The startup timeout on Win32 is reduced to 80% without assembler and to 50% with assembler usage (which is optional)
proto.c
- do not look up in filed tree and inserts in two steps but do it at once
- next few small speedups
- some often called elementary functions can be optionally implemented in assembler
- dispart some functions to see more exact result from profiling
packet-tpnc.c
- do not reallocate memory for each filed
svn path=/trunk/; revision=23643
est. Use g_ascii_strcasecmp() and g_ascii_strncasecmp(), and supply our
own versions if they're missing from GLib (as is the case with GLib
1.x).
In the code to build the list of named fields for Diameter, don't use
g_strdown(); do our own g_ascii_-style upper-case to lower-case mapping
in the hash function and use g_ascii_strcasecmp() in the compare
function.
We do this because there is no guarantee that toupper(), tolower(), and
functions that use them will, for example, map between "I" and "i" in
all locales; in Turkish locales, for example, there are, in both
upper case and lower case, versions of "i" with and without a dot, and
the upper-case version of "i" is "I"-with-a-dot and the lower-case
version of "I" is "i"-without-a-dot. This causes strings that should
match not to match.
This finishes fixing bug 2010 - an earlier checkin prevented the crash
(as there are other ways to produce the same crash, e.g. a bogus
dictionary.xml file), but didn't fix the case-insensitive string matching.
svn path=/trunk/; revision=23623
(where the initial length isn't readily available when item is first added)
Note that this still won't work where an initial length of 0 is given for
the item that will later be extended using proto_item_set_len(), as the
pointer value part of the zero-length array will reamin NULL...
svn path=/trunk/; revision=23253
packets in the Packet Details View.
This "appendix" bytes are not copied with the Copy functions or in the
Export Selected Packet Bytes.
svn path=/trunk/; revision=22887
routines and routines using those routines. GLib might use different
modifiers for 64-bit quantities than the platform's C library does.
svn path=/trunk/; revision=21990
In all the places where a cast to "long long" or "unsigned long long"
was done, use G_GINT64_MODIFIER and get rid of the cast, as
1) there's no guarantee that "%ll" works
and
2) there's no guarantee that "long long" works
(the latter definitely does *NOT* work with MSVC++; the former doesn't
work with regular printf in MSVC++, but it might work with the GLib
printf-based functions).
svn path=/trunk/; revision=21978
Fix compilation failures when building wireshark-0.99.6-SVN-21916 on an
x86_64-unknown-linux-gnu target with gcc version 4.1.2 20070403 (Red Hat
4.1.2-8).
The failures fall into two categories:
(1) Casts between pointers and 32-bit integers without an intermediary cast
via 'long' or 'unsigned long'. This results in a compiler warning complaining
about casts between a pointer and an integer of a different size.
(2) Passing values to "%lld" or similar printf-style format options that the
compiler thinks are a different size. Such values need to be cast to 'long
long' or 'unsigned long long'.
svn path=/trunk/; revision=21975
epan/filesystem.c
have get_plugin_dir() calling init_plugin_dir() if necessary
epan/epan.c and epan/report_err.c
move the report_failure family into the new report_err.c file, have epan_init() calling the initializer
epan/plugins.h and epan/proto.c
do not have init_plugins() calling the proto_reg functions instead do it in init_proto()
gtk/main.c and tshark.c
init_plugin_dir() has become suprefluous
capinfos.c and editcap.c
load the wiretap plugins
Makefiles
do what's needed to build withe the above changes.
svn path=/trunk/; revision=21935
The splash screen shows a progress bar and a percentage complete - like the progress dialog.
As dissectors are initialised and handed off the name is shown. However, the names of plugin dissectors are not shown.
The update to the make-dissector-reg shell script has been tested, though I think generally the python version is used.
svn path=/trunk/; revision=21716
proto_tree_add_bits_ret_val()
tvb_get_bits()
And modify
proto_tree_add_bits() not to return a value.
little endian is not yet implemented.
svn path=/trunk/; revision=21607
starting at the bit offset given for the number of bits indicated which wll also return
the value of the bits.
Experimental and for review, documentation to be updated.
svn path=/trunk/; revision=21556
* ptvcursor_push_subtree(), ptvcursor_pop_subtree() for pushing/popping
subtrees. Multiple levels of subtrees (256 max.), allocation per 8 levels.
