When there is no do_dissection cf is missing some variables
for cf_close() call. Therefore we have to set them explicitly.
Fixes: wireshark/wireshark#17021
This reverts commit 5df2925434.
The problem only showed up in tfshark.c, and was caused by tfshark.c
using stuff from ui/urls.h but not *including* ui/urls.h.
If you use it, GCC 9.3.0 seems to think there's a missing parenthesis
somewhere, just as the version of clang++ in my version of Xcode does,
even though other versions of GCC don't. I'm clearly missing something
obscure about C here; I give up.
Instead of grabbing the set of IDBs found at open time, have a loop
using wtap_get_next_interface_description() to read all unread IDBs run
after opening the input file, after reading a packet from the input
file, and after getting an EOF on the input file.
Add a routine wtap_uses_interface_ids() to check whether the file type
and subtype for a dump file uses interface IDs and requires IDBs. If
so, in the aforementioned loop, add the IDBs to the dump stream.
Add a routine wtap_dump_add_idb() to add IDBs to a dump stream. Have it
call a file-format-specific routine to add the IDBs; the only file type
that supports it is pcapng, and it 1) writes out the IDB and 2) adds it
to the set of IDBs for the stream.
Add a wtap_dump_params_init_no_idbs() routine that prevents the IDBs
from the input file from being used to initialize the output file; use
it in cases where we're using the aforementioned loop to copy over IDBs.
Don't require any IDBs to be present when opening a pcapng file for
writing; 1) the simplest pcapng file has just an SHB in it, 2) that
requirement causes dumps that don't provide IDBs at open time to fail,
and 3) the real issue is that we don't want packets with an interface ID
not corresponding to a known IDB, and we already have a check for that.
(There are some hacks here; eventually, when everything processes the
IDBs in such a loop, we may be able to get rid of the "two favors of
dump parameter initialization" hack.)
Fixes#15844.
Addresses the same issue in #15502, but there are other issues there
that also need to be addressed.
In addition, the merge code also needs to be changed to handle this.
Add ui/urls.h to define some URLs on various of our websites. Use the
GitLab URL for the wiki. Add a macro to generate wiki URLs.
Update wiki URLs in comments etc.
Use the #defined URL for the docs page in
WelcomePage::on_helpLabel_clicked; that removes the last user of
topic_online_url(), so get rid of it and swallow it up into
topic_action_url().
Add extcap_ensure_all_interfaces_loaded, which calls
extcap_load_interface_list if our interface list is empty. Call it in
each of our public functions that require a valid interface list.
Clean up the extcap API documentation and note which routines initialize
the interface list.
In tshark, don't unconditionally call extcap_register_preferences and
instead rely on lazy loading.
Change-Id: I8493ae5f4d703b0fd767246557d17723bcf207c6
Ping-Bug: 15295
Reviewed-on: https://code.wireshark.org/review/37750
Petri-Dish: Gerald Combs <gerald@wireshark.org>
Tested-by: Petri Dish Buildbot
Reviewed-by: Anders Broman <a.broman58@gmail.com>
In each of our executables we were calling "setlocale(LC_ALL, "")" at
startup. This told Windows that output was encoded using the current
system code page. Unless the code page was 65001 (UTF-8), this was a lie.
We write UTF-8 to stdout and stderr, so call "setlocale(LC_ALL, ".UTF-8)"
at startup on Windows. This lets the CRT translate our output correctly
in more cases.
Clarify and expand the OUTPUT section in the tshark man page.
Bug: 16649
Change-Id: If93231fe5b332c292946c7f8e5e813e2f543e799
Reviewed-on: https://code.wireshark.org/review/37560
Petri-Dish: Gerald Combs <gerald@wireshark.org>
Tested-by: Petri Dish Buildbot
Reviewed-by: Gerald Combs <gerald@wireshark.org>
wtap_cleanup() clears options which are still in use by the time
cf_close calls wtap_close. Be sure to close the capture file first.
