having two different versions, both broken in different ways.
Bump the count of total packets in the capture-from-pipe routine.
svn path=/trunk/; revision=5647
working on MacOS X.
It appears that the underlying problem with the timeout was that we
weren't treating MacOS X as a BSD, and the "select()" we were doing
presumably wasn't working as it doesn't work on BPF devices on many
BSDs; the workaround no longer appears to be necessary, with Michael's
fix to treat MacOS X as BSD.
(Presumably a select timeout with "tv_usec" set to 1000*1000
microseconds was treated as an error, or otherwise treated in such a way
that it didn't block waiting for the BPF device to say it could be
read.)
svn path=/trunk/; revision=5637
BPF, in at least some OS versions, acts like the other BPFs in some
versions of other BSDs, and doesn't work with "select()".
svn path=/trunk/; revision=5620
This fixes some bugs:
1. With the -S option under Linux, Capture/Stop or ^E was
ignored until the next packet was read. This is because
capture.c wasn't checking for EINTR from select(), which is
returned when the child receives SIGUSR1 from the parent.
2. When reading from a pipe, a spurious error message from
pcap_open_live() was written to stderr.
3. Error messages from the child in Sync mode were displayed in
a Warning alert box.
Also, there's a new subroutine, popup_errmsg(), to replace
several instances of duplicate code.
svn path=/trunk/; revision=5616
unused.
Put in a comment to note that if we fail to open the interface either as
a device or as a pipe, we report the error from the failed
"pcap_open_live()" (which explains why "pipe_open_live()" doesn't return
an error string).
svn path=/trunk/; revision=5381
returns radio information such as signal strength, channel, and data
rate in a pseudo-header. Add that pseudo-header.
Use the "802.11 with radio information" encapsulation type for Wireless
Sniffer files; extract the radio information from where it appears to be
in the header.
Add dissector code for that encapsulation type.
Fix an error in the code to put radio information into the AiroPeek
tree.
Make the "wrapped" flag for NetXRay/Windows Sniffer captures a
"gboolean".
svn path=/trunk/; revision=5122
Move the ringbuffer capture options from the "capture_file" structure to
the structure for capture options, as they're a property of an
in-progress capture, not a property of a particular capture file.
svn path=/trunk/; revision=4799
"capture_file" structure - they're a property of an in-progress capture,
not a property of an open capture file. Make them just variables.
The maximum number of packets to be captured should be a variable
separate from the "count" field in the "capture_file" structure - the
latter is a count of the packets in the capture file in question.
Have Boolean variables indicating whether a maximum packet count,
maximum capture file size, and maximum capture duration were specified.
If an option isn't set, and we're doing an "update list of packets in
real time" capture, don't pass the option to the child process with a
command-line argument.
Don't create "stop when the capture file reaches this size" or "stop
when the capture's run for this long" conditions if a maximum capture
file size or a maximum capture duration, respectively, haven't been
specified. Don't test or free a condition if it wasn't created.
Don't allow a 0 argument to the "-c" flag - the absence of a "-c" flag
is the way you specify "no limit on the number of packets".
Initialize the check boxes and spin buttons for the "maximum packets to
capture", "maximum capture size", and "maximum capture duration" options
to the values they had in the last capture. If an option wasn't
specified, don't read its value from the dialog box and set the
variable.
svn path=/trunk/; revision=4795
reading the capture file. Have callers of "wtap_snapshot_length()"
treat a value of 0 as "unknown", and default to WTAP_MAX_PACKET_SIZE (so
that, when writing a capture file in a format that *does* store the
snapshot length, we can at least put *something* in the file).
If we don't know the snapshot length of the current capture file, don't
display a value in the summary window.
Don't use "cfile.snap" as the snapshot length option when capturing -
doing so causes Ethereal to default, when capturing, to the snapshot
length of the last capture file that you read in, rather than to the
snapshot length of the last capture you did (or the initial default of
"no snapshot length").
Redo the "Capture Options" dialog box to group options into sections
with frames around them, and add units to the snapshot length, maximum
file size, and capture duration options, as per a suggestion by Ulf
Lamping. Also add units to the capture count option.
Make the snapshot length, capture count, maximum file size, and capture
duration options into a combination of a check box and a spin button.
If the check box is not checked, the limit in question is inactive
(snapshot length of 65535, no max packet count, no max file size, no max
capture duration); if it's checked, the spinbox specifies the limit.
Default all of the check boxes to "not checked" and all of the spin
boxes to small values.
Use "gtk_toggle_button_get_active()" rather than directly fetching the
state of a check box.
svn path=/trunk/; revision=4709
"epan/..." pathnames, so as to avoid collisions with header files in any
of the directories in which we look (e.g., "proto.h", as some other
package has its own "proto.h" file which it installs in the top-level
include directory).
