[torquedev] [patch] bind to ip on multihomed pbs_servers

Matti Bickel cat5 at minet.uni-jena.de
Mon Feb 11 13:09:48 MST 2008


Adrian Knoth <adi at drcomp.erfurt.thur.de> wrote:
> Last Friday, I've talked to Matti. He's pretty overloaded at the moment,
> exams, work, girlfriend...

Well, i expect that to change slightly after next week, so stay tuned.

> Anyway, here's what he has done so far:
> 
>    http://cluster.inf-ra.uni-jena.de/~adi/torque-r1542-ipv6.diff
> 
>    (based on r1542 checkout from the torque subversion trunk)

Not quite. 1542 was what i used for the initial checkout. I synced with
the sources using git and branches once in a while, keeping pace for
about a month. However trunk just ran away as my time did :/

> It doesn't even compile, but it gives you an impression of required
> changes. It's far behind the trunk, so the best would be to make this
> public, an open process.

I really hope to get you a compiling, completely broken and unfunctional
version by the end of february. Sounds promising, doesn't it?

> I would call this a dirty hack, it doesn't clean things up, the #ifdefs
> make it even worse.

No other chance of winning anything but ignorance if i push the ipv6
code over ipv4 users expecting their torque to be "just the same", i
guess.
I'll welcome anything to do away with the ugliness of the patch. I'll
concede right away that forcing ipv6 into torque has caused major pains,
both for the code and for me and i guess for the innocent bystanders
looking at the patch...

> What does the community say? Add IPv6? Merge it into trunk, perhaps
> breaking it for days? Rewrite from scratch? Completely start over with
> the IPv6 patch?

I'd say let it go into a regularly synced branch torque-ipv6.

> I don't see no use in letting Matti go on coding on r1542+ipv6, on the
> other hand, I believe he has failed to communicate with the developers
> due to his illusion to come up with the minimal, perfect patch, solving
> everything.

Indeed, i suggested providing something that at least compiles. It has
proven to be quite unrealistic, b/c i started with porting the libnet
functions, breaking the internal API horribly.

I started with the idea of "test first" in mind, but failed completly on
that matter as soon as i realized how big the patch would become and in
how many ways i would have to test the code.
It remains a wishlist item to have functioning unit tests for at least
the libnet code, and i will start working on that after february.

> It's even a shame that I have to present his results, but at least
> that's better than throwing it all away.

Thanks for taking the time i didn't had.
-- 
Regards, Matti Bickel
Signed/Encrypted email preferred (key 4849EC6C)
-------------- 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/20080211/dca72710/attachment.bin


More information about the torquedev mailing list