Bug: 16487
Change-Id: Id9ef1c0321865e9574b69439870a842efb2b209b
Fixes: v3.3.0rc0-853-g3662a69036 ("Maintain cf->state, because file cleanup depends on it.")
Reviewed-on: https://code.wireshark.org/review/36755
Petri-Dish: Peter Wu <peter@lekensteyn.nl>
Reviewed-by: Guy Harris <gharris@sonic.net>
pcapng has been the compile-time default since 2011. If there are any
users who would like to use the libpcap format, then they should use
runtime options instead (e.g. `tshark -P` or `editcap -F pcap`).
Change-Id: I54b70368cdc3ca78bc8617bc488cc687740a1eb9
Reviewed-on: https://code.wireshark.org/review/36721
Petri-Dish: Peter Wu <peter@lekensteyn.nl>
Reviewed-by: Gerald Combs <gerald@wireshark.org>
Tested-by: Petri Dish Buildbot
Reviewed-by: Guy Harris <gharris@sonic.net>
Instead of having programs that use the capchild library define
functions with known names, with the library routines calling back
routines with those names, have function pointers for those callbacks in
the capture_session structure, and have capture_session_init() set them.
Make the callback routines in TShark and in the ui library static.
Change-Id: Ia1ba6119c5ef7708e0f87b8420f200136ba41eae
Reviewed-on: https://code.wireshark.org/review/36583
Petri-Dish: Guy Harris <gharris@sonic.net>
Tested-by: Petri Dish Buildbot
Reviewed-by: Guy Harris <gharris@sonic.net>
If we're capturing to multiple files, whenever we are told about a new
file, we must close the old file, so we don't leak file descriptors and
wtap structures.
Have cf_close() handle the work of closing, the way it does in file.c,
and, when we *open* a file, set the state in capture_file to
FILE_READ_IN_PROGRESS.
Bug: 16457
Change-Id: I04a01c30571b7e3575dee5e7252a59bb1ee8abbc
Reviewed-on: https://code.wireshark.org/review/36580
Petri-Dish: Guy Harris <gharris@sonic.net>
Tested-by: Petri Dish Buildbot
Reviewed-by: Guy Harris <gharris@sonic.net>
2004 called, they want their libpcap/WinPcap back.
RHEL 6 initially shipped with libpcap 1.0; even old Enterprise(TM)
versions of OSes ship with something shinier than 0.7.x these days.
This lets us get rid of a bunch of #ifdefs and workaround code for
missing APIs.
Change-Id: I862cb027418b0a0c0f45a26979acea82f93f833b
Reviewed-on: https://code.wireshark.org/review/36383
Petri-Dish: Guy Harris <guy@alum.mit.edu>
Tested-by: Petri Dish Buildbot
Reviewed-by: Guy Harris <guy@alum.mit.edu>
Since dumpcap was split off tshark, tshark never had the need to use
file capabilities. Remove the unused header.
Change-Id: I76e9d09599a4276d4be5ba105d7c6e28e9dd96da
Reviewed-on: https://code.wireshark.org/review/35984
Petri-Dish: Peter Wu <peter@lekensteyn.nl>
Tested-by: Petri Dish Buildbot
Reviewed-by: Alexis La Goutte <alexis.lagoutte@gmail.com>
For long options, without corresponding short options, to be processed
they need to be assigned a value, preferably outside of the range of
all possible short options. The code in various places tries to stay
clear of these low values, but further coordination is missing, easily
leading to issues when option processing code gets extended and/or
reorganized.
This change introduces a single location from where each catagory of
command line long option can derive a base value, which should minimize
potential option value collisions.
Change-Id: Ic8861a347d0050f74002de3aa1fcfb01202866e5
Reviewed-on: https://code.wireshark.org/review/35459
Reviewed-by: Jaap Keuter <jaap.keuter@xs4all.nl>
Petri-Dish: Jaap Keuter <jaap.keuter@xs4all.nl>
Tested-by: Petri Dish Buildbot
Reviewed-by: Alexis La Goutte <alexis.lagoutte@gmail.com>
Documentation of the Tshark and dumpcap command line options between
help text, manual page and user's guide diverged over time. One aspect
of this is the implementation of more long options. This change tries to
update all documentation to be complete and in sync again.
