[ntar-workers] [Patch] tests automation

[Gianluca Varenni]

----- Original Message ----- 
From: "Sebastien Tandel" <sebastien at tandel.be>
To: <ntar-workers at winpcap.org>
Sent: Monday, January 15, 2007 4:15 PM
Subject: [ntar-workers] [Patch] tests automation


> Hi all,
>
> I tried ntar on a debian platform.
>
> There was only a simple error in error.c
> error.c : line 18. No eol : easy to solve by editing the file with vim
> but ... :)
> After this little step, the compilation on Debian was fine.

I have fixed the problem. I don't know why the error didnt show up on my 
linux distros...

>
> I made Makefiles for a basic tests automation system.
>
> *** make check ***
> will compile the testxxx if needed, execute the tests and check against
> testxxx.expected to see if it is correct. (until 16)
> testxxx.expected have been generated "by hand" the first time (not
> really but at least checked for the results).
>
> will output something like
> test001 [ PASS ]
> test002 [ FAIL ]
> test003 [ PASS ]
> test004 [ PASS ]
> test005 [ FAIL ]
> ...
>
>
> Note :
> 1) You first have to move testxxx_in.tar in their respective directories
> tests/testxxx, if needed.

No big deal. Actually we can have a script moving the binaries to the folder 
containing the testXXX_in.ntar  files.

> 2)  I really don't know if it will work on Windows as it uses diff.

Uhm... It can be a prerequisite, i'm not too much concerned about that. As a 
matter of facts, for almost all the projects i work on under Windows Cygwin 
is a prerequisite as I use a bunch of patching/diffing/scripting/...


> 3) As test0017 and test018 are waiting for user input, I did not
> integrate them in the tests automation.

It's fine. More than regression tests they are apps I used to measure the 
performance of the library on different platforms and hard drives.

What I'm more concerned about the about the automation system is that i 
think it will work on little endian systems only. The regression outputs I 
have created are little endian, on a big endian machine the file output is 
different. The library is able to deal with reading files in both 
endianness, but the generated files use the byte order of the host. The best 
approach would be to create different regression outputs for little- and 
big-endian machines, using the proper ones for the test automation.

In any case, good job! If you send me a unified diff with the makefile, I 
can apply it and commit it to the CVS.

I don't know if you have seen the discussion with Benn Bollay on the 
wireshark-dev mailing list, in any case I'm working in these weeks to add 
some new stuff to NTAR (some is already there, but experimental and not 
public). Namely better support for plugins (the APIs are there, the 
implementation is still not efficient, i need to fix the documentation and 
the mess i created with the include files), and support for seeks and 
backward reading in the file.

You can find the wireshark-dev thread here

http://www.wireshark.org/lists/wireshark-dev/200701/msg00173.html


>
>
> P.S. : The files permissions are generally set to 0710 which is not a
> good idea (don't know if it's a Windows feature but it's really ugly all
> these green text files on my screen :)).
>

Uhm... it can be an issue with how i repackaged the files in the tarball for 
*nix (it's totally possible i decompressed the zip on my Windows machine, 
and repackaged everything here). I'll do a CVS checkout on a linux distro 
and see if they are screwed up on the CVS as well.

Have a nice day
GV

>
>
> Regards,
>
> Sebastien Tandel
> _______________________________________________
> ntar-workers mailing list
> ntar-workers at winpcap.org
> https://www.winpcap.org/mailman/listinfo/ntar-workers