@bobfrankston/rmfmail 1.1.95 → 1.1.97

package/client/lib/rmf-tiny.js

@@ -326,6 +326,15 @@ export async function createTinyMceEditor(container, opts = {}) {
                 editor.selection.select(editor.getBody(), true);
                 editor.selection.collapse(pos === 0 /* true = start */);
                 editor.focus();
+                // Scroll the viewport to match the caret. For pos 0 that's
+                // the TOP of the body — on a reply the user types above the
+                // quoted block and must land looking at that empty line, not
+                // scrolled down inside the quote (Bob 2026-05-18: "the cursor
+                // should be at the top of the window").
+                if (pos === 0)
+                    editor.getWin()?.scrollTo(0, 0);
+                else
+                    editor.selection.scrollIntoView();
             }
             catch { /* */ }
         },

package/docs/rmf-tiny.md

@@ -1,6 +1,6 @@
-# `rmf-tiny` — optional TinyMCE editor adapter for rmfmail
+# `rmf-tiny` — TinyMCE editor adapter for rmfmail
 
-Status: design only. Not implemented.
+Status: **implemented and shipping.** Source: `y:/dev/utils/msgx/libs/rmf-tiny/src/adapter.ts`; built copy served as `client/lib/rmf-tiny.js` (via `bin/build-rmf-tiny.js`). Selectable in Settings → Compose → Editor. The design notes below are kept for history; see **Known TinyMCE quirks** at the bottom for the live gotcha list.
 
 ## Goal
 
@@ -154,3 +154,32 @@ Don't publish the adapter until the Word-paste case actually works end-to-end th
 - The adapter must not re-implement parts of TinyMCE — it's a thin wrapper. Re-implementing parts of GPL code in MIT is the legal trap; staying purely in adapter / interface territory is safe.
 - TinyMCE's bundle is large (~1MB minified). Users opting in pay that cost on the first compose-window load. Fine for opt-in; would be unacceptable as a default.
 - `rmf-tiny` major version should track TinyMCE's major version (rmf-tiny@7.x for tinymce@7.x) so peerDependency mismatches are obvious.
+
+## Known TinyMCE quirks
+
+TinyMCE frequently gets caret / scroll / block behaviour wrong. Each entry
+below is a quirk found in use and the workaround in `adapter.ts`. Add to
+this list whenever a new one turns up — it's the running gotcha log.
+
+- **Viewport doesn't follow `setCursor`.** After `setContent` of a long
+  reply, collapsing the selection to the start (caret at top, where the
+  user types above the quote) does NOT scroll the iframe — it can be left
+  scrolled down inside the quoted block. Fix: `setCursor(0)` explicitly
+  calls `editor.getWin().scrollTo(0, 0)`; non-zero positions call
+  `editor.selection.scrollIntoView()`. (Bob 2026-05-18.)
+
+- **Typing at a link's trailing edge extends the link.** With the caret
+  collapsed at the end of an `<a>`, contenteditable keeps appending typed
+  characters INTO the link — a whole sentence becomes link text. Fix: a
+  `keypress` handler hops the caret to just after the `<a>` first.
+
+- **Native spellcheck is force-enabled.** `adapter.ts`'s `init` handler
+  sets `spellcheck="true"` on the body. mailx runs its own nspell checker
+  and overrides this back to `"false"` in `wireSpellcheck` (with a
+  MutationObserver) so the two don't double-underline.
+
+- **`<p>` collapses to `<br>`.** Enter / paste sometimes produces a `<br>`
+  line break where a `<p>` paragraph was expected, so paragraph spacing is
+  lost until the lines are merged and Enter re-pressed. Not yet fixed —
+  candidate: `forced_root_block` / paste-normalization config. (Bob
+  2026-05-18, exact trigger still unclear.)

package/npmchanges.md

@@ -274,3 +274,29 @@ strip now stays visible with a single tab, showing that tab plus the
 "+" button; its tooltip names the Ctrl+T shortcut. Hidden only before
 the very first tab is seeded.
 
+## v1.1.95 — 2026-05-19
+
+Compose window opens centred near the top, not lower-right
+
+The floating compose/reply overlay was anchored to the bottom-right
+corner. It now opens horizontally centred with a small top margin. The
+title bar still drags it anywhere; this only changes initial placement.
+
+## v1.1.96 — 2026-05-19
+
+Alarm/snooze: log decisions to the daemon log so snooze is verifiable
+
+On review the snooze keying is actually correct — snoozeItem stores under
+item.uuid, which IS the full per-occurrence key (uuid:startMs@offset for
+calendar, uuid:dueMs for tasks) that collectDueAlarms reads back as
+snoozed[key]. They match.
+
+What was missing was visibility: the [alarm] decisions used plain
+console.log, which from the WebView never reaches the daemon log file —
+so "is snooze working?" couldn't be answered after the fact. Added an
+alog() helper that mirrors to the daemon log via logClientEvent, a
+"snooze saved" line (key + minutes + untilMs), and routed the existing
+fire-decision / dismiss / unknown-result logs through it. Next snooze is
+now traceable: the "snooze saved" line and the following "fire decision"
+line confirm the snooze took and for how long.
+

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@bobfrankston/rmfmail",
-  "version": "1.1.95",
+  "version": "1.1.97",
   "description": "Local-first email client with IMAP sync and standalone native app",
   "type": "module",
   "main": "bin/mailx.js",