Forum: Ruby-core Current status of 1.9.1 RC2

1e7e007723cfe59320d994a7db4ab4cf?d=identicon&s=25 Yugui (Yuki Sonoda) (Guest)
on 2009-01-13 14:08
(Received via mailing list)
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Hi folks,

I have a plan to release Ruby 1.9.1 RC2 on Jan 15 and we must release
1.9.1 on Jan 25. But there are 23 issues.

== #616 instance_eval and Module#to_s
< ko1
Scope for contants in instance_eval or module_eval is changed in 1.9.
The remaining problem is why the class is not named.

== #847, #848 debug.rb
This is a result of a bit of feature change in set_trace_func. This is
difficult to fix.
Will anyone fix debug.rb?  We need help.

== #865 patch for Rake
< Jim Weirich
Is the patch correct?  Then, please apply it.

== #866 manpages for RubyGems
< Jim Weirich
*I asked you about the license for the manpages 4 times and sent the
manpages to you.*

Can I distribute the pages with Ruby under GPL and Ruby's license?  Or
is the license unclear?  Then I'll give up the idea of distributing them
.


== #886 GC from blocking region
< ko1
Is this fixed in r21438 and r21441?

== #888 zlib 1.2.3 does not work with Rubygems 1.3.1 (in Ruby 1.9.1) on
Windows
Is this still reproducible?

== #935 child process forked from child thread does not call
ruby_cleanup
< mame
Is this still reproducible?  Or close the ticket.
I cannot reproduce.

== #940 ruby/config.h:1:1: warning: "PACKAGE_NAME" redefined
Can I regard this as rejected?

== #942 Build fails on IA-32 Linux with gcc 4.3.2
Is this still reproducible?

== #947 Signal handler block is never called when the program is too
short
< ko1
Show me your plan to fix this issue. Or should I give up to fix this by
1.9.1?

== #948 dl: cannot pass double value correctly on all x86_64 systems
I want to release 1.9.1 with this bug because dl2 x86_64 support seems
to be too difficult to fix.  Any idea?

== #956 Encoding: nl_langinfo(CODESET) on cygwin 1.5 always returns
US-ASCII
< nobu
Can I regard this as fixed?  Then close the ticket.

== #957 Cannot build in OSX 10.5.6 using gcc4.0.1 or gcc4.2 due to issue
with readline.h
< Kouji Takao
I cannot reproduce this issue on OSX 10.5.6 with gcc 4.0.1.
Can you reproduce this?

== #967 Results of const_missing are being cached incorrectly
< ko1
Show me your plan to fix this issue.

== #968 trap may fail to handle signals
< ko1
What do you think about mame's patch?

== #975 ruby curses extension does not support multibyte characters
< matz
Do you intend to fix this by 1.9.1?

== #981 String#encode() keeps invalid bytes when converting to US-ASCII
Any idea?

== #984 Test::Unit::Assertions#assert_match double counts number of
assertions
I'm working on this issue.

== #985 profile lib + iterator => 'NULL pointer given' (ArgumentError)
< ko1
Show me your plan to fix this issue.


== #1002 r21312 causes test failures on RubyGems
< Eric Hodel, nobu

I think r21312 broke an assumption for Rubygems' working correctly.
What do you think?

== #1003 autoload test failed on OpenSolaris
< Dae San, nobu
You are working on this issue? right?

== #1004 Encoding related test failure on OpenSolaris
< nobu
Can I regard this issue as rejected?



Regards,

- -- Yugui <yugui@yugui.jp>
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.9 (Darwin)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iEYEARECAAYFAklskicACgkQOXzH5JLb/AWegwCfRtIZ5JDskIn543mv52ps9qFu
1oEAn2zzJtWkRBM3UESTcCWCALs2eSiV
=MuTv
-----END PGP SIGNATURE-----
D414949a83ec0fb151c7f8dd03249871?d=identicon&s=25 Michael Selig (Guest)
on 2009-01-13 23:31
(Received via mailing list)
On Wed, 14 Jan 2009 00:07:04 +1100, Yugui (Yuki Sonoda) <yugui@yugui.jp>
wrote:


> == #942 Build fails on IA-32 Linux with gcc 4.3.2
> Is this still reproducible?

This problem appears to have been fixed.

Mike
F52e87b92cafb1e8c6d155076b56ecff?d=identicon&s=25 Martin Duerst (Guest)
on 2009-01-14 02:21
(Received via mailing list)
At 22:07 09/01/13, Yugui (Yuki Sonoda) wrote:
>-----BEGIN PGP SIGNED MESSAGE-----
>Hash: SHA1
>
>Hi folks,
>
>I have a plan to release Ruby 1.9.1 RC2 on Jan 15 and we must release
>1.9.1 on Jan 25. But there are 23 issues.

>== #956 Encoding: nl_langinfo(CODESET) on cygwin 1.5 always returns US-ASCII
>< nobu
>Can I regard this as fixed?  Then close the ticket.

I think it's fixed, but I want to do one or two more tests this
afternoon,
then I'll close it.


>== #981 String#encode() keeps invalid bytes when converting to US-ASCII
>Any idea?

I suggest Akira and/or Yui have a look at this.


