00:18 <drbrain> ok, I understand the bug now

00:18 <drbrain> in 1.9, String#<=> returns nil unless the other object is a String or has #to_str

00:19 <drbrain> now, if the other object does not have #to_str it tries other <=> self

00:22 <drbrain> n0kada: https://bugs.ruby-lang.org/issues/7870

04:24 <_ko1> oh...

07:57 <znz_v> biff: [ruby-changes:27236] nobu:r39288 (trunk): .gdbinit: revert colored prompt - http://mla.n-z.jp/?ruby-changes=27236

07:57 <znz_v> biff: [ruby-changes:27237] nobu:r39289 (trunk): configure.in: unexpand_shvar - http://mla.n-z.jp/?ruby-changes=27237

08:23 <znz_v> biff: [ruby-changes:27238] nobu:r39290 (trunk): mkmf.rb: remove extra topdir in VPATH - http://mla.n-z.jp/?ruby-changes=27238

09:44 <znz_v> biff: [ruby-changes:27239] nobu:r39291 (trunk): * configure.in: remove debug print. - http://mla.n-z.jp/?ruby-changes=27239

10:14 <_ko1> https://bugs.ruby-lang.org/projects/ruby-trunk/issues?c[]=tracker&c[]=status&c[]=priority&c[]=subject&c[]=assigned_to&c[]=updated_on&f[]=status_id&f[]=fixed_version_id&f[]=&group_by=&op[fixed_version_id]=%3D&op[status_id]=%3D&page=1&set_filter=1&sort=assigned_to%2Cpriority%3Adesc%2Cupdated_on%3Adesc&v[fixed_version_id][]=6&v[status_id][]=1&v[status_id][]=2

10:14 <_ko1> 64 tickets

10:28 <charliesome> _ko1: better than 86!

10:32 <_ko1> :)

10:32 <mame0> x86_64

10:33 <_ko1> charliesome: I found another problem about inline cache invalidate

10:33 <charliesome> _ko1: i decided that there are too many edge cases

10:33 <charliesome> so i am working on another patch

10:33 <_ko1> wow

10:33 <charliesome> invalidating by class hierarchy

10:33 <_ko1> like jruby?

10:34 <charliesome> yep

10:34 <charliesome> so i have a linked list of subclass weakrefs in rb_classext_t*

10:34 <charliesome> as well as a pointer to the linked list node in the parent class

10:34 <charliesome> and i modify the gc to unlink a class from its parent's linked list when it is collected

10:34 <_ko1> i don't like this approach because it is too complex

10:34 <_ko1> and need some time to invalidation

10:35 <charliesome> i think it's better than invalidating by id

10:36 <_ko1> I see

10:36 <charliesome> because in most cases, any method definitions will be done at classes near the bottom of the hierarchy

10:36 <_ko1> we need to make competition

10:36 <charliesome> haha yes!

10:36 <_ko1> my finding problem is

10:36 <_ko1> now VM caches `method_missing'

10:37 <_ko1> this affects your last patch and class hierarchy approach

10:38 <charliesome> ah

10:38 <_ko1> charliesome: what is your `corner cases' ?

10:38 <charliesome> with class hierarchy approach?

10:38 <_ko1> ah, we need to discuss on ML to record our knowlege about it

10:39 <charliesome> yea

10:39 <_ko1> i'll sent it

10:39 <_ko1> about it

10:39 <charliesome> ok, cool

10:39 <charliesome> so what i think:

10:39 <charliesome> classes need to know their subclasses (weakref)

10:39 <charliesome> modules need to know their iclasses (also weakref)

10:40 <charliesome> anything else?

10:40 <_ko1> care about prepend and refinement

10:41 <charliesome> refinements might be problematic, because i don't know how they are implemented

10:41 <_ko1> me too (hehe

10:42 <charliesome> :p

10:42 <charliesome> prepend just modifies the inheritance chain, am i correct?

10:42 <charliesome> ie. method lookups with prepend still follow a linear path

10:42 <_ko1> yes

10:43 <charliesome> so in theory: prepend should "just work" with class hierarchy invalidation

10:44 <_ko1> maybe yes

11:58 <znz_v> biff: [ruby-changes:27240] nobu:r39292 (trunk): compar.c: inversed comarison without infinite recursion - http://mla.n-z.jp/?ruby-changes=27240

12:46 <znz_v> biff: [ruby-changes:27241] naruse:r39293 (trunk): Revert r39289 - http://mla.n-z.jp/?ruby-changes=27241

19:51 <drbrain> I posted the IRC log for the last implementers meeting here: http://bugs.ruby-lang.org/projects/ruby/wiki/DevelopersMeeting20130215