[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: more (aranym) buildds needed



On Mon, Apr 07, 2008 at 12:29:44PM -0500, Stephen R Marenka wrote:

> > Currently I have a 2.8 Pentium4 with 2 GB RAM doing nothing. But it's not a
> > machine that's not supposed to be online 24/7. 
> buildds aren't very useful if they're not up 24/7. 

True. And that's only part of my dilemma... my dedicated roto server is
online 24/7 and has 2 GB RAM free for additional Xen VMs, but I ran out of
IPs at the moment and aranym would be needed to run headless... 

> Internet access is required for:
> 1) sid mirror - could work around with a local mirror, but that's a lot
> of disk space
> 2) incoming - could skip building out of incoming, but causes dependency
> headaches and give-backs if you're building off the top of the queue

I think I already used once a incoming mirror script that worked quite
well... 

> 3) w-b access - could manually queue packages, but it's a pain to
> maintain

Oh well, yes... how about having our own w-b again?

> 4) mail - easily batched

Mail is sometimes harder than it appears to be... (-;

> > But what's about the "removing packages"-idea like boost, flight simulators
> > or other heavy weight apps rarely used on m68k?
> I think we'll probably end up there, but so far all we have is hand
> waving as opposed to something concrete and dependency chain based.
> > I think getting aranym buildds up and running will be just an intermediate
> > solution. There are more and more packages to be, steadily increasing in
> > number, and it seems just a matter of time when all those aranym buildds
> > won't be enough again to keep up. 
> Maybe, but historically we stay caught up pretty well until we have a
> toolchain or other ugly dependency problem.

... and shortly before releases/freezes... ;)

> We just haven't recovered
> from the last one yet. (And massive binnmu's don't help, although a
> handful of aranym buildds could probably handle those hits.)

I could start aranym on that Pentium4 machine for some weeks (until mid/end
of May) if that helps. Then the machine will be relocated... 

> > Maybe we can get a natively built and uptodate core Debian for m68k and a
> > best-effort {stable}-m68k suite for other software (built by aranym
> > buildds)? And yes, it's difficult to tell which package should end up in
> > core or in {stable}-m68k... 
> Plus we move further away from stock debian. Not that I'm arguing, I
> think we'll have to find a way to do it sooner or later. Perhaps we'll
> point the direction for future debian changes. We've been talking about
> it for years, meanwhile brute force would probably help. :)

Well, my thinking was, that m68k would be re-included more likely into
testing migration/release when it just requires some core packages. Of
course the TLS problem has to be resolved until then. At least the m68k core
would then be in sync again with the rest of Debian. 

So, the way to go is for now: 
1) get some more CPU power on building packages
2) resolve the TLS issue (and other showstoppers)
3) choose some packages for a m68k core distribition and built all others on
aranym buildds
?

-- 
Ciao...                //        Fon: 0381-2744150 
      Ingo           \X/         SIP: 2744150@sipgate.de

gpg pubkey: http://www.juergensmann.de/ij_public_key.asc


Reply to: