FABRIC-2.md §I.9: fix the terminal phantom-linebreak defect

console_ensure_line_start() (hal/console.c) used to emit its newline
immediately via a path that deliberately skipped the tx-byte counter, so
that sk_repl_idle()'s unconditional per-beat call to it (repl.c, ~1s idle
heartbeat) could force a real newline the REPL's own prompt-reanchor logic
never noticed -- the prompt was never reprinted, and the next real
keystroke echoed onto the now-blank line, indistinguishable from Enter
having already been pressed at a bare prompt. Root-caused in the previous
commit (704573b); this commit applies the fix per explicit go-ahead.

Fix: defer the newline instead of emitting it eagerly.
console_ensure_line_start() now only sets a flag (g_pending_line_close);
the newline is realized -- for real, and counted by g_console_tx_count
like any other output -- on the next actual console_putc() call, or
silently discarded via the new console_cancel_deferred_line_start() if the
caller decides nothing was actually printed. sk_repl_idle() captures
tx_before_idle right after its console_ensure_line_start() call and cancels
the deferred newline at both of its exit points when console_tx_count()
hasn't moved. A beat with nothing to report now leaves the console
untouched; a beat that does print still closes the dangling prompt line
first, properly counted this time. The pre-existing n > 0 mid-edit gate in
sk_console_readline() is untouched -- independent purpose, not the bug.

Verified via the mandatory foreground 3-arch QEMU acceptance boot
(clean qemu, amd64 -> aarch64 -> riscv64, one at a time): all three reached
(zuse) ok>, all echoed the first real input on the same log line as the
prompt rather than a fresh line, all shut down cleanly via BYE. amd64's log
additionally shows live human backspace-correction still glued to the same
prompt line. Logs: logs/20260905-015854 (amd64), logs/20260905-020248
(aarch64), logs/20260905-020552 (riscv64); logs/20260905-015813 is a
foreground-rule-violation retry killed and redone correctly, kept per this
project's "never delete logs/" convention.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_019YcT3H2PQeyujrzjqS3Var

src/starkernel/repl.c

@@ -174,8 +174,18 @@ static void sk_repl_idle(VM *active_vm)
      * the prompt and inherit no {VMName} prefix (g_line_start is 0). A
      * fresh line first keeps every idle line prefix-tagged and readable,
      * matching what an interactive typing session expects. No-op when the
-     * console is already at a line boundary. */
+     * console is already at a line boundary.
+     *
+     * FABRIC-2.md §I.9 fix, 2026-09-05: this newline is now deferred (see
+     * console_ensure_line_start()'s own doc comment) -- it only actually
+     * reaches the console if something below really prints. tx_before_idle
+     * lets this function tell "nothing happened" apart from "something did"
+     * the same way the reanchor check further up this file already does,
+     * so a beat with nothing to report can cancel the deferred newline
+     * before returning, leaving the bare prompt line completely untouched
+     * instead of visibly snapping it to a fresh blank line every ~1s. */
     console_ensure_line_start();
+    uint64_t tx_before_idle = console_tx_count();
 
     /* Artemis Milestone 2d: xHCI Event Ring servicing. This is exactly the
      * "interrupt-driven, coarse cadence, cheap early-exit" trigger Section
@@ -357,8 +367,12 @@ static void sk_repl_idle(VM *active_vm)
      * a live parse truncates the rest of that line. Deferring the MSG-TICK
      * drain to the next prompt boundary is safe -- draining is best-effort
      * and simply resumes next beat. */
-    if (g_mama_interpreting || g_idle_pump_active)
+    if (g_mama_interpreting || g_idle_pump_active) {
+        if (console_tx_count() == tx_before_idle) {
+            console_cancel_deferred_line_start();
+        }
         return;
+    }
 
     g_idle_pump_active = 1;
     {
@@ -386,6 +400,10 @@ static void sk_repl_idle(VM *active_vm)
         }
     }
     g_idle_pump_active = 0;
+
+    if (console_tx_count() == tx_before_idle) {
+        console_cancel_deferred_line_start();
+    }
 }
 
 /*===========================================================================
@@ -596,17 +614,24 @@ int sk_console_readline(char* buf, int size, VM* active_vm, int reanchor_prompt)
 
             uint64_t now = heartbeat_ticks();
             /* n == 0 gate: sk_repl_idle() opens with console_ensure_line_start(),
-             * which unconditionally forces a newline whenever the console isn't
-             * already at a line boundary -- including mid-edit, with characters
-             * already typed and echoed (n > 0) but Enter not yet pressed. Without
-             * this gate, that forced break fires on every elapsed
-             * SK_IDLE_BEAT_INTERVAL regardless of whether sk_repl_idle() actually
-             * has anything to print, visually snapping the in-progress input line
-             * to a fresh (empty) line -- indistinguishable from Enter having been
-             * pressed. Deferring the whole idle beat while a line is being edited
-             * (same n > 0 guard the prompt reanchor below already uses) delays
-             * xhci/block-sync servicing by at most one more interval, which its
-             * own "coarse cadence, cheap early-exit" design already tolerates. */
+             * closing off a dangling prompt line before any chatter it might
+             * print (xhci attach/detach, block-sync, MSG-TICK pump). Before the
+             * FABRIC-2.md §I.9 fix (2026-09-05), that newline was unconditional
+             * and immediate, so it fired on every elapsed SK_IDLE_BEAT_INTERVAL
+             * regardless of whether sk_repl_idle() actually had anything to
+             * print -- including mid-edit (n > 0, characters typed but Enter
+             * not yet pressed), visually snapping the in-progress line to a
+             * fresh blank one, indistinguishable from Enter having been
+             * pressed. console_ensure_line_start()'s newline is now deferred
+             * and self-cancelling when nothing follows it (console.c), which
+             * fixes that regardless of n -- but this gate is kept for its own,
+             * independent reason: deferring the *whole* idle beat while a line
+             * is being edited (same n > 0 guard the prompt reanchor below
+             * already uses) means a genuine xhci/block-sync/MSG-TICK event
+             * cannot interrupt output mid-line while the user is actively
+             * typing, only delaying that servicing by at most one more
+             * interval, which its own "coarse cadence, cheap early-exit"
+             * design already tolerates. */
             if (n == 0 && now - g_last_beat_tick >= SK_IDLE_BEAT_INTERVAL) {
                 g_last_beat_tick = now;
                 sk_repl_idle(active_vm);