Re: [jruby-dev] Experiments: AbstractRubyStruct, AtomicStruct, AbstractRubyHash

2011-12-22 Thread James Abley
Thanks for that Charlie, looks interesting. Before anyone gets into that, what about the other candidates. I know of at least 2: * Doug Lea's ConcurrentHashMapV8 * Cliff Click's ConcurrentHashMap - http://high-scale-lib.cvs.sourceforge.net/viewvc/high-scale-lib/high-scale-lib/java/util/concurrent

[jruby-dev] [jira] (JRUBY-6297) SIGSEGV running application with Java 7u2 and master

2011-12-22 Thread Uwe Kubosch (JIRA)
Uwe Kubosch created JRUBY-6297: -- Summary: SIGSEGV running application with Java 7u2 and master Key: JRUBY-6297 URL: https://jira.codehaus.org/browse/JRUBY-6297 Project: JRuby Issue Type: Bug

[jruby-dev] Failed attempt at clever WaitReadable fix

2011-12-22 Thread Charles Oliver Nutter
Ok, so I thought I was really clever when I did this to fix the repeated .extend of WaitReadable into EAGAIN exceptions... https://github.com/jruby/jruby/commit/21ea42097e Unfortunately, this appears to break either stdlib (net/protocol) or RubyGems or some combination of the two. I'm looking for

[jruby-dev] [jira] (JRUBY-6298) Error trying to build head of jruby with rvm against java 1.7

2011-12-22 Thread Karl Baum (JIRA)
Karl Baum created JRUBY-6298: Summary: Error trying to build head of jruby with rvm against java 1.7 Key: JRUBY-6298 URL: https://jira.codehaus.org/browse/JRUBY-6298 Project: JRuby Issue Type: B

Re: [jruby-dev] Bump up default Xmx?

2011-12-22 Thread Uwe Kubosch
On 2011-12-20, at 07:19, Charles Oliver Nutter wrote: > I was thinking of bumping to 1GB. > > A 32-bit JVM can't go higher than a 4GB process space, and typically > you don't go above 2GB heap there in case there's a lot of native > memory that might push it close to the 32-bit process limit. So

Re: [jruby-dev] Bump up default Xmx?

2011-12-22 Thread Rob Heittman
JVM ergonomics can default to a very small (64Mb) max heap on 32-bit desktop-class systems; at least older JVMs certainly used to. That's not enough to do very much at all. It would be nice to just yield to the JVM's ergonomics when it is doing something useful, like the allocations 1.5 and up hav

Re: [jruby-dev] Bump up default Xmx?

2011-12-22 Thread Charles Oliver Nutter
I have thought about this in the past. The problem is that there's no way to know what settings the JVM will use (or supports) without launching the JVM. I have toyed with the idea of doing a separate execution of 'java' just to get information on the JVM, either on every startup (for just base JV

[jruby-dev] [jira] (JRUBY-6299) Slow perf in String#split or #join on 1.9 mode

2011-12-22 Thread Charles Oliver Nutter (JIRA)
Charles Oliver Nutter created JRUBY-6299: Summary: Slow perf in String#split or #join on 1.9 mode Key: JRUBY-6299 URL: https://jira.codehaus.org/browse/JRUBY-6299 Project: JRuby Issue