>== #1004 Encoding related test failure on OpenSolaris
>< nobu
>Can I regard this issue as rejected?

This is about TK, not about Encoding per se.


Regards,    Martin.



#-#-#  Martin J. Du"rst, Assoc. Professor, Aoyama Gakuin University
#-#-#  http://www.sw.it.aoyama.ac.jp       mailto:duerst@it.aoyama.ac.jp
1018fce89dab3bedcc380fe37d90c9c0?d=identicon&s=25 David Flanagan (Guest)
on 2009-01-14 19:57
(Received via mailing list)
Yugui (Yuki Sonoda) wrote:

> == #984 Test::Unit::Assertions#assert_match double counts number of
> assertions
> I'm working on this issue.
>

Yugui,

This is the only one of the 23 bugs that wasn't assigned in redmine that
I felt qualified to look at.  You said you were working on it, but I
took a look anyway.  My proposed patch is attached.  It should go
through Ryan, of course, since it is to lib/minitest/unit.rb

[Justification: I claim that subtracting one in assert_match,
assert_empty and assert_include (and the corresponding refute) methods
is the right thing  to do.  If it is intentional that these implicit
assert_respond_to assertions are being counted, then I think that
assert_instance_of and assert_kind_of should have similar behavior.]

  David Flanagan
308cbef6e86dfc49cce3b2d4cf42aedc?d=identicon&s=25 SASADA Koichi (Guest)
on 2009-01-15 01:45
(Received via mailing list)
Sorry for late response.

Yugui (Yuki Sonoda) wrote::
> == #616 instance_eval and Module#to_s
> < ko1
> Scope for contants in instance_eval or module_eval is changed in 1.9.
> The remaining problem is why the class is not named.

I think remaining issue is not critical.

> == #847, #848 debug.rb
> This is a result of a bit of feature change in set_trace_func. This is
> difficult to fix.
> Will anyone fix debug.rb?  We need help.

I will check Will Sobel's patch this nihgt.

> == #886 GC from blocking region
> < ko1
> Is this fixed in r21438 and r21441?

I hope so.

> == #947 Signal handler block is never called when the program is too short
> < ko1
> Show me your plan to fix this issue. Or should I give up to fix this by
> 1.9.1?

I think giving up is suitable if possible (I think it is not critical).

> == #967 Results of const_missing are being cached incorrectly
> < ko1
> Show me your plan to fix this issue.

I missed this issue.
I will count-up ruby_vm_global_state_version at const_missing as ad-hoc
(or only one?) solution.

I'll fix it this night.

> == #968 trap may fail to handle signals
> < ko1
> What do you think about mame's patch?

I'll check mame's patch this night.

> == #985 profile lib + iterator => 'NULL pointer given' (ArgumentError)
> < ko1
> Show me your plan to fix this issue.

I'll check matz' patch this night.
F24ff61beb80aa5f13371aa22a35619c?d=identicon&s=25 Yusuke ENDOH (Guest)
on 2009-01-15 14:08
(Received via mailing list)
Hi,

2009/1/13 Yugui (Yuki Sonoda) <yugui@yugui.jp>:
> == #935 child process forked from child thread does not call ruby_cleanup
> < mame
> Is this still reproducible?  Or close the ticket.

Yes, it is.


> I cannot reproduce.

Really?  I expected the issue will occur in any environment.


$ cat t.rb
at_exit { p [:exit, $$] }
Thread.new { p [fork ? :parent : :child, $$] }
sleep 1

$ ruby18 -v t.rb
ruby 1.8.7 (2009-01-14 revision 21505) [i686-linux]
[:child, 30417]
[:exit, 30417]
[:parent, 30416]
[:exit, 30416]

$ ruby19 -v t.rb
ruby 1.9.1p5000 (2009-01-15 trunk 21529) [i686-linux]
[:child, 30421]
[:parent, 30418]
[:exit, 30418]
Ede2aa10c6462f1d825143879be59e38?d=identicon&s=25 Charles Oliver Nutter (Guest)
on 2009-01-16 22:51
(Received via mailing list)
SASADA Koichi wrote:
>> == #967 Results of const_missing are being cached incorrectly
>> < ko1
>> Show me your plan to fix this issue.
>
> I missed this issue.
> I will count-up ruby_vm_global_state_version at const_missing as ad-hoc
> (or only one?) solution.
>
> I'll fix it this night.

This should be fixable by simply not caching const_missing results. I
don't think you need to have const_missing increment the global state
version. JRuby also uses a global version number for constants and this
strategy works fine.

- Charlie
5a837592409354297424994e8d62f722?d=identicon&s=25 Ryan Davis (Guest)
on 2009-01-17 21:11
(Received via mailing list)
On Jan 14, 2009, at 10:56 , David Flanagan wrote:

> [Justification: I claim that subtracting one in assert_match,
> assert_empty and assert_include (and the corresponding refute)
> methods is the right thing  to do.  If it is intentional that these
> implicit assert_respond_to assertions are being counted, then I
> think that assert_instance_of and assert_kind_of should have similar
> behavior.]

this behavior is intentional. assert_match and friends are compound
assertions. I don't see your point about assert_instance_of or
assert_kind_of. They only have one assertion each.
This topic is locked and can not be replied to.