Mouse cursor renders extremely large inside the app

[maosuarez]

Description

The mouse cursor appears extremely large while inside the application window. Unclear yet whether this is a WSL/environment issue or caused by the app itself — needs investigation to confirm root cause.

Platform

Linux (possibly WSL-specific, needs confirmation)

[maosuarez]

Ran the diagnosis plan (docs/plans/issue-6-oversized-cursor-linux.md) as far as possible from a WSL2 (Ubuntu) session — no static analysis substitute exists for the live-desktop checks, and per this project's own CLAUDE.local.md, WSLg is explicitly not the supported verification path (the dev loop builds/runs on native Windows over the UNC path). So this comment covers only the parts of §3.1/§3.2 that don't require a real X11/Wayland compositor.

Static-only checks (§3.2 D5, app-side candidates) — all NEGATIVE:

• grep -rn "cursor:" src → no matches (no app CSS sets a cursor rule)
• grep -rniE "zoom|transform:\s*scale|image-rendering" src/index.css src/App.css src/App.tsx → no matches
• grep -rn "set_zoom" src-tauri/src/ → no matches (no Tauri webview zoom call anywhere)
• Custom cursor assets (.cur, .ani, CSS url(...)) → none; only match is the Google Fonts @import url(...) in src/index.css
• src-tauri/tauri.conf.json: window is decorations: false, resizable: false, fixed 560x700, no Linux-specific bundle config beyond deb.depends: []
• src-tauri/src/lib.rs run(): no env::set_var / cursor/GTK-related calls anywhere, before or after tauri::Builder::default() — so today's binary does not touch XCURSOR_*/GDK_* at all (B-ENV isn't implemented; nothing to regress)

Net: this rules out D5(a–d) (app-side CSS/zoom/asset causes) as a false lead. It does not produce a verdict — D1 (text oversized?), D2 (control-app comparison), D3 (theme resolution), D4 (scale-factor mismatch), and D6 (native X11 + Wayland repro) all require an actual running app on a real compositor session, which isn't available here.

Phase A verdict: insufficient evidence, blocked at §3.0. Not claiming H1/H2/H3/H4/H5 — that would be guessing dressed as diagnosis, which the plan explicitly warns against (§4.3, "guard against over-correcting"). Labeling needs-linux-verifier per §3.0's stated fallback. No Phase B code changes made.

Added a short "known issue, cause unconfirmed" note to README.md's Linux section pointing back here — nothing more, since a confirmed cause doesn't exist yet to document.