[Ns-developers] regression testing and native win32 (mingw)

[Mathieu Lacage]

On Wed, 2008-09-10 at 01:16 +0100, Gustavo Carneiro wrote:
> I know this is not a priority right now, just bear with me.
> 
> Currently all mingw regression tests fail:
> 
>   1. Ascii traces differ: this is due to an extra carriage return generated
> from win32 and not available in the reference traces.  The diff option
> --strip-trailing-cr takes care of it and then the files match (or at least
> the one I tested manually).  Unfortunately that option is not always
> available; the msys diff command doesn't have it.  What is it better, avoid
> std::endl when generating ascii traces (use plain old \n instead), or work
> around it when comparing the files?  The issue is raised, solution may come
> later;

Since this is windows, all you have to do is setup mercurial's win32
filters to do the unix<->windows conversion automatically. I suspect
that if you use the mercurial tortoisehg binary, it might even be setup
for you already.

>   2. pcap files also differ.  The pcap files generated in win32 do not open
> in either Linux or win32 Wireshark: "The capture file appears to be damaged
> or corrupt.
> (pcap: File has 36569088-byte packet, bigger than maximum of 65535)".  I
> uploaded [1]the win32 version in case anyone wants to take a look; but no
> pressure.. :-)
> [1] http://telecom.inescporto.pt/~gjc/csma-broadcast-0-0.pcap

Can you file a bug ? This does indeed sound like a real issue.

Mathieu