Hi,
get_addr_name, in addr_resolv.c, returns a name corresponding to an address if the address
supports name resolution, and NULL if not. I've seen it is used only in column-utils,
conversations_table and hostlist_table. In all three of them, if it returns NULL, the code calls
address_to_str or address_to_str_buf immediately afterwards. These two return the address
conversion if they have received a valid address type, or g_assert if they received an AT_NONE.
I think it would be better if get_addr_name would:
- resolve to a name if the address supports it
- call address_to_str if it does not, but the address is valid
- return "NONE" if it is AT_NONE
In that way, the code would not need to check the result. If the code is broken, instead of
crashing -the way it would now in the mentioned files- it would just print "NONE" wherever an
address is expected; and it would be possible to do things like
g_snprintf(title2, 80, "Reverse: %s:%u to %s:%u (SSRC=%u)",
get_addr_name(&(user_data->ip_src_rev)),
user_data->port_src_rev,
get_addr_name(&(user_data->ip_dst_rev)),
user_data->port_dst_rev,
user_data->ssrc_rev);
in RTP Analysis, where the data is not always available -in this case, because there might be a
forward stream without a reverse one-, without needing to check every time for the existence.
I guess the reason against doing so would be that it enforces code correctness less than a
g_assert, but having "NONE" instead of the string you expected should be pretty obvious also.
Would this be OK?
Regards,
Francisco
