fix: resolve rendering issues and incorrect page splitting in T42 loader

src/teletext/io.py

@@ -107,21 +107,18 @@ def load_t42(file_path: str, progress_callback: Optional[Callable[[int, int], No
                 # Robustness check for VHS captures:
                 # If we see a row number that has already passed (e.g. Row 1 after Row 25)
                 # AND we didn't see a Row 0, it means a new page started but we missed the header.
-                # We should split into a new Page object to avoid data corruption.
+                # We only split on body rows (1-25) to avoid extension packets (26-31) triggering splits.
-                if target_page and row <= prev_row and row != prev_row: # Strictly less than (or handle duplicate rows?)
+                if target_page and 1 <= row <= 25 and row < prev_row:
-                    # In some captures, we might see the same row twice (Field 1/2).
-                    # If it's the SAME row number, we just append (overwrites in renderer).
-                    # If it's a LOWER row number, it's definitely a new cycle.
-                    
                     # Create a "Lost Header" page
-                    # We use page_number=0xFF to indicate unknown, but we keep mag.
                     target_page = Page(magazine=mag, page_number=0xFF, sub_code=0, language=0)
                     service.pages.append(target_page)
                     current_pages_by_mag[mag] = target_page
                 
                 if target_page:
                     target_page.packets.append(packet)
-                    last_row_by_mag[mag] = row
+                    # Only track body rows for sequence reversal detection
+                    if 1 <= row <= 25:
+                        last_row_by_mag[mag] = row
                 else:
                     # Packet without a header? Orphaned. Just keep in all_packets
                     pass
@@ -266,9 +263,8 @@ def parse_header(data: bytearray):
     
     sub_code = s1 | (s2 << 4) | (s3 << 8) | (s4 << 12)
     
-    # Control bits C12, C13, C14 are in Byte 8 (index 8)
+    # Control bits C12, C13, C14 are in Byte 13 (index 7 of data)
-    # They determine the National Option (Language)
+    c_bits_2 = decode_hamming_8_4(data[7])
-    c_bits_2 = decode_hamming_8_4(data[8])
     
     # Fix for Language Detection:
     # It seems C12 and C13 are swapped in the Hamming decoding or file format relative to expected values.

src/teletext/renderer.py

@@ -69,10 +69,10 @@ class TeletextCanvas(QWidget):
         self.flash_timer.timeout.connect(self.toggle_flash)
         self.flash_timer.start(500) # 500ms toggle (1Hz flash rate)
 
-        # Teletext is 40 columns x 25 rows
+        # Teletext is 40 columns x 26 rows (0-25)
         # We will render to a fixed size QImage and scale it
         self.cols = 40
-        self.rows = 25
+        self.rows = 26
         self.cell_w = 20
         self.cell_h = 24
         self.img_w = self.cols * self.cell_w
@@ -237,13 +237,13 @@ class TeletextCanvas(QWidget):
                 
         # Pass 1: Backgrounds
         occlusion_mask = [False] * 40
-        for r in range(25):
+        for r in range(26):
             packet = grid[r]
             occlusion_mask = self.draw_row(painter, r, packet, draw_bg=True, draw_fg=False, occlusion_mask=occlusion_mask)
 
         # Pass 2: Foregrounds
         occlusion_mask = [False] * 40
-        for r in range(25):
+        for r in range(26):
             packet = grid[r]
             occlusion_mask = self.draw_row(painter, r, packet, draw_bg=False, draw_fg=True, occlusion_mask=occlusion_mask)