Comment # 4
on bug 9949
from Evan Huus
As I commented on the duplicate: the majority of the memory usage appears to be
coming from se_ address-to-string conversions used for filling in columns.
There are two issues with this:
1. All the address_to_str functions (se_, ep_, and wmem) massively
over-allocate for simplicity; they grab 256 bytes, even if (as is the case
here) the resulting string is only a few bytes long. This is extremely
2. When running tshark the columns shouldn't be se-allocated anyways; I assume
it's necessary for Wireshark (since the columns are displayed continually) but
for tshark all the column data is free-able as soon as that packet has been
You are receiving this mail because:
- You are watching all bug changes.