Attempt to create decimal 2.2250738585072012e-308 sends JVM into tight infinite loop

From the articles about the issue, it seems this has been accepted as a bug with internal bug number 1949967.

I have a working solution I'm going to commit. There may be improvements suggested by commenters on my blog post:

http://blog.headius.com/2011/02/working-around-java-doubleparsedouble.html

commit 4c712963885c0117b95066d927520a6a738c2a65
Author: Charles Oliver Nutter <headius@headius.com>
Date: Wed Feb 2 12:20:05 2011 -0600

Fix JRUBY-5441: Attempt to create decimal 2.2250738585072012e-308 sends JVM into tight infinite loop

There were only two places we called Double.parseDouble. I modified both places to use my own logic that splits up the mantissa and exponent and uses BigDecimal math to produce a result.

Will this be backported to 1.5.6?

This isn't fixed yet.

Punting off RC2.

I tried another attempt at this, by checking whether the normalized string form (produced by BigDecimal.toString just before calling .doubleValue) is .equals to the known bad form. My change just then replaced the doubleValue call with a literal value close to the bad form that's known to parse, and other things failed.

We'll have to poke around more for a workaround. Nothing so far passes all tests.

I made another attempt that tries to check if the value is in the proper range, but this version fails other specs that expect specific minimum float values.

https://gist.github.com/815270

Oracle has released a hotfix that applies to earlier JDKs as well. http://www.oracle.com/technetwork/java/javase/fpupdater-tool-readme-305936.html

I expect Apple will get updates released shortly.

Indeed, Apple's latest JDK contains the fix.

$ jruby -v -e 'p 2.2250738585072012e-308'
jruby 1.6.0 (ruby-1.8.7-p330) (2011-04-11 b695384) (Java HotSpot(TM) 64-Bit Server VM 1.6.0_24) [darwin-x86_64-java]
2.2250738585072e-308

We do still have a pragmatic problem of supporting older versions of JDK.

Should we incorporate the fix outlined here: http://extremestjython.blogspot.com/2011/02/no-more-java-double-trouble.html

Yes, I think we should.

Fixed on master (66b5b33) and jruby-1_6:

commit 9791492574e39c78e9bc62d9dc71081e2810e2dc
Author: Charles Oliver Nutter <headius@headius.com>
Date: Mon Jul 11 19:40:43 2011 -0500

Fix JRUBY-5541 on older JDKs: Attempt to create decimal 2.2250738585072012e-308 sends JVM into tight infinite loop

• Using code from Oti's blog: http://extremestjython.blogspot.com/2011/02/no-more-java-double-trouble.html, which he has released into the public domain.