Change-Id: Ie8bee013df8d209080fcf288072774f18f9ff51f
Reviewed-on: https://code.wireshark.org/review/35261
Reviewed-by: Jaap Keuter <jaap.keuter@xs4all.nl>
Petri-Dish: Jaap Keuter <jaap.keuter@xs4all.nl>
Tested-by: Petri Dish Buildbot
Reviewed-by: Michael Mann <mmann78@netscape.net>
EPIPE almost certainly means "the next program after us in the pipeline
exited before we were finished writing", so this isn't a real error, it
just means we're done. (We don't get SIGPIPE because libwireshark
ignores SIGPIPE to avoid getting killed if writing to the MaxMind
process gets SIGPIPE because that process died.)
Presumably either that program exited deliberately (for example, "head
-N" read N lines and printed them), in which case there's no error to
report, or it terminated due to an error or a signal, in which case
*that's* the error and that error has been reported.
(We don't do that for EINVAL, as that's presumably a real error. It
shows up on Windows in bug 16192, but what we probably want to do there
is to, on Windows, use _doserrno, check for the equivalent Windows
errors, and, for the default case, convert _doserrno to the appropriate
string, using Windows APIs, and report *that* string; the MS C library
converts a whole bunch of Windows errors to EINVAL, thus losing
information and making it harder to determine what the real error is.
Therefore, I'm just marking this with Ping-Bug, as it's only fixing the
problem on UN*Xes.)
Change-Id: I94c392f478561e29501facd657487716a5882295
Ping-Bug: 16192
Reviewed-on: https://code.wireshark.org/review/35053
Petri-Dish: Guy Harris <guy@alum.mit.edu>
Tested-by: Petri Dish Buildbot
Reviewed-by: Guy Harris <guy@alum.mit.edu>
That means the packet-count-during-capture stuff is scattered amongst
fewer locations.
Move capture_info.h into ui; it's now a header that declares routines
whose implementations are GUI-platform-dependent.
Change-Id: I475815724a4766f6bc2511e67ebae14865e1a9d1
Reviewed-on: https://code.wireshark.org/review/26249
Petri-Dish: Michael Mann <mmann78@netscape.net>
Tested-by: Petri Dish Buildbot
Reviewed-by: Michael Mann <mmann78@netscape.net>
Bug: 15825
Change-Id: Iec8dff38dd89e3947f3fe7053e38101c3ad7b1b2
Reviewed-on: https://code.wireshark.org/review/33523
Petri-Dish: Guy Harris <guy@alum.mit.edu>
Tested-by: Petri Dish Buildbot
Reviewed-by: Michael Mann <mmann78@netscape.net>
Those routines exist on both Windows and UN*X, but they don't do
anything on UN*X (they could if it were ever necessary).
That eliminates some #ifdefs, and also means that the gory details of
initializing Winsock, including the Winsock version being requested,
are buried in one routine.
The initialization routine returns NULL on success and a pointer to a
g_malloc()ated error message on failure; report the error to the user,
along with a "report this to the Wireshark developers" suggestion.
That means including wsutil/socket.h, which obviates the need to include
some headers for socket APIs, as it includes them for you.
Change-Id: I9327bbf25effbb441e4217edc5354a4d5ab07186
Reviewed-on: https://code.wireshark.org/review/33045
Petri-Dish: Guy Harris <guy@alum.mit.edu>
Tested-by: Petri Dish Buildbot
Reviewed-by: Guy Harris <guy@alum.mit.edu>
Make sure we link each application that calls WSAStartup with ws2_32.lib.
Pass version 2.2 to WSAStartup. Wikipedia says it was introduced in 1996,
so we should be OK.
