[torquedev] Re: [torqueusers] Torque and IPv6

Garrick Staples garrick at usc.edu
Thu Aug 16 10:12:37 MDT 2007


On Thu, Aug 16, 2007 at 05:55:51PM +0200, Matti Bickel alleged:
> Garrick Staples <garrick at usc.edu> wrote:
> > Not currently.  There was some work started and I even created an ipv6 branch
> > in svn, but that worked seems to have stalled after not to long.  Now the ipv6
> > branch is quite stale.
> 
> I've downloaded a full checkout and i'm currently looking into it,
> estimating what it would take to port torque to use ipv6.
> I've two further question:
> 
> Are there any coding conventions jotted down? From pbs_ifl.h i get the
> impression you're still supporting K&R style C? Is there any system
> *that* ancient still in operation or are there other constraints that i
> have to be aware of?

I don't really worry about K&R anymore, only to follow convention in an
existing file or area for readability.

You'll find two whitespace styles throughout the code: the older OpenPBS style,
and the newer CRI style.  New code should follow the CRI style.


> What kind of unit/acceptance testing do you do, if at all? I've found
> some perl code in trunk/t. I'm very much planning to include unit tests
> of the ipv6 stuff. Do you have any preference wrt a unit testing
> framework?

We don't any such framework.  The perl scripts are for verifiying and
troubleshooting installation problems.

I have a test cluster that I use for new code, but nothing formal or strict.


> I've also noticed a few things in relation to the networking code:
>  * torque uses pbs_net_t, which is defined as unsigned long
>   - this is not guaranteed to evaluate to a 4 byte long structure, needs
>   to replaced by addr_info struct

I've always wondered about that.  I don't know why the original OpenPBS
developers used a type that was the same length or longer than the correct
type.  I assume it pre-dates a standard.

Obviously any ipv6 code needs to move to addr_info.  If pbs_net_t goes away, I don't think anyone will shed any tears.


>  * Garrick, in your ipv6 branch you did #ifdef ENABLE_IPV6 to include
>    ipv6 code, i gather my code should follow that convention?

Yes, please, at least for now.

 
> I'm currently doing cursory searches through the code alongside using
> Suns socket scrubber for analysis. So it's more than likely i'll come
> back with questions as they turn up.
> 
> I've created a private branch in my checkout from todays trunk, so i can
> work on the implementation. I'd like to share that branch publically, so
> it gets maximum exposure with the short integration cycles i'm used to
> from extreme programming. How's write access to the repo handled here?

Get some code out, we'll review, and get you write access.

 
> Thanks for reading through this, looking forward to gettin' my hands
> dirty ;)
> -- 
> Regards, Matti Bickel
> Encrypted/Signed Email preferred

My email is signed for your pleasure!

-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: not available
Url : http://www.supercluster.org/pipermail/torquedev/attachments/20070816/9dc2d7e0/attachment.bin


More information about the torquedev mailing list