Brown, Mark C (GSE GCSM) wrote:
I pulled the latest source from the CVS with your changes checked in this morning:
Yeah, I decided not to do the "cleans up a bit of the DL_HP_RAWDLS stuff", as the configure script treats HP-UX other than 9.0, 10.0x, and 10.1x as 10.20 or 11.x, so presumably it's not "HP-UX prior to 9.0", it's "HP-UX 10.0x and 10.1x".
I've no idea whether the code still works on 10.0x or 10.1x, nor whether anybody uses recent versions of libpcap on 10.0x or 10.1x; I haven't seen any bugs reported, which means that either
1) I missed the report;
2) the code works;
3) nobody's tried it on 10.0x or 10.1x (which means that if it doesn't work, perhaps nobody cares);
4) the code doesn't work, but nobody bothered reporting it.
I did the rest of the changes.
A single typo:
--- pcap-dlpi.c_orig 2005-04-16 20:47:40 -0400
+++ pcap-dlpi.c 2005-04-17 08:32:35 -0400
@@ -1116,7 +1116,7 @@
}
static int
-recv_ack(int fd, int size, const char *what, char *bufp, char *ebuf,
int *uerr)
+recv_ack(int fd, int size, const char *what, char *bufp, char *ebuf,
int *uerror)
{
union DL_primitives *dlp;
struct strbuf ctl;
Checked in.
Testing with DLSAP 22 & 24 busy:
# ./tcpdump -i lan0 -c 1 tcpdump: WARNING: recv_ack: bind: UNIX error - Device busy
That shouldn't be printed - if "pcap_open_live()" succeeds but leaves text in the error string buffer, that's now a way to supply warnings to the caller - so I checked in a change to clear the error buffer if we get EBUSY back from a bind attempt in dl_dohpuxbind().
Both checkins are on the main and x.9 branch. - This is the tcpdump-workers list. Visit https://lists.sandelman.ca/ to unsubscribe.
