silently invalidating another:
   and dismiss it immediately. If a dialog flashes shut the instant it opens,
   suspect this pattern before assuming a mis-click — take a screenshot a frame
   later and retry with the click and the opening action clearly separated.
-- **The shared `/data/target` build output is not campaign-exclusive.** If
-  other worktrees/sessions on the same box can rebuild concurrently, the
-  binary under test can be silently overwritten mid-campaign by an unrelated
-  build. For any run spanning hours, build to a private path and hash-verify
-  (`sha256sum`) before each relaunch rather than trusting the shared path
-  throughout.
+- **A shared Cargo target directory is not campaign-exclusive.** If other
+  worktrees/sessions on the same box can rebuild concurrently, the binary under
+  test can be silently overwritten mid-campaign by an unrelated build. For any
+  run spanning hours, set `CARGO_TARGET_DIR` to a private path before building
+  and hash-verify (`sha256sum`) before each relaunch.
 
 ## Scenario index
 

docs/gui-testing/scenarios/TEMPLATE.md

@@ -24,7 +24,12 @@ cp .env.example "$DATADIR/.env"
 # Confirm no conflicting instance is already using this display/data dir
 pgrep -af dash-evo-tool
 
-DISPLAY=:99 DASH_EVO_DATA_DIR="$DATADIR" nohup /data/target/debug/dash-evo-tool >/tmp/<scenario-slug>.log 2>&1 &
+TARGET_DIR=$(cargo metadata --format-version 1 --no-deps | \
+  python3 -c 'import json, sys; print(json.load(sys.stdin)["target_directory"])')
+BIN="$TARGET_DIR/debug/dash-evo-tool"
+test -x "$BIN"
+LOG="$DATADIR/<scenario-slug>.log"
+DASH_EVO_DATA_DIR="$DATADIR" nohup "$BIN" >"$LOG" 2>&1 &
 ```
 
 ## Procedure