[#20190] Behavior: autoload calls rb_require() directly — Evan Phoenix <evan@...>

Hi everyone,

79 messages 2008/12/01
[#20200] Re: Behavior: autoload calls rb_require() directly — Yukihiro Matsumoto <matz@...> 2008/12/02

Hi,

[#20215] Re: Behavior: autoload calls rb_require() directly — Evan Phoenix <evan@...> 2008/12/02

[#20217] Re: Behavior: autoload calls rb_require() directly — Dave Thomas <dave@...> 2008/12/02

[#20301] Re: Behavior: autoload calls rb_require() directly — Yukihiro Matsumoto <matz@...> 2008/12/04

Hi,

[#20316] Re: Behavior: autoload calls rb_require() directly — Charles Oliver Nutter <charles.nutter@...> 2008/12/04

Yukihiro Matsumoto wrote:

[#20317] Re: Behavior: autoload calls rb_require() directly — Paul Brannan <pbrannan@...> 2008/12/04

On Fri, Dec 05, 2008 at 03:25:42AM +0900, Charles Oliver Nutter wrote:

[#20323] Leave my open classes alone (was Behavior: autoload calls rb_require() directly) — Dave Thomas <dave@...> 2008/12/04

[#20325] Re: Leave my open classes alone (was Behavior: autoload calls rb_require() directly) — Charles Oliver Nutter <charles.nutter@...> 2008/12/04

Dave Thomas wrote:

[#20328] Re: Leave my open classes alone (was Behavior: autoload calls rb_require() directly) — Yukihiro Matsumoto <matz@...> 2008/12/04

Hi,

[#20334] Re: Leave my open classes alone (was Behavior: autoload calls rb_require() directly) — Charles Oliver Nutter <charles.nutter@...> 2008/12/04

Yukihiro Matsumoto wrote:

[#20384] Re: Leave my open classes alone (was Behavior: autoload calls rb_require() directly) — Brent Roman <brent@...> 2008/12/05

[#20329] Re: Behavior: autoload calls rb_require() directly — daz <dooby@...10.karoo.co.uk> 2008/12/04

Charles Oliver Nutter wrote:

[#20335] Re: Behavior: autoload calls rb_require() directly — Charles Oliver Nutter <charles.nutter@...> 2008/12/04

daz wrote:

[#20341] Re: Behavior: autoload calls rb_require() directly — "Michael Selig" <michael.selig@...> 2008/12/04

On Fri, 05 Dec 2008 09:58:02 +1100, Charles Oliver Nutter

[#20344] Re: Behavior: autoload calls rb_require() directly — Charles Oliver Nutter <charles.nutter@...> 2008/12/05

Michael Selig wrote:

[#20214] Proposal: deferred blocks — "Yehuda Katz" <wycats@...>

Several people have played around with solutions that use ParseTree or

11 messages 2008/12/02

[#20235] autoload and concurrency — "Yehuda Katz" <wycats@...>

Merb uses autoload rather extensively. We have lately observed some

31 messages 2008/12/03
[#20236] Re: autoload and concurrency — Jim Deville <jdeville@...> 2008/12/03

This seems like a strong argument in favor of Ruby-core:20225.

[#20240] Re: autoload and concurrency — Charles Oliver Nutter <charles.nutter@...> 2008/12/03

Jim Deville wrote:

[#20242] Re: autoload and concurrency — Charles Oliver Nutter <charles.nutter@...> 2008/12/03

Charles Oliver Nutter wrote:

[#20245] Re: autoload and concurrency — "Yehuda Katz" <wycats@...> 2008/12/03

Also, this just illustrates that it's possible. In the case of Merb, we

[#20247] Re: autoload and concurrency — Tomas Matousek <Tomas.Matousek@...> 2008/12/03

I think it has already been concluded that autoload and require are inherently broken in presence of threads. See ruby-core:19860.

[#20251] Re: autoload and concurrency — Charles Oliver Nutter <charles.nutter@...> 2008/12/03

Require could be made safe if only one thread were allowed to execute

[#20241] [Bug #814] NoMethodError: undefined method `read_nonblock' for #<OpenSSL::SSL::SSLSocket:0x1a64f9a0> — Aaron Patterson <redmine@...>

Bug #814: NoMethodError: undefined method `read_nonblock' for #<OpenSSL::SSL::SSLSocket:0x1a64f9a0>

19 messages 2008/12/03
[#22538] [Bug #814] NoMethodError: undefined method `read_nonblock' for #<OpenSSL::SSL::SSLSocket:0x1a64f9a0> — Tony Arcieri <redmine@...> 2009/02/26

Issue #814 has been updated by Tony Arcieri.

