Comment # 12
on bug 9027
from Evan Huus
Ah, thank you, that makes sense now!
So in this case, I don't think the answer is to 'fix' the TCP dissector in some
way. As soon as it passes the first TVB to the sub-dissector, we can no longer
assume that the TVB is freeable, since the sub-dissector could keep a reference
to it for all sorts of other reasons (ie tree items keep a tvb reference in
order to highlight the correct bytes in the correct pane when selected in the
GUI).
This means that your suggestion #2 is the correct one (I think). Instead of
freeing the old data, its chain should be merged with the new data so they are
all freed together.
This problem goes away with reassembly tvbs, but that's not backportable
anyways :)
You are receiving this mail because:
- You are watching all bug changes.