diff options
author | shyouhei <shyouhei@b2dd03c8-39d4-4d8f-98ff-823fe69b080e> | 2011-06-02 10:02:20 +0000 |
---|---|---|
committer | shyouhei <shyouhei@b2dd03c8-39d4-4d8f-98ff-823fe69b080e> | 2011-06-02 10:02:20 +0000 |
commit | 61ac67c14b03d52826b08ce89bb03a188dbfd6c3 (patch) | |
tree | ad63ff20dd63df63b25dcb3c55c89cff66b7e822 /ChangeLog | |
parent | 6e96241d318cf4a177c4ed59ea334cb687f4efd6 (diff) |
merge revision(s) 31731:31734:
* variable.c (rb_autoload_load): There is a race condition while
autoloading. When two or more threads touch a single autoloaded
constant at a time, one of them does the require, but others
behave oddly. To fix this situation we now refrain from
deleting the autoload table while someone is doing the autoload.
That deletion is deferred to a point where a require ended
successfully. Doing so make it possible for multiple threads to
enter autoloading at the same time but the require is protected
against multiple simultaneous entrance anyway so all but one
thread gets blocked at that point. So with it, touching a
constant that gets autoloaded cause those threads to block until
there is another one that does the same thing.
[ruby-core:36308] (#921)
* variable.c (rb_const_get_0): ditto.
* variable.c (autoload_node): ditto.
* variable.c (autoload_delete): ditto.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/branches/ruby_1_8@31732 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
Signed-off-by: URABE, Shyouhei <shyouhei@ruby-lang.org>
* variable.c (rb_const_get_0): Fix previous change. There were
possibilities when an autoload-specified library lacks
definition of the constant it was bound to. Once after such
library had already beed loaded, the autoload engine shall not
reload it. Instead the interpreter have to consider such
constant nonexistent. It results in a const_missing situation.
* variable.c (rb_autoload_load): ditto.
* variable.c (autoload_node): ditto.
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/branches/ruby_1_8@31734 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
Signed-off-by: URABE, Shyouhei <shyouhei@ruby-lang.org>
git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/branches/ruby_1_8_7@31904 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
Diffstat (limited to 'ChangeLog')
-rw-r--r-- | ChangeLog | 35 |
1 files changed, 35 insertions, 0 deletions
@@ -1,3 +1,38 @@ +Thu Jun 2 18:33:51 2011 URABE Shyouhei <shyouhei@ruby-lang.org> + + * variable.c (rb_const_get_0): Fix previous change. There were + possibilities when an autoload-specified library lacks + definition of the constant it was bound to. Once after such + library had already beed loaded, the autoload engine shall not + reload it. Instead the interpreter have to consider such + constant nonexistent. It results in a const_missing situation. + + * variable.c (rb_autoload_load): ditto. + + * variable.c (autoload_node): ditto. + +Thu Jun 2 18:28:58 2011 URABE Shyouhei <shyouhei@ruby-lang.org> + + * variable.c (rb_autoload_load): There is a race condition while + autoloading. When two or more threads touch a single autoloaded + constant at a time, one of them does the require, but others + behave oddly. To fix this situation we now refrain from + deleting the autoload table while someone is doing the autoload. + That deletion is deferred to a point where a require ended + successfully. Doing so make it possible for multiple threads to + enter autoloading at the same time but the require is protected + against multiple simultaneous entrance anyway so all but one + thread gets blocked at that point. So with it, touching a + constant that gets autoloaded cause those threads to block until + there is another one that does the same thing. + [ruby-core:36308] (#921) + + * variable.c (rb_const_get_0): ditto. + + * variable.c (autoload_node): ditto. + + * variable.c (autoload_delete): ditto. + Mon May 30 10:58:17 2011 Hiroshi Nakamura <nahi@ruby-lang.org> * lib/logger.rb (Logger::ProgName): do not depend on subversion |