Ping-Bug: 15711
Change-Id: I431839e930e7c646669af7373789640b5180ec28
Reviewed-on: https://code.wireshark.org/review/33033
Petri-Dish: Gerald Combs <gerald@wireshark.org>
Reviewed-by: Tomasz Moń <desowin@gmail.com>
Tested-by: Petri Dish Buildbot
Reviewed-by: Gerald Combs <gerald@wireshark.org>
Ethernet packets without the CRC are 1514 bytes long, not 1500 bytes
long; using 1514 bytes will avoid a reallocation for a full-sized
Ethernet packet.
Change-Id: Ie8da3f13bf3df07e23e4478b7dcf84f06dec6a9d
Reviewed-on: https://code.wireshark.org/review/32761
Reviewed-by: Guy Harris <guy@alum.mit.edu>
That makes it - and the routines that implement it - work more like the
seek-read routine.
Change-Id: I0cace2d0e4c9ebfc21ac98fd1af1ec70f60a240d
Reviewed-on: https://code.wireshark.org/review/32727
Petri-Dish: Guy Harris <guy@alum.mit.edu>
Tested-by: Petri Dish Buildbot
Reviewed-by: Guy Harris <guy@alum.mit.edu>
Catch signals/ctrl events when we're reading a capture, and stop reading
if we get one of those. When we close a print stream, restore the color
as appropriate.
Change-Id: I3dd936964560fb3902befe0fd2e961f80437ca72
Ping-Bug: 15659
Reviewed-on: https://code.wireshark.org/review/32716
Petri-Dish: Guy Harris <guy@alum.mit.edu>
Tested-by: Petri Dish Buildbot
Reviewed-by: Guy Harris <guy@alum.mit.edu>
Put the pass 1 of a two-pass read, pass 2 of a two-pass read, and only
pass of a one-pass read into separate routines, returning success/read
error/write error status codes.
This makes the processing a bit cleaner, and makes it easier to have the
file-reading code catch signals/control events.
Change-Id: I58cd9e4b86f219f3afa2dc61b57f41978fc2f853
Reviewed-on: https://code.wireshark.org/review/32711
Petri-Dish: Guy Harris <guy@alum.mit.edu>
Tested-by: Petri Dish Buildbot
Reviewed-by: Guy Harris <guy@alum.mit.edu>
There's nothing to draw.
Maybe we should also avoid it if we didn't get any packets.
Change-Id: If76f7909f78e66b7302d0ab2caa284ca36c43bfb
Reviewed-on: https://code.wireshark.org/review/32649
Reviewed-by: Guy Harris <guy@alum.mit.edu>
This makes it match the "Plugins" tab of the "About" dialog.
While we're at it, use the same code to enumerate extcap plugins in that
dialog.
Change-Id: I50f402a7ab5d83d46baab070d145558ed8f688f4
Reviewed-on: https://code.wireshark.org/review/32589
Reviewed-by: Guy Harris <guy@alum.mit.edu>
capture_input_drops() doesn't, and shouldn't, modify or free or... the
interface name, so make the pointer to it a const pointer.
Change-Id: Iafc5c5dd9939225b3aeb8a8e36c5bdeecc394e12
Reviewed-on: https://code.wireshark.org/review/32465
Reviewed-by: Guy Harris <guy@alum.mit.edu>
capture_input_cfilter_error_message() doesn't, and shouldn't, modify or
free or... the error message, so make the pointer to it a const pointer.
Change-Id: Ic14ac306add328df369af4b6e149c856f4283912
Reviewed-on: https://code.wireshark.org/review/32464
Reviewed-by: Guy Harris <guy@alum.mit.edu>
Set the output_file_name variable for -w regardless of whether we were
built with libpcap or not. If we were built with libpcap, also pass the
flag and its argument to capture_opts_add_opt().
In the reading-a-file code (rather than the doing-a-live-capture code),
use output_file_name as the name of the output file, regardless of
whether we were built with libpcap or not.
