Hello,

In message "[ruby-core:17530] Re: We'll release 1.8.6/1.8.7 this Friday"
    on Jul.03,2008 16:16:34, <luislavena / gmail.com> wrote:
> > Can you test with reverting r17290 (socket changes)?

Oh, sorry, I'd mistaken.
These are select changes, not socket changes.


> Reverting the specified revision fix the issue:
> 
> 1588 tests, 15068 assertions, 3 failures, 0 errors
> http://pastie.org/private/xk8thrwcbzeb2h1s7lm3tw
> 
> Those are the same errors that are failing under p114, which is nice.

Good news :)


> One comment about r17290 is that it previously worked, since before
> 1.8.7 was released I've been able to use ruby_1_8 branch, which
> included the same backport.

Hmm...

I consider four senarios.

  1) drb changes cause this problem.

  2) ext/socket changes cause this problem.

  3) win32/win32.c changes cause this problem.

  4) another changes cause this problem.

First, drb was not changed between p114 and p230. So, drb is
not guilty.
Second, ext/socket was changed, but it doesn't seems to affect.
Third, win32/win32.c contains the changes about select. These
are very suspicious.
Fourth, ... I have no idea :(

So, I suggested to revert r17290.


On ruby_1_8 branch, ext/socket changes are older than win32's
select changes.
So, probably ext/socket changes doesn't cause it, too.

BTW, r17290 is tagged as p207.
So, if p207 has same problem, ruby_1_8_6 needs something to
backport from ruby_1_8.
If p207 doesn't has the problem, something between p208 and
p230 causes it.
Luis, could you test p207, please?


Regards,
-- 
U.Nakamura <usa / garbagecollect.jp>