I recently found that there are some dissectors which allocate
file-scope memory every time they dissect a packet (not just on the
first pass, when the packet is unvisited). Can anybody think of a
reason for this, or should it be considered a bug?
I have attached a trivial (if hacky) patch which adds an assertion if
file-scope memory is used during the second pass of two-pass tshark.
Applying this and then running "fuzz-test.sh -2" on the menagerie
results in many failures.
This is not a proposal or anything, just an idea I had that produced
some unexpected results. I am curious if this is a valid use or known
issue or something.
Cheers,
Evan
Attachment:
wmem_file_scope_second_pass.patch
Description: Binary data