[Ns-developers] ns-3.6 Daily Bug Status for Wednesday, October 7, 2009

[craigdo@ee.washington.edu]

Dear All,

News and Comments:
==================

As they say, I have some good news and some bad news.

The good news is that we got George's patch into ns-3-dev tonight and we
have closed all of our P1 bugs according to the plan outlined yesterday.
Thanks, George and Josh!

The bad news is that more problems have come up.  See "Discussion" for
details.

Bugs:
=====

As of Wednesday night (PST), October 7, 2009, by my reckoning we are down to
zero P1 bugs filed against ns-3-dev.  This won't last long as I begin filing
bugs on the new issues tomorrow, though.

Discussion
==========

You probably noticed that the buildbots reported build problems earlier
tonight.  Thanks to Sebastien Vincent for noticing and fixing a problem with
Python bindings.  We also had an issue with an implicit type conversion
between double and unsigned int that has been fixed as well.  Linux builds
seem to be happier now.

The builds are still broken on darwin-ppc (osx-ppc-g++-4.0 and
osx-ppc-g++-4.2) though:

[384/845] cxx: src/common/pcap-writer.cc ->
build/debug/src/common/pcap-writer_1.o
In file included from debug/ns3/high-precision-128.h:24,
                 from debug/ns3/high-precision.h:29,
                 from debug/ns3/nstime.h:29,
                 from debug/ns3/simulator.h:27,
                 from ../src/common/pcap-writer.cc:31:
debug/ns3/cairo-wideint-private.h:67:2: error: #error Cannot find
definitions for fixed-width integral types (uint8_t, uint32_t, etc.)
Waf: Leaving directory
`/Users/buildslave/slave/full-darwin-ppc-g++-4.0/build/build'

We need to get these running again ASAP.

I haven't seen successful results for Cygwin and I haven't run the tests on
osx-intel so I'm not sure yet how they are faring.

The next glitch is that the ns3-tcp-cwnd TestSuite is failing since we
applied the TCP fixes.  This may represent a change to correct behavior
based on George's patch, but I want to take some time to verify this.  If I
run the test suite under -m (allow multiple failures) the suite crashes
which is a little troubling.

I have enabled valgrind under test.py (run using ./test.py -g) and this has
revealed some problems.

The two TCP test suites are crashing under valgrind:

  CRASH: TestSuite ns3-tcp-interoperability
  CRASH: TestSuite ns3-tcp-cwnd 

We also have a memory leak problem in one of the test suites:

  VALGR: TestSuite pcap-file-object

Three examples (apparently otherwise untested) show memory leak problems
too:

  VALGR: Example ipv6/icmpv6-redirect
  VALGR: Example wireless/simple-wifi-frame-aggregation
  VALGR: Example mesh/mesh

87 of 93 tests passed (87 passed, 0 failed, 2 crashed, 4 valgrind errors)

I'm afraid I'm getting old and 02h00 is my limit.  I am giving up for the
night.  If one of you kind folks on the other side of the world could step
up and take a look at some of these problems I would be eternally grateful.

Next Steps:
===========

Code freeze for ns-3.6 was planned for today, Wednesday, October 7, 2009.
This has obviously slipped, and ns-3.6-RC1 along with, until we can at least
get our product built on all of our supported platforms.

Regards,

-- Craig