The GDB debug script tools/gdb/gdb_scripts/wt_debug_script_update.py mishandles WT_UPDATE_TOMBSTONE entries in dump_update_chain().

Problem

val_bytes = gdb.selected_inferior().read_memory(wt_val['data'], wt_val['size'])
can_bson = wt_val['type'] == 3
if can_bson:
    try:
        obj = bson.decode_all(val_bytes)[0]
        ...
    except:
        pass
print('  ' + '\n  '.join(str(wt_val).split('\n')) + " " + str(obj) + " =>")

Two issues:

1. read_memory(wt_val['data'], wt_val['size']) is called unconditionally. For WT_UPDATE_TOMBSTONE (type=4) — and WT_UPDATE_RESERVE (type=2) — size is 0 and data is uninitialized / NULL. The read is unsafe and the resulting bytes are meaningless.
2. The printed line gives no indication of the WT_UPDATE type, so a tombstone prints identically to a STANDARD value whose BSON decode failed (both show None). The operator cannot distinguish a delete from a non-decodable value while walking an update chain in gdb.

WT_UPDATE type values (from src/include/btmem.h):

Fix

• Guard read_memory with size > 0 so tombstone/reserve entries skip the read.
• Decode BSON only when type == WT_UPDATE_STANDARD and size > 0.
• Prefix each printed update with [MODIFY] / [RESERVE] / [STANDARD] / [TOMBSTONE] derived from wt_val['type'].

The import bson at the top of the file is intentionally left as a hard import — silently falling back would let the operator believe the script "decoded nothing" rather than realizing pymongo is missing.

Affected files

• tools/gdb/gdb_scripts/wt_debug_script_update.py