I've just patched the new VoIP call analysis feature from Alejandro
Vaquero.
I checked in changes to remove some compiler warnings - one of which was
warning about a variable that might not have been set in all code paths
where it was used (the H225_RELEASE_COMPLET case in "H225calls_packet()"
wasn't setting "comment" if "q931_frame_num" wasn't equal to
"pinfo->fd->num" - the new code sets it to "H225 No Q931 Rel Cause" in
that case; BTW, is it a common convention to leave the "." out of ITU-T
recommendation names?), and some of which revealed some problems in code
to append to a filter expression (memory leaks, possible freeing of the
result of "gtk_entry_get_text()".
Some additonal warnings:
graph_analysis.c:141: warning: `on_destroy' defined but not used
graph_analysis.c:149: warning: `dialog_graph_set_title' defined but not used
voip_calls.c: In function `SIPcalls_packet':
voip_calls.c:334: warning: `tmp_sipinfo' might be used uninitialized in
this function
voip_calls.c: In function `isup_calls_packet':
voip_calls.c:521: warning: unused variable `tapinfo'
voip_calls.c: In function `mtp3_calls_packet':
voip_calls.c:593: warning: `forward' might be used uninitialized in this
function
voip_calls.c: In function `q931_calls_packet':
voip_calls.c:763: warning: unused variable `tapinfo'
voip_calls.c: In function `H225calls_packet':
voip_calls.c:853: warning: `tmp_h323info' might be used uninitialized in
this function
The "might be used uninitialized" should be checked to make sure there
aren't code paths where it could be uninitialized.
Should the global variables be stored in a "voip_calls_tapinfo_t"
structure? Using it would suppress the "unused variable" warnings, and
might also make the code more easily modifiable to handle multiple taps.