Don't add "-I" flags to search "epan", as that's no longer necessary
(and we want includes of "epan" headers to fail if the "epan/" is left
out, so that we don't re-introduce includes lacking "epan/").
svn path=/trunk/; revision=4586
and "Automatic scrolling in live capture" options from the preference
settings for them, so that the preference settings affect the initial
values of those options, but changing those values in a capture don't
affect the preferences, and don't automatically get saved when you save
the preferences.
If we're building without libpcap, don't have an "Automatic scrolling in
live capture" option anywhere.
svn path=/trunk/; revision=4514
(by deleting the main window or selecting File->Quit or typing ^Q) while
an "Update list of packets in real time" capture is in progress, we can
abort the capture.
Arrange that "fork_child" is -1 when there is no capture child, so said
routine knows when it can kill the child.
When we exit, kill off any capture child, using that routine, and, if
we're exiting due to a request to delete the main window and, if a read
is in progress (from an "Update list of packets in real time" capture),
don't delete the main window - just set the "Read aborted" flag, so that
the code doing the read will see that flag (it will be called because
the pipe to the capture child is closed due to the child exiting) will
see that and clean up and exit itself.
svn path=/trunk/; revision=4498
"gboolean", as it's a Boolean value, and move it to the beginning of the
structure in Tethereal, as it is in Ethereal.
From Graeme Hewson:
Check for "pcap_dispatch()" returning -1, meaning an error
occurred; if it does, stop capturing, and report the error.
If we get a signal in tethereal, stop the capture with a
"longjmp()", rather than by clearning the "go" flag;
"pcap_dispatch()", on many platforms, keeps reading rather than
returning a captured packet count of 0 if the system call to
read packets returns -1 with an errno of EINTR, so the
"pcap_dispatch()" won't be broken out of if the signal handler
returns.
Fix a typo in an error message.
svn path=/trunk/; revision=4471
files to get that big.
From Thomas Wittwer and Matthias Nyffenegger:
Support for "ring buffer mode", wherein there's a ring buffer of N
capture files; as each capture file reaches its maximum size (the ring
buffer works only with a maximum capture file size specified), Ethereal
rolls over to the next capture file in the ring buffer, replacing
whatever packets might be in it with new packets.
svn path=/trunk/; revision=4323
Rename WTAP_ENCAP_PRISM to WTAP_ENCAP_PRISM_HEADER, to match
DLT_PRISM_HEADER.
Add in missing capture support for WTAP_ENCAP_PRISM_HEADER when
capturing with "pcap_open_live()" rather than reading the capture from a
pipe.
svn path=/trunk/; revision=4299
of packet data captured.
Make the "BYTES_ARE_IN_FRAME()" macro take a "captured length of the
packet" argument.
Add some length checks to capture routines.
svn path=/trunk/; revision=4235
for AIX 5.x's non-standard libpcap, where "pcap_datalink()" doesn't
return DLT_ values, it returns RFC 1573 ifType values.
Put that wrapper, and the routine to get the interface list, in a
separate file, for packet-capture utility routines, so not everybody who
includes "util.h" needs to include <pcap.h>.
Fix up the Wiretap hack for dealing with said incompatibility to use the
correct ifType value for Token Ring.
svn path=/trunk/; revision=4184
byte count of zero, don't bother allocating a buffer for that message,
as we wouldn't do anything with that buffer.
Null-terminate the error message once we read it, before using it as a
string.
svn path=/trunk/; revision=3551
a "Match Selected" on it - we can't do a "Match Selected" if the field
has no value (e.g., FT_NULL) and has a length of 0.
If we unselect the current packet, we don't have a protocol tree, so we
don't have a currently selected field - clear the "Match Selected" menu
item and the display in the status line of information about the
currently selected field.
Move the low-level statusbar manipulation into "gtk/main.c", in routines
whose API doesn't expose anything GTK+-ish.
"close_cap_file()" calls one of those routines to clear out the status
bar, so it doesn't need to take a pointer to the statusbar widget as an
argument.
"clear_tree_and_hex_views()" is purely a display-manipulating routine;
move it to "gtk/proto_draw.c".
Extract from "tree_view_unselect_row_cb()" an "unselect_field()" routine
to do all the work that needs to be done if the currently selected
protocol tree row is unselected, and call it if the currently selected
packet list row is unselected (if it's unselected, there *is* no
protocol tree, so no row can be selected), as well as from
"tree_view_unselect_row_cb()".
Before pushing a new field-description message onto the statusbar, pop
the old one off.
Get rid of an unused variable (set, but not used).
svn path=/trunk/; revision=3513
control whether we have a child process do the capturing; a user might
want the packet list to be updated as packets arrive but *not* want it
to scroll so that the most recently arrived packets are shown.
