Apple Silicon requires CMake 3.19.2, but the binaries provided
for 3.19.2 only run on MacOS 10.10 and later, so we have more
bifurcation of the CMake we try to install. Get rid of some of
the old 2.x paths to compensate.
In Git's packfile transfer protocol[1], a pkt-line can start with \1, \2, or \3
to represent whether this line contains packfile data (1), progress information (2),
or an error (3). Parse out this information as a field to make it easier for people
debugging to filter to the pkt-lines of interest to them.
[1] https://www.kernel.org/pub/software/scm/git/docs/technical/pack-protocol.html
Related to #17093
fixed the display filter explanation, corrected the screenshot, added the levels description
minor changes to export PDUs to file section
adding important part about exporting PDUs by AndersBroman comment
applying SME suggestions
Set the Packet List scrollbar page step to be equal to the height
of the scrollbar.
This makes page stepping using the packet list slider more natural.
In OverlayScrollBar return the real ScrollBar sliderPosition to ensure
the correct value is used when handling the actionTriggered signal in
vScrollBarActionTriggered().
This improves turning on and off auto scroll during capture when page
stepping using the packet list slider, because the value is propagated
after this signal.
Old behaviour is to read the entire file into memory at once; provide
the XML tree as the first packet; and then individual `<msg>` elements
as subsequent packets. It did this by writing to a temporary pcapng
file.
This change causes the XML file to only be read a chunk at a time (and
be read directly, not through an intermediate pcapng). This means much
larger files can be loaded, at the cost of no longer showing the raw XML
as the first packet. This is not a loss because the file can be loaded
in MIME Files Format (or a text editor) to see the XML.
Much of the logic from the old functions `create_temp_pcapng_file()` and
`write_packet_data()` has been relocated into the new function
`nettrace_msg_to_packet()`, and is used to directly generate packet data
for wiretap instead of writing it to a temporary file.
Also includes some initial "code smell" fixes:
- Removed some duplicate `#define`s from epan/exported_pdu.h
- Replaces some magic numbers with macros from epan/exported_pdu.h
- Replaces other magic numbers with the CLEN() macro to make it easier
to see (and debug) where sizes/offsets come from
- Use `g_strstr_len()` instead of `strstr()` to remove the need to
insert string terminators
- Uses direct pointer math instead of indexing into a byte array
This compiles and runs, and seems to produce the same results as the old
reader (except for the XML packet). Consider it a proof of concept; it
needs further revision before being review-ready.
The patch changes:
- Removed first_packet_mac_addr staff. It was commented out many years ago...
- Removed delta_timestamp item. Not used.
- Sequence verification takes into account timestamp therefore it is able to detect delayed packets more clearly. As consequence of it, #16330 is solved.
- If packet is delayed, it is not used in calculation of diff/jitter/skew. It just mess output. As consequence of it, #16330 is solved.
I checked output with many RTP streams and looks well. But I have no
sample with wrapped timestamp and I have just a few samples with
missing/reordered packets. Nevertheless all are calculated same way as
before and #16330 is solved too.
When the refid contains non-ascii chars, the conversion function
returns a string longer than 4 chars. This results in an invalid
string if the output is limited to 4 bytes. Incidentally this
results in an invalid PDML output as well that caught this bug
in the first place.
Fix: #17112.
Every supported distribution has at least the 3.3 branch of GnuTLS
(stable branch starting in April 2014). That branch was maintained
for bug-fixes until July 2018, so some distributions (e.g. RHEL7,
SUSE Enterprise 12) are still on it, keeping us from requiring 3.4 yet.
Also clarify a comment about when the Mac OS build of gnutls started
being compiled with pkcs11 support.
When waveform start much later than at time 0s, centering of waveform
was incorrect. The reason was that range was taken before rescalingAxes.
Patch solves the issue.
In MPTCP v1, in the MPC handshake, the first key is carried by
the SYN/ACK packet: when the mptcp analysis context is created,
the forward direction comes from the server. That is the opposite
of what used to happen with MPTCP v0 - the initial SYN already
carried the sender key.
As a result, when a later MP_JOIN subflows attach to an existing
MPTCP connection v1, the mptcp_subflow meta points to the wrong
flow, and the relative DSN/ACK_SEQ decoding for such flows gives
pseudo-random results.
This change addresses the issue by swapping the MPTCP meta for
MP_JOIN flows after mptcp_get_meta_from_token(), if the negotiated
version for the relevant MPTCP connection is v1.
Reason for #16452 is that RTP Player dialog is destroyed before tap
finishes when Escape key pressed. The only solution I found is to check
whether dialog does exists when tap finishes. If so, data are processed
and shown. If dialog was destroyed in meanwhile, no data processing is done.
Anders Broman