cdrom: cancel stale INT2 on new command (PSX-CD-007)

Cancel the previous command's outstanding second-response interrupt when
a new CD command is accepted (psx-spx command flow; DuckStation
BeginCommand semantics). Without this, a Pause(0x09) INT2 survives into
the guest's next Setmode/Setloc/ReadN sequence and fires mid-read as a
stale COMPLETE — LIBDS drivers treat it as a read failure, corrupting
the streamed resource chain.

Cross-validated: C-12 Final Resistance mission 1 PC=0 null-callback
termination disappeared after absorbing this fix (DEVLOG 2026-08-05).
MOHU Mission 1 freeze in a DsRead FATAL/retry loop on single-sector
marker reads.

Also guard beetle SIGSTOP trap behind #ifndef _WIN32 (Windows has no
SIGSTOP).

(cherry picked from commit 2c43fc69c66242a86183156a6f1e60b2f81b1e47)
(cherry picked from commit c03fe5ba3d49241dfaedb344db6ad9c977070722)

runtime/src/beetle_libretro.cpp

@@ -75,7 +75,9 @@ static void cdcmd_trace_callback(uint8_t cmd, uint8_t nargs,
                                  uint32_t pc) {
     /* Symmetric self-stop trap (mirror of the runtime's PSX_CD_TRAP_CMD):
      * SIGSTOP on the Nth matching CD command so a debugger can freeze the
-     * oracle at the exact boot point and diff its state against ours. */
+     * oracle at the exact boot point and diff its state against ours.
+     * POSIX-only, same guard as debug_server.c — Windows has no SIGSTOP. */
+#ifndef _WIN32
     {
         static long trap_cmd = -2, trap_nth = 1, hits = 0;
         if (trap_cmd == -2) {
@@ -86,6 +88,7 @@ static void cdcmd_trace_callback(uint8_t cmd, uint8_t nargs,
         }
         if ((long)cmd == trap_cmd && ++hits == trap_nth) raise(SIGSTOP);
     }
+#endif
     BeetleCdcmdEntry *e = &s_cdcmd_trace[s_cdcmd_idx];
     e->seq = s_cdcmd_seq++; e->pc = pc; e->cmd = cmd; e->nargs = nargs;
     e->a0 = a0; e->a1 = a1; e->a2 = a2;

runtime/src/cdrom.c

@@ -264,6 +264,7 @@ static uint8_t cd_muted;
 #define XA_SUBMODE_REALTIME 0x40
 #define CDROM_SECTOR_MODE2 0x02
 
+
 #define CDROM_SKIP_NONE 0
 #define CDROM_SKIP_XA_AUDIO_REALTIME 1
 #define CDROM_REQUEST_BFRD 0x80u
@@ -2004,6 +2005,26 @@ static void exec_command(uint8_t cmd) {
     trace_cdrom('C', 0, cmd, 0);
     /* ENQUEUE: a CD command was issued (aux = command byte). */
     event_ring_record_aux(EV_ENQ, (uint8_t)SRC_CD_CMD, (uint32_t)cmd);
+    /* A new command CANCELS the previous command's outstanding second
+     * response (psx-spx command flow; DuckStation BeginCommand "command
+     * cancellation"). The sub-CPU runs one transaction at a time: once a
+     * new command is accepted, the superseded command's INT2 is never
+     * delivered. Two-phase handlers below already overwrite `pending`;
+     * this clears it for single-phase commands (GetStat/Setmode/Setloc/…)
+     * and for ReadN/ReadS, which do not use `pending` at all.
+     *
+     * Without this, a Pause(0x09) INT2 scheduled ~1.1M cycles out (see
+     * pause_complete_delay_cycles) survives into the guest's next
+     * Setmode/Setloc/ReadN sequence and fires MID-READ as a stale
+     * COMPLETE. LIBDS drivers treat that as a read failure: MOHU Mission 1
+     * wedged in a phase-locked DsRead FATAL/retry loop on its single-sector
+     * marker reads, corrupting the streamed resource chain
+     * (medal-of-honor-underground DEVLOG 2026-08-05). GT1 race loading
+     * shows the same stale-Pause overlap during ReadS streaming. */
+    if (pending.pending) {
+        trace_cdrom('X', 0, pending.cmd, 0);
+        pending.pending = 0;
+    }
     /* Snapshot the param FIFO before handlers consume it, so the
      * command-history record at the end sees the original params. */
     uint8_t cmd_params[PARAM_FIFO_SIZE];