[#20416] [Feature #839] Add code on each line of a backtrace output to the screen — Roger Pack <redmine@...>

Feature #839: Add code on each line of a backtrace output to the screen

12 messages 2008/12/08

[#20483] encoding of symbols — Dave Thomas <dave@...>

If I have a source file like this:

50 messages 2008/12/11
[#20494] Re: encoding of symbols — Yukihiro Matsumoto <matz@...> 2008/12/11

Hi,

[#20496] Re: encoding of symbols — Yukihiro Matsumoto <matz@...> 2008/12/12

Hi,

[#20522] Re: encoding of symbols — Charles Oliver Nutter <charles.nutter@...> 2008/12/13

Yukihiro Matsumoto wrote:

[#20526] Re: encoding of symbols — Brian Candler <B.Candler@...> 2008/12/13

On Sat, Dec 13, 2008 at 08:33:13PM +0900, Charles Oliver Nutter wrote:

[#20540] Re: 1.9 character encoding (was: encoding of symbols) — "Michael Selig" <michael.selig@...> 2008/12/14

On Sun, 14 Dec 2008 01:01:44 +1100, Brian Candler <B.Candler@pobox.com>

[#20545] Re: 1.9 character encoding (was: encoding of symbols) — "Michael Selig" <michael.selig@...> 2008/12/14

On Sun, 14 Dec 2008 11:57:55 +1100, Michael Selig

[#20562] Re: 1.9 character encoding (was: encoding of symbols) — Yukihiro Matsumoto <matz@...> 2008/12/15

Hi,

[#20619] Re: 1.9 character encoding (was: encoding of symbols) — danielcavanagh@... 2008/12/17

> You're right. When we have two strings with identical byte sequence

[#20502] [Bug #863] Openssl issues with fresh compile on Ubuntu — Brian Takita <redmine@...>

Bug #863: Openssl issues with fresh compile on Ubuntu

11 messages 2008/12/12

[#20557] [Bug #877] [win32] Ruby Standard Library (maybe smth else): Wrong Encoding in Files, Directories and Environment Variables — "Dmitry A. Ustalov" <redmine@...>

Bug #877: [win32] Ruby Standard Library (maybe smth else): Wrong Encoding in Files, Directories and Environment Variables

14 messages 2008/12/14

[#20564] [Bug #883] Failure: test_handle_special_CROSSREF_no_underscore(TestRDocMarkupToHtmlCrossref) — Kazuhiro NISHIYAMA <redmine@...>

Bug #883: Failure: test_handle_special_CROSSREF_no_underscore(TestRDocMarkupToHtmlCrossref)

9 messages 2008/12/15

[#20576] [Bug #888] zlib 1.2.3 does not work with Rubygems 1.3.1 (in Ruby 1.9.1) on Windows — Chauk-Mean PROUM <redmine@...>

Bug #888: zlib 1.2.3 does not work with Rubygems 1.3.1 (in Ruby 1.9.1) on Windows

14 messages 2008/12/15

[#20578] [Feature #889] erb.rb should use Array and << for eoutvar and not String and concat — Thomas Enebo <redmine@...>

Feature #889: erb.rb should use Array and << for eoutvar and not String and concat

15 messages 2008/12/15

[#20668] [Feature #905] Add String.new(fixnum) to preallocate large buffer — Charles Nutter <redmine@...>

Feature #905: Add String.new(fixnum) to preallocate large buffer

60 messages 2008/12/18
[#20671] Re: [Feature #905] Add String.new(fixnum) to preallocate large buffer — Yukihiro Matsumoto <matz@...> 2008/12/19

Hi

[#20674] Re: [Feature #905] Add String.new(fixnum) to preallocate large buffer — Charles Oliver Nutter <charles.nutter@...> 2008/12/19

Yukihiro Matsumoto wrote:

[#20697] Re: [Feature #905] Add String.new(fixnum) to preallocate large buffer — Jim Weirich <jim.weirich@...> 2008/12/19

[#20703] Re: [Feature #905] Add String.new(fixnum) to preallocate large buffer — Charles Oliver Nutter <charles.nutter@...> 2008/12/19

Jim Weirich wrote:

[#20704] Re: [Feature #905] Add String.new(fixnum) to preallocate large buffer — Dave Thomas <dave@...> 2008/12/19

[#28461] [Feature #905] Add String.new(fixnum) to preallocate large buffer — caleb clausen <redmine@...> 2010/03/04

Issue #905 has been updated by caleb clausen.