This takes a few twists out of the maze of #ifdefs, all different.
Change-Id: I828f1b04dacbf0ea4f3aff36f26cb9a3ffcbc480
Reviewed-on: https://code.wireshark.org/review/32011
Petri-Dish: Guy Harris <guy@alum.mit.edu>
Tested-by: Petri Dish Buildbot
Reviewed-by: Guy Harris <guy@alum.mit.edu>
Be sure to map "ssl" to "tls" instead of silently ignoring it.
Change-Id: If1edc10ead4a9f25ee4802e1395390dc3c51796f
Reviewed-on: https://code.wireshark.org/review/31894
Petri-Dish: Peter Wu <peter@lekensteyn.nl>
Tested-by: Petri Dish Buildbot
Reviewed-by: Peter Wu <peter@lekensteyn.nl>
Align the usage help text for '-' as filename for stdin and stdout
with the text used for wireshark.
Change-Id: I67011b8234616940b7878fd5768c9e2a9e79f9f0
Reviewed-on: https://code.wireshark.org/review/31838
Petri-Dish: Stig Bjørlykke <stig@bjorlykke.org>
Tested-by: Petri Dish Buildbot
Reviewed-by: Peter Wu <peter@lekensteyn.nl>
Capturing with "tshark -i lo" results in capture_opts->descr being set
to "Loopback" via:
#3 0x55c5f575720c in fill_in_interface_opts_from_ifinfo capture_opts.c:547:33
#4 0x55c5f5750dc5 in capture_opts_add_iface_opt capture_opts.c:695:9
#5 0x55c5f574b6bd in capture_opts_add_opt capture_opts.c:843:18
#6 0x55c5f5785efc in main tshark.c:1087:21
but tshark overwrites it, presumably to offer textual descriptions like
"Standard input" for "-i -". Fix this memory leak, reported by ASAN for
three tests from case_tshark_capture that capture from Loopback.
Change-Id: I4f393c4440bde7a621271cca3066bef3d57e250a
Reviewed-on: https://code.wireshark.org/review/31756
Petri-Dish: Peter Wu <peter@lekensteyn.nl>
Tested-by: Petri Dish Buildbot
Reviewed-by: Peter Wu <peter@lekensteyn.nl>
Found by clang-tidy.
Change-Id: I7fe1c3219758d8daf411d094e2df17916f2eb57b
Reviewed-on: https://code.wireshark.org/review/31330
Petri-Dish: Dario Lombardo <lomato@gmail.com>
Tested-by: Petri Dish Buildbot
Reviewed-by: Anders Broman <a.broman58@gmail.com>
Restore the "main" name since that is used everywhere else except for
Windows. On Windows, "main" is renamed via a macro to avoid a conflict
with "wmain" and to allow it to be called in cli_main.c.
For those wondering, GUI applications (such as Qt) have a different
entry point, namely WinMain. In Qt5, src/winmain/qtmain_win.cpp defines
WinMain, but seems to convert its arguments from Unicode to CP_ACP
(ASCII). It might not support UTF-8, but I did not verify this.
Change-Id: I93fa59324eb2ef95a305b08fc5ba34d49cc73bf0
Reviewed-on: https://code.wireshark.org/review/31208
Petri-Dish: Peter Wu <peter@lekensteyn.nl>
Reviewed-by: Anders Broman <a.broman58@gmail.com>
cmdarg_err() is for reporting errors for command-line programs and
command-line errors in GUI programs; it's not something for any of the
Wireshark libraries to use.
The various routines for parsing numerical command-line arguments are
not for general use, they're just for use when parsing arguments.
Change-Id: I100bd4a55ab8ee4497f41d9651b0c5670e6c1e7f
Reviewed-on: https://code.wireshark.org/review/31281
Petri-Dish: Guy Harris <guy@alum.mit.edu>
Tested-by: Petri Dish Buildbot
Reviewed-by: Guy Harris <guy@alum.mit.edu>
Add interface name (colon delimited) to SP_DROPS ('D') message so when dropped
packets are outputted, they include the interface name for clarity.