"prefs.capture_auto_scroll", not "auto_scroll_live", should control
whether we scroll a real-time-update capture's packet list;
"auto_scroll_live" isn't set by the capture dialog box,
"prefs_capture_auto_scroll" is.
svn path=/trunk/; revision=3388
binaries, so users only need to make sure they have that version
installed in order to have Ethereal (and tcpdump, and snort, and so on)
accept "lanN"-style names (i.e., names of the sort reported by lanscan
and handled by ifconfig), rather than "dlpiN".
Get rid of the patches to update libpcap, get rid of the discussion in
"README.hpux" of patching libpcap and just say "get 0.6.2", and make the
notes on HP-UX kernel patches to fix problems with capturing outgoing
packets a separate item in the list of items in "README.hpux".
Also update the error messages Ethereal and Tethereal display if they
can't open a device and the error is "can't find PPA for XXX" to say
"get 0.6.2" rather than "patch libpcap and recompile.
svn path=/trunk/; revision=3288
DLT_HDLC to it.
Make a separate dissector for Cisco HDLC, and add a dissector for Cisco
SLARP. Have the PPP dissector call the Cisco HDLC dissector if the
address field is the Cisco HDLC unicast or multicast address. Use the
Cisco HDLC dissector for the Cisco HDLC Wiretap encapsulation type.
Add a new dissector table "chdlctype", for Cisco HDLC packet types
(they're *almost* the same as Ethernet types, but 0x8035 is SLARP, not
Reverse ARP, and 0x2000 is the Cisco Discovery protocol, for example),
replacing "fr.chdlc".
Have a "chdlctype()" routine, similar to "ethertype()", used both by the
Cisco HDLC and Frame Relay dissectors. Have a "chdlc_vals[]"
"value_string" table for Cisco HDLC types and protocol names. Split the
packet type field in the Frame Relay dissector into separate SNAP and
Cisco HDLC fields, and give them the Ethernet type and Cisco HDLC type
"value_string" tables, respectively.
svn path=/trunk/; revision=3133
list of packets in real time" capture so that "!" always indicates an
error, with the "!" preceded by a count of characters in the error
message and followed by the text of the error, and so that those error
messages can be sent after the capture has started.
Use that to report capture errors, and errors writing to the capture
file, while the capture is under way.
Use #defines for the message type characters in that protocol.
svn path=/trunk/; revision=3018
Print the "Capturing on <interface>" message, the running count of
packets captured, and error messages to the standard error in Tethereal,
so that you can pipe the output of a live capture that's printing
packets to a program or script without that script having to worry about
parsing stuff other than dissected packet summaries or details (tcpdump
does the same).
svn path=/trunk/; revision=3017
capturing; if we succeed, display the packet drops count as the "Drops"
value in the status line and as the "Dropped packets" statistics in the
summary dialog box, otherwise don't display it at all.
In Tethereal, attempt to get the packet statistics from libpcap when
capturing; if we succeed, and if there were any dropped packets, print
out the count of dropped packets when the capture finishes.
svn path=/trunk/; revision=3016
compiled capture filter program, so remove it, and remove the include of
<pcap.h> from "file.h"; instead, have local "struct bpf_program"
structures where needed, and have those files that need stuff from
<pcap.h> include it.
This cleans stuff up a bit, and should eliminate a pile of compile
warnings with Visual C++ due to <pcap.h> and some GTK+/GLib header file
(or files they include) both defining "inline".
svn path=/trunk/; revision=2954
doesn't work like the read timeout in BPF - the timer doesn't start
until at least one packet has arrived.
I think that's the way read timeouts should work on *all* packet capture
mechanisms, but it does mean that Solaris will, on a quiet net, exhibit
the same symptoms that Linux used to exhibit before we put in a
"select()" call to wait until either packets arrive or a timer expires -
the "pcap_dispatch()" call blocks until a packet arrives, so the display
doesn't get updated and Ethereal doesn't respond to user input until a
packet arrives.
Furthermore, Linux isn't the only OS that lacks any read timeout
on its packet capture mechanism; the others will also have that problem.
We therefore do the "select()" on *all* platforms other than the BSDs
(where the timer starts when the read is done, and can be used for
polling); I don't know whether it's necessary on Digital UNIX, but I
suspect it's necessary on SunOS 4.x (as the 5.x "bufmod" is probably
derived from the 4.x one, and the 5.x one, as per the above, starts the
timer when a packet arrives), and it may even be necessary on 3.x, those
(BSD, SunOS including 5.x, and Digital UNIX) apparently being the only
UNIXes that appear to have such a read timeout.
svn path=/trunk/; revision=2790
on it, such as the exit status if it exited "normally" but unexpectedly.
On UNIX systems, #define the various POSIX <sys/wait.h> macros (and the
non-POSIX WCOREDUMP()" macro) if they're not defined by <sys/wait.h> (or
if we don't have <sys/wait.h>), and use them to dissect the exit status.
svn path=/trunk/; revision=2788