Re: [Jikesrvm-core] OS X and 2.3.5 release

[Chris H. <hof...@cs...>]

Sorry for the delay. As Eliot said, I've been tied up with some personal 
business for the past few weeks. On top of that most of my funding now 
comes from sources other than Eliot, so I don't have nearly the time 
available that I'd like to work on JikesRVM.

We suspect the crash reporter is the main issue, but aren't certain. The 
OSX machines used to hang every month or so and we weren't sure what was 
the cause. We changed our file servers and suddenly they locked up 
frequently (right now they lock up on almost every run of the regression 
suite) and examination of the logs makes it pretty clear that this is 
happening as the crash log is being updated.

The person who pointed out the setting for turning off crashreporter 
also said he thought doing this globally might break some other tools we 
use. It's worth testing, though.

There is a more serious problem at the moment, though, in that JikesRVM 
won't build at all. This is because the Jikes build scripts are passing 
arguments that are specific to GNU's libtool to g++, but on OSX g++ is 
using a different libtool. GNU's libtool is called glibtool.

This seems to be related to some sort of upgrade to OSX or its 
development tools rather than to changes in JikesRVM. One of the final 
things I was able to confirm before losing all my "spare" time was that 
the last couple of releases of JikesRVM no longer build on our machines, 
whereas they most certainly did a few weeks ago -- though I had to 
disable the automatic regression, I did occasionally run an OSX 
regression by hand when I knew the machines were not otherwise in use or 
when a bug had been reported.

I hope this would be an easy thing to fix for someone who is a true OSX 
programmer, instead of a Linux programmer who has accumulated enough 
bookmarks in his web browser to pretend to be an OSX programmer (e.g: me).

I'm sorry I've been unable to give this the attention it deserves. If 
Brian is willing to become the official host of the regression test, 
this might be a good idea. I can introduce him to the set of scripts I 
use to kick off the standard JikesRVM regression suite.

Chris

Brian D. Carlstrom wrote:
> David P Grove writes:
>  > A longer term issue is that if we cannot find an OS X box on which nightly 
>  > regression tests can be reliably run, then I think we may be forced to 
>  > drop OS X as a supported platform.  
> 
> Is the problem with reliably running nightly regression tests a problem
> with the CrashReporter? I thought it could simply be turned off:
> 
>     http://sourceforge.net/mailarchive/message.php?msg_id=10791152
> 
> If it is access to a machine for regression I could probably run the
> regressions on one of our G5 boxes.
> 
> I'm asking because Mac OS X is still important to us even with our move
> of some of our boxes to linux/ppc. My advisor at Stanford knows someone
> to talk to at Apple but I want to be informed before trying to get then
> to look at the issue.
> 
> -bri
> 
> 
> -------------------------------------------------------
> This SF.net email is sponsored by Demarc:
> A global provider of Threat Management Solutions.
> Download our HomeAdmin security software for free today!
> http://www.demarc.com/info/Sentarus/hamr30
> _______________________________________________
> Jikesrvm-core mailing list
> Jik...@li...
> https://lists.sourceforge.net/lists/listinfo/jikesrvm-core
> 

-- 
Chris Hoffmann -- Dept. of Computer Science/UMass at Amherst
http://www-ali.cs.umass.edu/~hoffmann