general packet analyzing, but useful when developing wiretap code or
other such low-level development.
Default preference is *not* to show File Offset.
svn path=/trunk/; revision=4120
Frascone.
Small white-space fix.
Display the preference level in router advertisements as signed, not
unsigned, as per RFC 1256, which says it's a "signed, twos-complement
value".
svn path=/trunk/; revision=4118
that they're external when the plugin is compiled, and there's no
definition to satisfy the extern, so plugins fail to compile.
Add a "plugins/plugin_api_defs.h" header file that defines them, and
include that in one (and only one) source file for a particular plugin.
svn path=/trunk/; revision=4117
and update the URL for the DHCP parameter value assignments; support a
code of 90 for DHCP authentication, as that's what RFC 3118 specifies.
Show symbolic values for some of the fields in the DHCP authentication
option, and base the decision on whether to dissect the authentication
data as HMAC-MD5 data on whether that's what the protocol and algorithm
field specified. Show the replay detection value as a 64-bit number
only if it's specified to be a monotonically-increasing counter;
otherwise, just show it as bytes.
svn path=/trunk/; revision=4116
It makes no difference if they really are function declarations;
however, in plugins, when building on OSes that don't let
dynamically-loaded modules access functions in the main program (e.g.,
Windows), when compiling a plugin, <plugin_api.h> defines the names of
those functions as (*pointer_name), so they turn into declarations of
pointer variables pointing to the functions in question, and, on
platforms with a def/ref model in the linker, if a plugin has more than
one source file that gets linked into the plugin, the linker may get
upset at two definitions of the same variable.
svn path=/trunk/; revision=4114
IP addresses - either use "tvb_memcpy()", or just use
"proto_tree_add_item()" and leave the fetching up to the core Ethereal
code.
In the DVMRP dissector, replace a bunch of other "proto_tree_add_XXX()"
calls with "proto_tree_add_item()", and use
"tvb_reported_length_remaining()" rather than "tvb_length_remaining()"
(if we didn't capture enough data because the snapshot length was too
low, we *want* an exception to be thrown, so that the frame is marked as
a "Short Frame").
svn path=/trunk/; revision=4109
little-endian byte order.
Use "proto_tree_add_item()" wherever possible.
Get rid of line commented out with a C++ comment, as not all C compilers
accept C++ comments.
Don't put the user name into the protocol tree if it's not there (we
really should do TCP segment reassembly for this, but this is at least a
good first cut).
When checking whether data exists in the packet, use
"tvb_offset_exists()", don't compare the offset in the tvbuf with
"pinfo->len" - "pinfo->len" is the length of the entire packet, not of
the tvbuff.
Fix some references to "pi" to refer to "*pinfo" instead.
svn path=/trunk/; revision=4108
If the byte being investigated is 0, don't just do a "continue";
this might be one of the passes on which we have to propagate
carries, so skip the processing of that byte's bits, but fall
through to the carry processing.
In the carry processing loop, don't use "i" as the loop index,
as we're already using it as the outer loop index.
In "u64toh()", use lower-case letters rather than upper-case letters, as
we use "%x" rather than "%X" to print 8, 16, 24, and 32-bit integral
values in the "proto.c" code.
svn path=/trunk/; revision=4103
former depends on having "guint64" and the latter depends on
"%ll[douxX]" being what's used to print 64-bit integers, and there are
platforms on which Etheeal runs that don't have "guint64" or that don't
use "%ll[douxX]" to print 64-bit integers.
Get rid of the routines to extract 64-bit integers into "gint64"s and
"guint64"s, as per Ronnie Sahlberg's suggestion, to discourage people
from writing code that won't work on all platforms; they should be using
FT_UINT64, or the routines in "int-64bit.c", instead.
svn path=/trunk/; revision=4102
without requiring compiler support for them, and updates to the
Diameter, L2TP, NFS, and NLM dissectors to use it and to the ONC RPC
dissector to allow ONC RPC subdissectors to use it.
svn path=/trunk/; revision=4099
"snprintf()" returns a negative number, that's an error, and we assume
"errno" was set and return NULL, otherwise we cast its return value to
"size_t" and compare it with the size of the buffer we were given, and,
if it was bigger, we know that "snprintf()" didn't generate all the
characters it could be cause they wouldn't have fit, so we set "errno"
to ENOSPC and return NULL.
svn path=/trunk/; revision=4095
doesn't exist, or is out of date with respect to "config.h.win32", it's
remade - stuff in "ftypes" and "dfilter" includes "config.h", and it
should get the "config.h" in "epan".
svn path=/trunk/; revision=4091
that, when using cryptographic authentication, "The checksum field in
the standard OSPF header is not calculated, but is instead set to 0";
treat a packet checksum value of 0 as meaning "no checksum present".
svn path=/trunk/; revision=4090
there were 2 functions which accepted 'maxlength' == -1, but the function
prototypes had maxlength as a guint --- fixed.
svn path=/trunk/; revision=4087
pass -1 in the cases where it's not called from "dissect_proposal()",
i.e. where there *is* no protocol ID to pass it, and have it dissect
the transform ID only as a number if the protocol ID isn't one of the
ones we know about.
Give the payload dissectors other than "dissect_transform()" an extra
"int" argument, so that their signature is the same as that of
"dissect_transform()", put "dissect_transform()" back in the
"strfuncs[]" table, and get rid of the special-casing of
"dissect_transform()" in "dissect_payloads()".
svn path=/trunk/; revision=4086
formatting the string as text.
In "add_pointer_param()", make the offset in the text item for the
string the offset in the string area, not the offset of the pointer
itself.
Fix some items in item lists to have "lm_null" rather than a null
pointer, so that if we don't find an entry for the detail level, we
don't blow up dereferencing a null pointer, we just dissect all the
items based on the type in the descriptor string.
If we don't have a data descriptor for response data, just put in an
entry for the data; that may happen if we only dissected part of the
request to which the stuff we're dissecting is a reply, e.g. because the
snapshot length didn't let us capture all of the request.
svn path=/trunk/; revision=4085
common routine that loops over payloads.
Have that routine check for a payload of type "None", which means that
there's extra data after a payload that claimed to be the last one (by
virtue of having a next payload type of "None").
When dissecting a Security Association payload, make sure we have enough
data for the Domain of Interpretation field before putting it into the
tree, dissect the situation as a 4-byte quantity followed by a set of
sub-payloads only of the DOI is "IPSEC" (otherwise dissect it as raw
data), and make sure we have enough data for the 4-byte situation field
before putting it into the tree.
When dissecting a Proposal payload, show the raw bytes of the SPI.
svn path=/trunk/; revision=4081
- Handle "short" as a union discriminant type
- Handle case where union discriminant may be a typedef
- Change integer template to long template
- Initial code for \n and \t as union case labels.
svn path=/trunk/; revision=4076
Guy Harris