[#28491] [Feature #905] Add String.new(fixnum) to preallocate large buffer — Kurt Stephens <redmine@...> 2010/03/05

Issue #905 has been updated by Kurt Stephens.

[#20695] [Bug #907] Various system() and backticks problems on Windows — "James M. Lawrence" <redmine@...>

Bug #907: Various system() and backticks problems on Windows

13 messages 2008/12/19

[#20706] [Feature #908] Should be an easy way of reading N characters from am I/O stream — Michael Selig <redmine@...>

Feature #908: Should be an easy way of reading N characters from am I/O stream

39 messages 2008/12/19
[#21816] [Feature #908] Should be an easy way of reading N characters from am I/O stream — Michael Selig <redmine@...> 2009/02/03

Issue #908 has been updated by Michael Selig.

[#21825] Re: [Feature #908] Should be an easy way of reading N characters from am I/O stream — Tanaka Akira <akr@...> 2009/02/04

In article <4988d2fa997f8_8527a9e32018e7@redmine.ruby-lang.org>,

[#21826] Re: [Feature #908] Should be an easy way of reading N characters from am I/O stream — "Michael Selig" <michael.selig@...> 2009/02/04

Hi,

[#22100] Re: [Feature #908] Should be an easy way of reading N characters from am I/O stream — Tanaka Akira <akr@...> 2009/02/14

In article <op.uotab6oa9245dp@kool>,

[#22125] Re: [Feature #908] Should be an easy way of reading N characters from am I/O stream — Michal Suchanek <hramrach@...> 2009/02/15

2009/2/14 Tanaka Akira <akr@fsij.org>:

[#22146] Re: [Feature #908] Should be an easy way of reading N characters from am I/O stream — Tanaka Akira <akr@...> 2009/02/15

In article <a5d587fb0902141711q780f0d24jef9be9b8bbe69b2a@mail.gmail.com>,

[#22182] Re: [Feature #908] Should be an easy way of reading N characters from am I/O stream — Michal Suchanek <hramrach@...> 2009/02/16

2009/2/15 Tanaka Akira <akr@fsij.org>:

[#22213] Re: [Feature #908] Should be an easy way of reading N characters from am I/O stream — Tanaka Akira <akr@...> 2009/02/18

In article <a5d587fb0902160252u56b50cfdv8e0fd36bb4f0b1b3@mail.gmail.com>,

[#22215] Re: [Feature #908] Should be an easy way of reading N characters from am I/O stream — Michal Suchanek <hramrach@...> 2009/02/18

2009/2/18 Tanaka Akira <akr@fsij.org>:

[#22238] Re: [Feature #908] Should be an easy way of reading N characters from am I/O stream — "Michael Selig" <michael.selig@...> 2009/02/18

On Thu, 19 Feb 2009 02:21:21 +1100, Michal Suchanek <hramrach@centrum.cz>

[#22253] Re: [Feature #908] Should be an easy way of reading N characters from am I/O stream — Tanaka Akira <akr@...> 2009/02/19

In article <op.upklh9q19245dp@kool>,

[#22281] Re: [Feature #908] Should be an easy way of reading N characters from am I/O stream — Martin Duerst <duerst@...> 2009/02/20

At 19:00 09/02/19, Tanaka Akira wrote:

[#22332] Re: [Feature #908] Should be an easy way of reading N characters from am I/O stream — Tanaka Akira <akr@...> 2009/02/22

In article <6.0.0.20.2.20090220134502.0823ee98@localhost>,

[#22338] Re: [Feature #908] Should be an easy way of reading N characters from am I/O stream — "Michael Selig" <michael.selig@...> 2009/02/22

On Mon, 23 Feb 2009 03:00:41 +1100, Tanaka Akira <akr@fsij.org> wrote:

[#22356] Re: [Feature #908] Should be an easy way of reading N characters from am I/O stream — Michal Suchanek <hramrach@...> 2009/02/23

2009/2/22 Michael Selig <michael.selig@fs.com.au>:

[#20723] [Bug #911] ArgumentError in Resolv#getaddress — Federico Builes <redmine@...>

