Eric Hodel wrote: > I saw Jim made a flurry of cleanups, and I made some changes to 'gem > pristine' to make the require_gem -> gem transition to flow more > smoothly. > > Since its holiday time this will keep the pending 0.9.1 release in > the public eye a little better, and hopefully we can release 0.9.1 in > the following week. > > Are there any issues pending that I don't know about or have forgotten?
I am talking with Michael Studman about a patch to the Specification hash method. Evidently, some gems have bad values in the specification that is cause problems with the "hash" method and the "==" operator. The problem is manifest on JRuby more often than on standard Ruby just because JRuby uses a different hash algorithm that happens to compare the bad specs more often than the C version of Ruby. We have beefed up the hash and == operators in Specification, but Michael believes there are still some problems in this area. His patch changes the hash function so that the distribution of specs in the hash are different, making it less likely that the bad specs collide in the same hash bucket. But that's a solution that only "probably" fixes the issue. I'm working with him to identify the exact problem to come up with a better fix. BTW, I've asked Michael to join this list so the issue gets better visibility. -- -- Jim Weirich [EMAIL PROTECTED] http://onestepback.org -- In theory, practice and theory are the same. -- In practice, they are different. _______________________________________________ Rubygems-developers mailing list Rubygems-developers@rubyforge.org http://rubyforge.org/mailman/listinfo/rubygems-developers
