Nightly regression tests on PPC OSX

Initial triage for 3.0.1 release; suggesting postponing to 3.0.2; move back to 3.0.1 if you disagree.

The machine is there (paprika.anu.edu.au). I have upgraded the OS and dealt with the problem of it going into a power saving sleep. Daniel has just committed Tony's buildit fixes.

I think all that remains is for someone who knows what they're doing to set up the crontab suitably.

I suggest we leave this as a 3.0.1 until someone actually tries. If it turns out to be difficult, by all means defer to 3.0.2. But if it is trivial it would be silly to defer any longer. It's a nice piece of h/w sitting idle and an OS going untested.

So apparently the problem was that the developer tools were not installed. I have installed them now.

I have set up regressions to run and they seem to have kicked off ok.

If they report back ok overnight we can close this.

Thanks for setting up paprika. This tracker is for PPC OS/X not Intel OS/X. Is there a PPC OS/X machine we can also run a daily regression test on?

Sorry I didn't realize.

No, we don't have a PPC OSX machine. It was stolen (unbelievably) a year or so ago.

Since a) Apple has not been shipping PPC for some years now, b) we have decent PPC AIX & Linux coverange, c) we now have OS IA32, and d) I suspect we're never likely to get another PPC OSX machine, I suggest we close this tracker.

My desktop machine is a PPC OS/X mac mini, I can do regression tests on it in the night UK time. I'll copy the keys over and set this up if it's ok?

Sure.

If you'd like it to function like the other machines (ie make it available to the core team), the thing to do is create an account called "regression" and copy the keys over from one of the other regression machines. You could optionally also create an account called test and use the same keys. However, since the machine is also your desktop, I can understand that you may not want to do this.

You'll also need to make the various benchmarks available on the machine.

Importance of PPC OSX continuing to decline... at this point I think a good case can be made for just removing support for the platform entirely. Would cleanup some of the PPC calling conventions to only have 2 ABIs to support instead of 3.

sigh. wrong fix target.
reopneing to fix.