DMR-данные (SMS) на TYT MD-UV380 под OpenGD77

Восстановлено из транскрипта сессии после потери рабочего дерева на сервере сборки.

Что внутри:
- четыре новых модуля прошивки (приём, сборка исходящего, меню Messages);
- правки существующих файлов в виде скриптов, с порядком применения;
- инструменты стенда: снятие экрана рации по USB, дамп кадров, эталонная сборка пакета;
- NOTES.md: что проверено в эфире, что не доделано, где грабли.

Работало в эфире в августе 2026: данные ходят в обе стороны в режиме хотспота,
рация принимает SMS и показывает на экране. Отправка собрана и сверена с эталоном
сервера побайтово, но в эфире не проверялась.

Claude-Session: https://claude.ai/code/session_01GJtvG5WwqYNe1b7s9LLtgy

patches/patch_diag4.py

@@ -0,0 +1,72 @@
+#!/usr/bin/env python3
+"""Expose the SMS diagnostics through a RAM window that can be read over USB.
+
+Drawing counters on the screen turned out to be unreliable: the notification is
+overwritten by the normal screen redraw, so a screenshot usually shows nothing even
+when reception is happening. Instead the counters are mirrored into the tail of the
+raw audio buffer, which the CPS protocol can read out at any moment (area 7).
+
+Layout at rawBuffer[2376..2399]:
+    'S' 'M' 'S' 'D'      magic, so a stale buffer cannot be mistaken for data
+    u16 bursts           data bursts handed to the assembler
+    u16 headers          data headers parsed
+    u16 assembled        messages assembled completely
+    u16 lastBlocks       block count announced by the last header
+    u16 collected        blocks collected for the message in progress
+    u16 state            0 = idle, 1 = collecting
+    u16 textLen          length of the last extracted text
+"""
+W = "/home/viktor/ogd77rel/uv_src/OpenGD77_MDUV380_DM1701_20260130/MDUV380_firmware/application"
+
+p = W + "/source/functions/sms.c"
+s = open(p).read()
+
+if "SMS_DIAG_OFFSET" in s:
+    print("already patched")
+    raise SystemExit(0)
+
+s = s.replace('#include "functions/ticks.h"\n',
+              '#include "functions/ticks.h"\n#include "functions/sound.h"\n', 1)
+
+anchor = "void smsInit(void)\n"
+assert anchor in s
+helper = """// Diagnostics mirror: the tail of the raw audio buffer is readable over USB (CPS area 7),
+// which is far more dependable than trying to keep text on the screen.
+#define SMS_DIAG_OFFSET  2376
+
+static void smsDiagPublish(void)
+{
+	volatile uint8_t *d = &audioAndHotspotDataBuffer.rawBuffer[SMS_DIAG_OFFSET];
+	uint16_t values[7];
+
+	values[0] = diagBursts;
+	values[1] = diagHeaders;
+	values[2] = diagAssembled;
+	values[3] = diagLastBlocks;
+	values[4] = blocksCollected;
+	values[5] = (uint16_t)rxState;
+	values[6] = diagTextLen;
+
+	d[0] = 'S'; d[1] = 'M'; d[2] = 'S'; d[3] = 'D';
+
+	for (uint8_t i = 0; i < 7; i++)
+	{
+		d[4 + (i * 2)] = (uint8_t)(values[i] & 0xFF);
+		d[5 + (i * 2)] = (uint8_t)(values[i] >> 8);
+	}
+}
+
+"""
+s = s.replace(anchor, helper + anchor, 1)
+
+# track the extracted text length too
+s = s.replace("static uint8_t           diagLastBlocks = 0;\n",
+              "static uint8_t           diagLastBlocks = 0;\nstatic uint16_t          diagTextLen = 0;\n", 1)
+s = s.replace("\tlatest.text[t] = '\\0';\n",
+              "\tlatest.text[t] = '\\0';\n\tdiagTextLen = t;\n", 1)
+
+# publish on every tick, so a read over USB always sees the current state
+s = s.replace("void smsTick(void)\n{\n", "void smsTick(void)\n{\n\tsmsDiagPublish();\n\n", 1)
+
+open(p, "w").write(s)
+print("sms.c: diagnostics mirrored into the readable RAM window")