Comment # 21
on bug 8287
from Dirk
your comments were helpful in rethinking the current state. My previous code
has evolved over time as more "lookup" stuff was added. I agree that everything
should be combined into a simple struct object describing the state of a
request/response. I've rewritten that part and the code looks much better. It
now only has a single lookup function which can return pointers to the
previous, current and next struct with one iteration. I've also used a single
linked list, this has ironically removed the need for wmem_realloc() in this
code.
If the capture file has too much packet loss and retransmission and the TCP
stream reassembly is not clean, my request/response tracking can get confused.
I think however that our users are aware that with problematic capture files
Wireshark's analysis may be flawed and there's also nothing I can do about it.
I've included your suggestions, except for reformatting the definitions of the
new hf_* variables. I personally don't like the arbitrary line breaks used in a
lot of existing dissectors and prefer to have one hf_ variable defined on a
single line. (If whoever likes to commit these changes eventually has another
strong opinion I'm fine if the comitter reformats those)
You are receiving this mail because:
- You are watching all bug changes.
