Recent changes to bugs

Switch from gethostbyname() to getaddrinfo()

Robert Scheck — Mon, 29 Jun 2015 18:23:49 -0000

Use of gethostbyname (in src/addr.c) may prevent the program from working properly on IPv6 systems. getaddrinfo should be used in preference.

Updated config.guess and config.sub scripts required

Tony — Fri, 26 Aug 2011 14:04:18 -0000

Hello,

When building on Debian kFreeBSD and DragonFly BSD, ./configure complains that it can't recognize the operating system and needs updated versions of config.guess|sub scripts [1]. Recent versions of automake(1.11) or autoreconf(2.68) will update these using the --force --install options, though I found it easier to simply use the latest versions from the git repo[2].

If you could update these, it would be much appreciated.

Thanks,

Tony

[1] http://www.gnu.org/s/hello/manual/automake/Auxiliary-Programs.html#Auxiliary-Programs
[2] http://git.savannah.gnu.org/gitweb/?p=config.git;a=tree

ip_checksum() segfaults when payload is 0 or 1 bytes

Anonymous — Mon, 24 Aug 2009 18:37:27 -0000

The ip_cksum_add() function (called by the exported function "ip_checksum()") has a bug: Duff's device, used here for loop unrolling, assumes at least 1 iteration.
In the case of the length of the payload being 0 or 1 bytes, the number of iterations will be 0. This will cause "cksum += *sp++" 16 times, whereas it should not have been executed at all. Thus, the code will walk off the end of the buffer, and possibly onto another memory page.

The worst is that a segfault will occur. The best is that an incorrect checsksum will be generated. The absolute best is if the 16 bytes of memory following the passed-in buffer are mapped, and zero, then the correct checksum will be generated.

The fix is to add a conditional around the switch() statement such that the cases are not executed unless len > 1.

missing prerequisite headers?

asgr — Mon, 26 May 2008 09:31:47 -0000

I'm trying to configure 1.11 on OSX 10.5.2, and experienced a number of warnings about missing headers.

configure: WARNING: net/if.h: present but cannot be compiled
configure: WARNING: net/if.h: check for missing prerequisite headers?
configure: WARNING: net/if.h: see the Autoconf documentation
configure: WARNING: net/if.h: section "Present But Cannot Be Compiled"
configure: WARNING: net/if.h: proceeding with the preprocessor's result
configure: WARNING: net/if.h: in the future, the compiler will take precedence

configure: WARNING: netinet/in_var.h: present but cannot be compiled
configure: WARNING: netinet/in_var.h: check for missing prerequisite headers?
configure: WARNING: netinet/in_var.h: see the Autoconf documentation
configure: WARNING: netinet/in_var.h: section "Present But Cannot Be Compiled"
configure: WARNING: netinet/in_var.h: proceeding with the preprocessor's result
configure: WARNING: netinet/in_var.h: in the future, the compiler will take precedence

configure: WARNING: netinet/ip_fw.h: present but cannot be compiled
configure: WARNING: netinet/ip_fw.h: check for missing prerequisite headers?
configure: WARNING: netinet/ip_fw.h: see the Autoconf documentation
configure: WARNING: netinet/ip_fw.h: section "Present But Cannot Be Compiled"
configure: WARNING: netinet/ip_fw.h: proceeding with the preprocessor's result
configure: WARNING: netinet/ip_fw.h: in the future, the compiler will take precedence

Is this libdnet, missing library or is it OSX related (does libdnet still work with Leopard as it looks as if Apple is using a new firewall app).

configure failed on Fedora Core 6

Anonymous — Sun, 19 Aug 2007 19:13:22 -0000

checking how to run the C++ preprocessor... /lib/cpp
configure: error: C++ preprocessor "/lib/cpp" fails sanity check

(as reported by ./configure && make)
BLAG60001 distro

mispackaged libdnet 1.12

Jeff Nathan — Tue, 03 Jul 2007 22:20:13 -0000

The 1.12 tarball has a copy of trunk inside of it. So, you end up with libdnet-1.12/src and libdnet-1.12/trunk/src . Funky!

No library suffix on SUSE 10.2

James Oakley — Wed, 06 Sep 2006 16:00:21 -0000

For some reason, when I compile libdnet on SUSE 10.2
(On my alpha system and on SUSE's buildservers), the
libraries have no suffix. Here's a snippet of make install
output:

mkdir -p -- /var/tmp/libdnet-root/usr/lib
/bin/sh ../libtool --mode=install /usr/bin/install -c
libdnet.la /var/tmp/libdnet-root/usr/lib/libdnet.la
/usr/bin/install -c .libs/libdnet.1.0.1 /var/tmp/libdnet-root/usr/lib/libdnet.1.0.1
(cd /var/tmp/libdnet-root/usr/lib && rm -f libdnet.1 && ln -s
libdnet.1.0.1 libdnet.1)
(cd /var/tmp/libdnet-root/usr/lib && rm -f libdnet && ln -s
libdnet.1.0.1 libdnet)
/usr/bin/install -c .libs/libdnet.lai /var/tmp/libdnet-root/usr/lib/libdnet.la
/usr/bin/install -c .libs/libdnet.a /var/tmp/libdnet-root/usr/lib/libdnet.a

Note the lack of '.so' in any of those filenames.

The libtool version is 1.5 22.

dnet 1.11 for python crashes under XP SP2

Anonymous — Tue, 11 Apr 2006 21:49:21 -0000

I have have had a number of crashes under Windows XP
with SP 2. My machine has a number of interfaces,
including 802.11, 802.3, and VMWare interfaces. Eth2 is
a connected Broadcom NetXtreme Gigabit adapter.

Here is the simplest crash in python 2.4.3

import dnet
for a in dnet.intf(): print a

Crash.

if I use pcap, I can find an interface:

import pcap

>>> a = pcap.pcap()
>>> print a.name
eth2
>>> print a.readpkts()
[(1144789999.9557691, <read-only buffer ptr 0x010AF2AC,
size 60 at 0x009EB1A0>),
(1144789999.96422, <read-only buffer ptr 0x010AF2FC,
size 60 at 0x009EB200>), ( ...

works fine.

I can't use the interface name with dnet to get an
interface:

>>> import dnet
>>> a = dnet.eth('eth2')

Crash.

In case it helps, here is the stack trace when it dies.
from windbg formated: frame #, three raw args, function
info

00 7c8633d5 00000002 0021f074 ntdll!KiFastSystemCallRet
(FPO: [0,0,0])
01 00000002 0021f074 00000001
ntdll!ZwWaitForMultipleObjects+0xc (FPO: [5,0,0])
02 0021f28c 00000000 00000000
kernel32!UnhandledExceptionFilter+0x82d (FPO: [Non-Fpo])
03 00000000 0021f28c 7c34246e MSVCR71!_XcptFilter+0x15f
(FPO: [Non-Fpo])
WARNING: Stack unwind information not available.
Following frames may be wrong.
04 00000014 00000000 7ffdf000 python+0x11d2
05 1d001062 00000000 78746341
kernel32!BaseProcessStart+0x23 (FPO: [Non-Fpo])

dnet.arp() provides a loop that gives me correct
arptable info.

Thanks,
Jesse

bug in intf-win32.c

Anonymous — Tue, 11 Apr 2006 01:06:03 -0000

static void
_ifcombo_add(struct ifcombo *ifc, DWORD idx)
{
if (ifc->cnt == ifc->max) {
if (ifc->idx) {
ifc->max *= 2;
ifc->idx = realloc(ifc->idx,
sizeof(ifc->idx[0] * ifc-
>max));
} else {
ifc->max = 8;
ifc->idx = malloc(sizeof(ifc-
>idx[0] * ifc->max));
}
}
ifc->idx[ifc->cnt++] = idx;
}

error:
ifc->idx = realloc(ifc->idx, sizeof(ifc->idx[0] * ifc-
>max));
ifc->idx = malloc(sizeof(ifc->idx[0] * ifc->max));

ok:
ifc->idx = realloc(ifc->idx,(sizeof(ifc->idx[0]) *
ifc->max));
ifc->idx = malloc(sizeof(ifc->idx[0]) * ifc->max);

configure fails on RH ES4

wkenzie — Mon, 27 Feb 2006 21:24:01 -0000

problem with __cplusplus

what details do you need in order to help me get
through this?