Bug: 13498
Change-Id: I68cdde4f20a574580f089dc5096d815cde5d3357
Reviewed-on: https://code.wireshark.org/review/31218
Petri-Dish: Michael Mann <mmann78@netscape.net>
Tested-by: Petri Dish Buildbot
Reviewed-by: Anders Broman <a.broman58@gmail.com>
Make the time stamp precision a 4-bit bitfield, so, when combined with
the other bitfields, we have 32 bits. That means we put the flags at
the same structure level as the time stamp precision, so they can be
combined; that gets rid of an extra "flags." for references to the flags.
Put the two pointers next to each other, and after a multiple of 8 bytes
worth of other fields, so that there's no padding before or between them.
It's still not down to 64 bytes, which is the next lower power of 2, so
there's more work to do.
Change-Id: I6f3e9d9f6f48137bbee8f100c152d2c42adb8fbe
Reviewed-on: https://code.wireshark.org/review/31213
Petri-Dish: Guy Harris <guy@alum.mit.edu>
Tested-by: Petri Dish Buildbot
Reviewed-by: Guy Harris <guy@alum.mit.edu>
Error:
CMakeFiles/tshark.dir/tshark.c.o -c ../tshark.c
../tshark.c: In function 'real_main':
../tshark.c:706:24: error: variable 'max_packet_count' might be clobbered by 'longjmp' or 'vfork' [-Werror=clobbered]
int max_packet_count = 0;
^~~~~~~~~~~~~~~~
cc1: all warnings being treated as errors
ninja: build stopped: subcommand failed.
Change-Id: I92c0e2a57f5d0d2e39188cc18f0b25e1d5136a43
Reviewed-on: https://code.wireshark.org/review/31149
Petri-Dish: Dario Lombardo <lomato@gmail.com>
Reviewed-by: Anders Broman <a.broman58@gmail.com>
Option '-c' is not only a capture option, it also affects offline reads
as documented in the tshark manual. Fixes failing tests since
v2.9.1rc0-18-g5bf37f63a8 ("text2pcap: allow to set interface name").
Change-Id: Iffe4fd60f62766282e1a8b02a942673ba4e605f0
Reviewed-on: https://code.wireshark.org/review/31130
Petri-Dish: Peter Wu <peter@lekensteyn.nl>
Tested-by: Petri Dish Buildbot
Reviewed-by: Peter Wu <peter@lekensteyn.nl>
It really shouldn't free the comment passed to it, as the caller
allocated it, and knows how to free it if necessary; it might not even
have been allocated.
Make the comment argument a "const char *" to 1) allow passing string
constants etc. and 2) to catch any attempts to free it in
exp_pdu_open().
Make the callers free it after exp_pdu_open() returns.
(Alternatively, we could have exp_pdu_open() take the file name argument
and generate the comment itself, so that all code paths generate the
same comment.)
Change-Id: I6e6924b05565761b641a6c3b4d9a2e97f4264e1b
Ping-Bug: 15365
Reviewed-on: https://code.wireshark.org/review/31105
Reviewed-by: Guy Harris <guy@alum.mit.edu>
Have a ws_init_version_info() routine that, given an application name
string:
constructs the app-name-and-version-information string, and
saves it;
adds the initial crash information on platforms that support it,
and saves it.
Have show_version() use the saved information and take no arguments.
Add a show_help_header() routine to print the header for --help
command-line options, given a description of the application; it prints
the application name and version information, the description, and the
"See {wireshark.org URL}" line.
Use those routines in various places, including providing the
"application name" string in pcapng SHBs.
Change-Id: I0042a8fcc91aa919ad5c381a8b8674a007ce66df
Reviewed-on: https://code.wireshark.org/review/31029
Petri-Dish: Guy Harris <guy@alum.mit.edu>
Tested-by: Petri Dish Buildbot
Reviewed-by: Guy Harris <guy@alum.mit.edu>
That means that code is only in one place, rather than having copies of
it in each of those programs.