Bug #911: ArgumentError in Resolv#getaddress

14 messages 2008/12/20

[#20797] [Bug #921] autoload is not thread-safe — Charles Nutter <redmine@...>

Bug #921: autoload is not thread-safe

15 messages 2008/12/22

[#20893] [Bug #932] incorrect case statement in "ext/dl/test/test_base.rb" causes library problems on openSUSE 11.1 64-bit — Ed Borasky <redmine@...>

Bug #932: incorrect case statement in "ext/dl/test/test_base.rb" causes library problems on openSUSE 11.1 64-bit

8 messages 2008/12/26

[#20978] Definable != is a Bad Thing™ — Ryan Davis <ryand-ruby@...>

> >> class X; def == o; :great; end; def != o; :horrible; end; end

20 messages 2008/12/30
[#20979] Re: Definable != is a Bad Thing™ — Yukihiro Matsumoto <matz@...> 2008/12/30

Hi,

[#20994] [Bug #956] Encoding: nl_langinfo(CODESET) on cygwin 1.5 always returns US-ASCII — Tom Link <redmine@...>

Bug #956: Encoding: nl_langinfo(CODESET) on cygwin 1.5 always returns US-ASCII

10 messages 2008/12/30

[#20999] Supporting Thread.critical=with native threads — Shri Borde <Shri.Borde@...>

Hi,

27 messages 2008/12/30
[#21000] Re: Supporting Thread.critical=with native threads — Charles Oliver Nutter <charles.nutter@...> 2008/12/30

Shri Borde wrote:

[#21002] Re: Supporting Thread.critical=with native threads — Eric Hodel <drbrain@...7.net> 2008/12/31

On Dec 30, 2008, at 15:00 PM, Shri Borde wrote:

[#21010] Re: Supporting Thread.critical=with native threads — Shri Borde <Shri.Borde@...> 2008/12/31

http://www.ruby-doc.org/core/classes/Thread.html#M000461 says:

[#21245] Re: Supporting Thread.critical=with native threads — Charles Oliver Nutter <charles.nutter@...> 2009/01/10

I'm starting come around to Shri's idea of critical= being represented

[#21353] Re: Supporting Thread.critical=with native threads — Shri Borde <Shri.Borde@...> 2009/01/14

Is opening a bug the recommended way to get the spec changed for 1.9*?

[#21359] Re: Supporting Thread.critical=with native threads — Brent Roman <brent@...> 2009/01/15

[ruby-core:20797] [Bug #921] autoload is not thread-safe

From: Charles Nutter <redmine@...>
Date: 2008-12-22 17:10:34 UTC
List: ruby-core #20797
Bug #921: autoload is not thread-safe
http://redmine.ruby-lang.org/issues/show/921

Author: Charles Nutter
Status: Open, Priority: Normal

Currently autoload is not safe to use in a multi-threaded application. To put it more bluntly, it's broken.

The current logic for autoload is as follows:

1. A special object is inserted into the target constant table, used as a marker for autoloading
2. When that constant is looked up, the marker is found and triggers autoloading
3. The marker is first removed, so the constant now appears to be undefined if retrieved concurrently
4. The associated autoload resource is required, and presumably redefines the constant in question
5. The constant lookup, upon completion of autoload, looks up the constant again and either returns its new value or proceeds with normal constant resolution

The problem arises when two or more threads try to access the constant. Because autoload is stateful and unsynchronized, the second thread may encounter the constant table in any number of states:

1. It may see the autoload has not yet fired, if the first thread has encountered the marker but not yet removed it. It would then proceed along the same autoload path, requiring the same file a second time.
2. It may not find an autoload marker, and assume the constant does not exist.
3. It may see the eventual constant the autoload was intended to define.

Of these combinations, (3) is obviously the desired behavior. (1) can only happen on native-threaded implementations that do not have a global interpreter lock, since it requires concurrency during autoload's internal logic. (2) can happen on any implementation, since while the required file is processing the original autoload constant appears to be undefined.

I have only come up with two solutions:

* When the autoload marker is encountered, it is replaced (under lock) with an "autoload in progress" marker. All subsequent threads will then see this marker and wait for the autoloading process to complete. the mechanics of this are a little tricky, but it would guarantee concurrent autoloads would only load the target file once and would always return the intended value to concurrent readers.
* A single autoload mutex, forcing all autoloads to happen in serial.

There is a potential for deadlock in the first solution, unfortunately, since two threads autoloading two constants with circular autoloaded constant dependencies would ultimately deadlock, each waiting for the other to complete. Because of this, a single autoload mutex for all autoloads may be the only safe solution.


----------------------------------------
http://redmine.ruby-lang.org

In This Thread

Prev Next