amyboard: report full sketch traceback (line numbers) to web editor (#1059)

When a sketch fails to load, run_sketch() reported only
"<ExcType>: <message>" back to the web editor over the 'X' sysex frame —
the full traceback (with file/line numbers) was generated by
sys.print_exception() but only went to the serial console, never the
website. In simulate mode the web already shows the full traceback (it
scrapes it off the JS console), so this gap only affected control mode
on real hardware.

Capture sys.print_exception() into an io.StringIO and report that
instead, via a new _format_exc_for_report() helper. Kept as a single
sysex frame (no chunking): real MicroPython tracebacks are ~150-300
bytes and a single frame reliably carries ~768 raw bytes over Web MIDI
(ZDUMP_STREAM_RAW_CHUNK in amy/src/transfer.c). The helper bounds output
to 700 bytes, keeping the tail (MicroPython prints most-recent-call-last,
so the exception line is last) and falling back to the old one-liner if
capture ever fails. The web display already renders multi-line text, so
no web-side change is needed.

The sketch.loop() runtime-error path is intentionally left serial-only:
it runs every ~60ms and reporting from there would spam sysex frames.

Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>

Author: bwhitman

tulip/shared/amyboard-py/amyboard.py

@@ -512,6 +512,31 @@ def start_amy():
 
 _sketch_seq = None  # Keep reference to prevent GC
 
+# Cap on the traceback text we ship in a single 'X' sysex frame. One sysex
+# frame is reliably delivered over Web MIDI up to ~768 raw bytes (see
+# ZDUMP_STREAM_RAW_CHUNK in amy/src/transfer.c); we stay comfortably under that
+# after base64 expansion. Real MicroPython sketch tracebacks are ~150-300 bytes,
+# so this only ever trims pathologically deep stacks.
+_SKETCH_ERROR_MAX = 700
+
+def _format_exc_for_report(e):
+    """Full traceback (with file/line numbers) as a string, bounded so it fits
+    in a single error sysex frame. MicroPython prints most-recent-call-last, so
+    the exception type/message is the LAST line — when we must truncate we keep
+    the tail (the part the user needs) and drop the outermost frames."""
+    try:
+        from io import StringIO
+        buf = StringIO()
+        sys.print_exception(e, buf)
+        text = buf.getvalue().strip()
+    except Exception:
+        # If capture fails for any reason, fall back to the old one-liner so the
+        # web still gets *something* rather than nothing.
+        text = "%s: %s" % (type(e).__name__, e)
+    if len(text) > _SKETCH_ERROR_MAX:
+        text = "...(traceback truncated)\n" + text[-_SKETCH_ERROR_MAX:]
+    return text
+
 def _report_sketch_error(detail):
     """Push a sketch load error back to the web editor over sysex, framed as
     F0 00 03 45 'X' <base64 text> F7, so the UI can surface the failure (banner +
@@ -570,7 +595,7 @@ def run_sketch():
         # Route through stderr_write so the web console sees it alongside the
         # other diagnostics (stdout may not be wired up in some hosts).
         tulip.stderr_write("sketch.py load failed: %s: %s" % (type(e).__name__, e))
-        _report_sketch_error("%s: %s" % (type(e).__name__, e))
+        _report_sketch_error(_format_exc_for_report(e))
         try:
             sys.print_exception(e)
         except Exception:
@@ -588,7 +613,7 @@ def run_sketch():
             import sketch
         except Exception as e2:
             tulip.stderr_write("default sketch.py load also failed: %s: %s" % (type(e2).__name__, e2))
-            _report_sketch_error("%s: %s" % (type(e2).__name__, e2))
+            _report_sketch_error(_format_exc_for_report(e2))
             try:
                 sys.print_exception(e2)
             except Exception: