Re: [tcpdump-workers] Build and check on a big-endian system

2014-02-16 Thread Guy Harris

On Feb 16, 2014, at 10:58 AM, François-Xavier Le Bail fx.leb...@yahoo.com 
wrote:

 From: Guy Harris g...@alum.mit.edu
 
 On Feb 15, 2014, at 1:21 PM, Michael Richardson m...@sandelman.ca wrote:
 
 The ESP tests are failing because you haven't got libssl-dev.
 
 Yes - some tests fail if the full set of support libraries weren't available 
 when tcpdump was built.
 
 Perhaps we want to suppress some tests if we don't have the appropriate 
 libraries - or have the tests compare against the results of a tcpdump built 
 without the library or libraries in question rather than the results of one 
 built with them.
 
 
 We could maybe add a text in INSTALL file, like:
 
 diff --git a/INSTALL.txt b/INSTALL.txt
 index 132e118..4dc61d3 100644
 --- a/INSTALL.txt
 +++ b/INSTALL.txt
 @@ -9,6 +9,9 @@ You will need an ANSI C compiler to build tcpdump. The 
 configure
  will abort if your compiler is not ANSI compliant. If this happens, use
  the generally available GNU C compiler (GCC).
  
 +To build with crypto support, you need SSL library (libssl-dev).
 +If you don't have this library, you will have some make check errors.

Denis appears to have fixed that with

commit 9e2042ebe86c41d373930dbd89b56a6a2f06fcf0
Author: Denis Ovsienko infrastat...@yandex.ru
Date:   Mon Feb 17 00:22:37 2014 +0400

make OpenSSL-specific tests conditional

That still leaves the NFLOG tests; that requires looking for DLT_NFLOG in the 
version of libpcap being used to build tcpdump, which would be a bit more work.

An alternative might be to have the configure script run a test for DLT_NFLOG 
and somehow do something to make make check skip the NFLOG test.  A 
generalization might be to have some support for if ... end in the 
tests/TESTLIST file, with the tests checking some name specified in a file, and 
have the configure script add entries to that file as necessary.
___
tcpdump-workers mailing list
tcpdump-workers@lists.tcpdump.org
https://lists.sandelman.ca/mailman/listinfo/tcpdump-workers


Re: [tcpdump-workers] Build and check on a big-endian system

2014-02-15 Thread Michael Richardson

François-Xavier Le Bail fx.leb...@yahoo.com wrote:
 At present, the Travis test do a x86_64 build and a 'make check'.

 So the checks are done on a little-endian system.

 He could be useful to add a build and check on a big-endian system.

 What do you think about it?

I would love it.
As far as I know, Travis does not have any big-endian systems.
(Nor is it particularly easy to install the right 32-bit x86 libraries on the
same travis instance, because a few of the devel libraries like libusb, have
conflicts between the 64-bit and 32-bit packages on ubuntu/debian)

Do you?

The wireshark folks have buildbot infrastructure, and we were suppposed to
set up a buildbot server, but it never got done.
If you wanted to take that on, it would be very much appreciated.

If you are at IETF89 London, I could set up accounts for you on bpf, but
really the buildbot server could run anywhere.

--
]   Never tell me the odds! | ipv6 mesh networks [
]   Michael Richardson, Sandelman Software Works| network architect  [
] m...@sandelman.ca  http://www.sandelman.ca/|   ruby on rails[


___
tcpdump-workers mailing list
tcpdump-workers@lists.tcpdump.org
https://lists.sandelman.ca/mailman/listinfo/tcpdump-workers


Re: [tcpdump-workers] Build and check on a big-endian system

2014-02-15 Thread François-Xavier Le Bail
 From: Michael Richardson m...@sandelman.ca

 François-Xavier Le Bail fx.leb...@yahoo.com wrote:
      At present, the Travis test do a x86_64 build and a 'make 
 check'.
 
      So the checks are done on a little-endian system.
 
      He could be useful to add a build and check on a big-endian system.
 
      What do you think about it?
 
 I would love it.
 As far as I know, Travis does not have any big-endian systems.
 (Nor is it particularly easy to install the right 32-bit x86 libraries on the
 same travis instance, because a few of the devel libraries like libusb, have
 conflicts between the 64-bit and 32-bit packages on ubuntu/debian)
 
 Do you?

I did not test, but it seems promising :
http://www.aurel32.net/info/debian_mips_qemu.php

 The wireshark folks have buildbot infrastructure, and we were suppposed to
 set up a buildbot server, but it never got done.
 If you wanted to take that on, it would be very much appreciated.
 
 If you are at IETF89 London, I could set up accounts for you on bpf, but
 really the buildbot server could run anywhere.

Greetings,
Francois-Xavier

___
tcpdump-workers mailing list
tcpdump-workers@lists.tcpdump.org
https://lists.sandelman.ca/mailman/listinfo/tcpdump-workers


Re: [tcpdump-workers] Build and check on a big-endian system

2014-02-15 Thread Michael Richardson

François-Xavier Le Bail fx.leb...@yahoo.com wrote:
 Do you?

 I did not test, but it seems promising :
 http://www.aurel32.net/info/debian_mips_qemu.php

so, you are saying we should install a qemu mips emulator inside travis to do
the testing... wow.. fork the tree, enable travis, and try it out...

--
]   Never tell me the odds! | ipv6 mesh networks [
]   Michael Richardson, Sandelman Software Works| network architect  [
] m...@sandelman.ca  http://www.sandelman.ca/|   ruby on rails[



___
tcpdump-workers mailing list
tcpdump-workers@lists.tcpdump.org
https://lists.sandelman.ca/mailman/listinfo/tcpdump-workers


Re: [tcpdump-workers] Build and check on a big-endian system

2014-02-15 Thread François-Xavier Le Bail
 From: Michael Richardson m...@sandelman.ca

 so, you are saying we should install a qemu mips emulator inside travis to do
 the testing... wow.. fork the tree, enable travis, and try it out...

FYI, digging more, I used this link:
http://people.debian.org/~aurel32/qemu/powerpc/
with the file: debian_wheezy_powerpc_standard.qcow2

The start and build process is at present slow (for my first test).
Don't know if we can have a better response time.

# qemu-system-ppc --version
QEMU emulator version 1.1.2 (Debian 1.1.2+dfsg-6a)
# qemu-system-ppc -hda debian_wheezy_powerpc_standard.qcow2 -nographic
inside : apt-get : gcc make git flex bison autoconf
I build and check (slow!). result:
root@debian-powerpc:~/tcpdump# make check 21 | tee trace
(cd tests  ./TESTrun.sh)
    lmp-v : passed
    print-x   : passed
    print-xx  : passed
    print-X   : passed
    print-XX  : passed
    print-A   : passed
    print-AA  : passed
    bgp_vpn_attrset   : passed
    mpbgp-linklocal-nexthop   : passed
    eapon1    : passed
    esp0  : passed
    esp1  : TEST FAILED
    esp2  : TEST FAILED
    esp3  : TEST FAILED
    esp4  : TEST FAILED
    esp5  : TEST FAILED
    espudp1   : TEST FAILED
    isakmp1   : passed
    isakmp2   : passed
    isakmp3   : passed
    isakmp4   : passed
    lmp   : passed
    mpls-ldp-hello    : passed
    ospf-gmpls    : passed
    ikev2four : passed
    ikev2fourv    : passed
    ikev2fourv4   : passed
    ikev2pI2  : TEST FAILED
    dio01 : passed
    dio02 : passed
    dio03 : passed
    dao01 : passed
    daoack01  : passed
    e1000g    : passed
    forces01  : passed
    forces01vvv   : passed
    forces01  : passed
    qinq  : passed
    qinqv : passed
    sflow1    : passed
    sflow2    : passed
    ahcp-vv   : passed
    babel1    : passed
    babel1v   : passed
    babel_auth    : passed
    babel_pad1    : passed
    pppoe : passed
    pppoes    : passed
    pppoes_id : passed
    igmpv3-queries    : passed
    icmpv6    : passed
    icmpv6_opt24-v    : passed
    spb   : passed
    spb_bpduv4    : passed
    dcb_ets   : passed
    dcb_pfc   : passed
    dcb_qcn   : passed
    evb   : passed
    ripv1v2   : passed
    ripv2_auth    : passed
    dhcpv6-aftr-name  : passed
    dhcpv6-ia-na  : passed
    dhcpv6-ia-pd  : passed
    dhcpv6-ia-ta  : passed
    dhcpv6-ntp-server : passed
    dhcpv6-sip-server-d   : passed
    dhcpv6-domain-list    : passed
    zmtp1v    : passed
    pgmv  : passed
    epgmv : passed
    pgm_zmtp1v    : passed
    epgm_zmtp1v   : passed
    msnlb : passed
    msnlb2    : passed
    mptcp : passed
    mptcp-fclose  : passed
    tfo   : passed
    802.11_exthdr : passed
    802.11_rx-stbc    : passed
    of10_p3295-vv : passed
    of10_s4810-   : passed
    of10_pf5240-vv    : passed
    of10_7050q-v  : passed
    geonet-calm-fast  : passed
    nflog-e   : passed
    syslog-v  : passed
    ipv6-bad-version.pcap : passed
    loopback  : passed
    dccp_partial_csum_v4_simple   : TEST FAILED
    dccp_partial_csum_v4_longer   : TEST FAILED

   9 tests failed
  81 tests passed


Failed test: esp1

1,8c1,8
 IP 192.1.2.23  192.1.2.45: ESP(spi=0x12345678,seq=0x1), length 116: IP 
192.0.2.1  192.0.1.1: ICMP echo request, id 28416, seq 1280, length 64 
(ipip-proto-4)
 IP 192.1.2.23  192.1.2.45: 

Re: [tcpdump-workers] Build and check on a big-endian system

2014-02-15 Thread François-Xavier Le Bail
 From: Michael Richardson m...@sandelman.ca


 The ESP tests are failing because you haven't got libssl-dev

Right, thanks !
remain :
    dccp_partial_csum_v4_simple   : TEST FAILED
    dccp_partial_csum_v4_longer   : TEST FAILED
___
tcpdump-workers mailing list
tcpdump-workers@lists.tcpdump.org
https://lists.sandelman.ca/mailman/listinfo/tcpdump-workers


Re: [tcpdump-workers] Build and check on a big-endian system

2014-02-15 Thread Guy Harris

On Feb 15, 2014, at 1:21 PM, Michael Richardson m...@sandelman.ca wrote:

 The ESP tests are failing because you haven't got libssl-dev.

Yes - some tests fail if the full set of support libraries weren't available 
when tcpdump was built.

Perhaps we want to suppress some tests if we don't have the appropriate 
libraries - or have the tests compare against the results of a tcpdump built 
without the library or libraries in question rather than the results of one 
built with them.
___
tcpdump-workers mailing list
tcpdump-workers@lists.tcpdump.org
https://lists.sandelman.ca/mailman/listinfo/tcpdump-workers