Comment # 7
on bug 9579
from Peter Wu
The above was a simplification of the issue, but (skipping disabled code) there
are still parts I don't fully understand (first_pass, branch_taken).
Replacing the goto by a continue has as side-effect that *branch_taken gets
incremented and possibly reset to -1 if !ch->func.
This BER thing seems very complex, can't branch_taken just be a boolean?
(looking at epan/dissectors/packet-mms.c:7032 for example).
Replacing goto by continue (and then the choice_try_again label can be removed
too) fixes the overflow issue and assuming that branch_taken is a boolean (-1
false, others true), then I think it is safe to make the change.
I have not tested it since it will "obviously" fix the ASAN bug, but will
possibly modify the output and I cannot tell whether the new output will be
correct or not.
Remember to compile with -DCMAKE_C_FLAGS='-fsanitize=address'
-DCMAKE_CXX_FLAGS='-fsanitize=address' (or similar CFLAGS, CXXFLAGS and
possibly LDFLAGS for autotools) to enable ASAN.
You are receiving this mail because:
- You are watching all bug changes.