* Two new functions creating an item in the tree and pushing a subtree at the
same time. These two functions accept an undefined length
(SUBTREE_UNDEFINED_LENGTH). The length of the item is set at the next pop.
1) ptvcursor_add_with_subtree
2) ptvcursor_add_text_with_subtree
- get rid of potential memory leaks with g_new in ptvcursor_new().
- Documentation of the new ptvcursor functions in README.developer
svn path=/trunk/; revision=21276
http://bugs.wireshark.org/bugzilla/show_bug.cgi?id=552
by enforcing that header fields have names of length > 0. This should fix
the display of those fields and also make them filterable (which was the
subject of the bug). Abbreviations are (still) optional: if they are empty
then the field is not filterable.
Update README.developer with this information.
Add header field names in several dissectors where they were missing.
In packet-arp.c give "packet-storm-detected" a name (as above) but also set it
as _GENERATED.
Also remove trailing white space from all the files checked in.
svn path=/trunk/; revision=21018
Here is an updated patch for proto_tree_add_item and the
range_string structure. The new macro RVALS() can be used as the macro
VALS() in the declaration of your hf_register_info with another
structure (range_string). Be aware that you *have to* ORed the value of
the field display with BASE_RANGE_STRING constant and it can 'only' be
used with FT_(U)INT* types in a header_field_info.
svn path=/trunk/; revision=20805
exception rather than aborting the program; using it means that
dissector bugs show up as such rather than as malformed packets.
svn path=/trunk/; revision=20532
32-bit numbers. Separate signed and unsigned accessors have been
added and used where appropriate.
Definitely not for 0.99.5.
svn path=/trunk/; revision=20472
if set, and if the program isn't running with additional privileges,
it'll treat the directory in which the program is found as the data
directory.
If, on Windows, the version-number subdirectory of {data
directory}\plugins doesn't exist (which is assumed to mean that the
program is being run from the build directory), or if, on UN*X,
WIRESHARK_RUN_FROM_BUILD_DIRECTORY is set, the plugin directory is the
"plugins" subdirectory of the data directory, and all subdirectories of
that directory are scanned for plugins, as the "plugins" subdirectory of
the build directory contains subdirectories for the plugins; this means
that if we're running from the build directory, we'll find the plugins
we built in the build tree.
When generating the wireshark-filter man page, run tshark with
WIRESHARK_RUN_FROM_BUILD_DIRECTORY set, so it uses the plugins from the
build to generate the list of filters.
svn path=/trunk/; revision=20261
This should mean that all fvalue_set() for FT_STRING[Z] are always with already_copied==FALSE
(funny that we never saw someone trying to g_free("[ Null ]") which might have happened before)
svn path=/trunk/; revision=20245
this primarily removes code and simplifies (==eliminates) the need to track the data that is allocated and should potentially be slightly faster than a slab allocator.
however these functions are called A LOT so there might be a performance hit when using emem with full debugging canary values and all the bells and whistles activated.
this change also makes any future attempt to parallellize dissection of frames easier if we just make the ep allocator allocate from a threads specific ep pool.
(something we would have to do anyway to make ep allocations multithreaded)
this works in all my tests so far but needs more test coverage.
svn path=/trunk/; revision=20194
proto_can_match_selected() routines, to more clearly separate the two
functions - but have them both call the same underlying routine, so
they both make the same decisions as to whether a match-selected string
can be constructed or not.
svn path=/trunk/; revision=19976
proto_construct_match_selected_string() to indicate what it does - and
have it return a Boolean indication of whether the string could be
built, returning the string through a pointer, and, if that pointer is
null, have it just return the Boolean and not construct the string.
Get rid of proto_can_match_selected() -
proto_construct_match_selected_string() can be used for that, which
means we have only one piece of code that knows whether a "match
selected" string can be constructed or not.
Have proto_construct_match_selected_string() support matching
zero-length FT_NONE (and FT_PCRE, but that shouldn't happen) fields even
if there's no epan_dissect_t, as such a match just checks whether the
field is present.
svn path=/trunk/; revision=19967
checks that we do when we try to construct the filter expression for
"match selected" - this means we don't just assert that all FT_NONEs are
filterable, as they aren't.
svn path=/trunk/; revision=19961
proto_construct_dfilter_string() the default, so you add explicit cases
only when the type needs to be treated specially, so we don't end up
with types where we forget to have a case.
svn path=/trunk/; revision=19959