CLI programs that, on Windows, should get UTF-8 arguments rather than
arguments in the local code page should:
include the top-level cli_main.h header;
define the main function as real_main();
be built with the top-level cli_main.c file.
On UN*X, cli_main.c has a main() program, and just passes the arguments
on to real_main().
On Windows, cli_main.c has a wmain() function that converts the UTF-16
arguments it's handed to UTF-8 arguments, using WideCharToMultiByte() so
that it doesn't use any functions other than those provided by the
system, and then calls real_main() with the argument count and UTF-8
arguments.
Change-Id: I8b11f01dbc5c63fce599d1bef9ad96cd92c3c01e
Reviewed-on: https://code.wireshark.org/review/31017
Petri-Dish: Guy Harris <guy@alum.mit.edu>
Tested-by: Petri Dish Buildbot
Reviewed-by: Guy Harris <guy@alum.mit.edu>
1) The default build configuration is to select PCAPNG as
output format, but it can be selected as PCAP. Some of the
command line tools have the option to select the output
format and default towards the build configuration.
This has to be reflected in their help output also.
2) Various documentation files are still stating that PCAP is
the default format of various tools. With the default build
configuration being PCAPNG these have to be adjusted as well.
(with lack of dynamic content the documentation can only refer
to the default build configuration format).
Change-Id: I51d19642a7ed8c99817971c1f25d20972095021e
Signed-off-by: Jaap Keuter <jaap.keuter@xs4all.nl>
Reviewed-on: https://code.wireshark.org/review/30951
Petri-Dish: Anders Broman <a.broman58@gmail.com>
Tested-by: Petri Dish Buildbot
Reviewed-by: Anders Broman <a.broman58@gmail.com>
Add a new secrets API to the core, one that can outlive the lifetime of
a single capture file. Expose decryption secrets from wiretap through a
callback and let the secrets API route it to a dissector.
Bug: 15252
Change-Id: Ie2f1867bdfd265bad11fc58f1e8d8e7295c0d1e7
Reviewed-on: https://code.wireshark.org/review/30705
Petri-Dish: Peter Wu <peter@lekensteyn.nl>
Tested-by: Petri Dish Buildbot
Reviewed-by: Anders Broman <a.broman58@gmail.com>
The (optional) JSON-GLib library adds dependencies on GObject, GIO. For
statically linked oss-fuzz builds it also adds libffi and more. To avoid
these dependencies, replace JSON-GLib by some custom code. This allows
`tshark -G elastic-mapping` to be enabled by default without extra deps.
API design goals of the new JSON dumper library:
- Small interface without a lot of abstraction.
- Avoid memory allocations if possible (currently none, but maybe
json_puts_string will be replaced to improve UTF-8 support).
- Do not implement parsing, this is currently handled by jsmn.
Methods to open/close array/objects and to set members are inspired by
the JsonGlib interface. The interfaces to write values is inspired by
the sharkd code (json_puts_string is also borrowed from that).
The only observed differences in the tshark output:
- JSON-GLib ignores duplicates, json_dumper does not and may produce
duplicates and currently print two "ip.opt.sec_prot_auth_unassigned".
- JSON-GLib adds a space before a colon (unimportant formatting detail).
- (Not observed, but UTF-8 strings will be wrong like bug 14948.)
A test was added to catch changes in the tshark output. I also fuzzed
json_dumper with libFuzzer + UBSAN/ASAN and fixed an off-by-one error.
Change-Id: I0c85b18777b04d1e0f613a3d59935ec59be87ff4
Link: https://www.wireshark.org/lists/wireshark-dev/201811/msg00052.html
Reviewed-on: https://code.wireshark.org/review/30732
Petri-Dish: Peter Wu <peter@lekensteyn.nl>
Tested-by: Petri Dish Buildbot
Reviewed-by: Anders Broman <a.broman58@gmail.com>
Uli Heilmeier