[#111565] [Ruby master Bug#19293] The new Time.new(String) API is nice... but we still need a stricter version of this — "matsuda (Akira Matsuda) via ruby-core" <ruby-core@...>

Issue #19293 has been reported by matsuda (Akira Matsuda).

8 messages 2023/01/01

[#111572] [Ruby master Bug#19297] Don't download content from internet to execute Ruby test suite — "vo.x (Vit Ondruch) via ruby-core" <ruby-core@...>

Issue #19297 has been reported by vo.x (Vit Ondruch).

12 messages 2023/01/02

[#111579] [Ruby master Feature#19300] Move public objects from Kernel to Object — "zverok (Victor Shepelev) via ruby-core" <ruby-core@...>

Issue #19300 has been reported by zverok (Victor Shepelev).

15 messages 2023/01/02

[#111581] [Ruby master Bug#19301] Fix Data class to report keyrest instead of rest parameters — "bkuhlmann (Brooke Kuhlmann) via ruby-core" <ruby-core@...>

SXNzdWUgIzE5MzAxIGhhcyBiZWVuIHJlcG9ydGVkIGJ5IGJrdWhsbWFubiAoQnJvb2tlIEt1aGxt

8 messages 2023/01/02

[#111604] [Ruby master Misc#19304] Kernel vs Object documentation — "zverok (Victor Shepelev) via ruby-core" <ruby-core@...>

Issue #19304 has been reported by zverok (Victor Shepelev).

8 messages 2023/01/03

[#111674] [Ruby master Feature#19314] String#bytesplice should support partial copy — "shugo (Shugo Maeda) via ruby-core" <ruby-core@...>

Issue #19314 has been reported by shugo (Shugo Maeda).

8 messages 2023/01/06

[#111678] [Ruby master Feature#19315] Lazy substrings in CRuby — "Eregon (Benoit Daloze) via ruby-core" <ruby-core@...>

Issue #19315 has been reported by Eregon (Benoit Daloze).

11 messages 2023/01/06

[#111693] [Ruby master Bug#19316] YJIT crash in 3.2.0 — "jdashton (J Daniel Ashton) via ruby-core" <ruby-core@...>

Issue #19316 has been reported by jdashton (J Daniel Ashton).

12 messages 2023/01/06

[#111696] [Ruby master Feature#19317] Unicode ICU Full case mapping — "noraj (Alexandre ZANNI) via ruby-core" <ruby-core@...>

SXNzdWUgIzE5MzE3IGhhcyBiZWVuIHJlcG9ydGVkIGJ5IG5vcmFqIChBbGV4YW5kcmUgWkFOTkkp

7 messages 2023/01/06

[#111712] [Ruby master Feature#19322] Support spawning "private" child processes — "kjtsanaktsidis (KJ Tsanaktsidis) via ruby-core" <ruby-core@...>

SXNzdWUgIzE5MzIyIGhhcyBiZWVuIHJlcG9ydGVkIGJ5IGtqdHNhbmFrdHNpZGlzIChLSiBUc2Fu

14 messages 2023/01/07

[#111739] [Ruby master Feature#19324] Enumerator.product => Enumerable#product — "zverok (Victor Shepelev) via ruby-core" <ruby-core@...>

Issue #19324 has been reported by zverok (Victor Shepelev).

18 messages 2023/01/08

[#111740] [Ruby master Bug#19325] Windows support lacking. — "dsisnero (Dominic Sisneros) via ruby-core" <ruby-core@...>

Issue #19325 has been reported by dsisnero (Dominic Sisneros).

11 messages 2023/01/08

[#111742] [Ruby master Feature#19326] Please add a better API for passing a Proc to a Ractor — sdwolfz via ruby-core <ruby-core@...>

SXNzdWUgIzE5MzI2IGhhcyBiZWVuIHJlcG9ydGVkIGJ5IHNkd29sZnogKENvZHJ1yJsgR3XImW9p

13 messages 2023/01/08

[#111789] [Ruby master Feature#19333] Setting (Fiber Local|Thread Local|Fiber Storage) to nil should delete value in order to avoid memory leaks. — "ioquatix (Samuel Williams) via ruby-core" <ruby-core@...>

Issue #19333 has been reported by ioquatix (Samuel Williams).

10 messages 2023/01/11

[#111792] [Ruby master Bug#19334] Defining many instance variables and accessing them is slow in Ruby 3.2.0 — "mame (Yusuke Endoh) via ruby-core" <ruby-core@...>

Issue #19334 has been reported by mame (Yusuke Endoh).

12 messages 2023/01/12

[#111812] [Ruby master Bug#19340] Ruby master 'make install' not installing rbs gem — "MSP-Greg (Greg L) via ruby-core" <ruby-core@...>

Issue #19340 has been reported by MSP-Greg (Greg L).

8 messages 2023/01/14

[#111842] [Ruby master Feature#19347] Add Dir.fchdir — "jeremyevans0 (Jeremy Evans) via ruby-core" <ruby-core@...>

Issue #19347 has been reported by jeremyevans0 (Jeremy Evans).

9 messages 2023/01/16

[#111873] [Ruby master Bug#19351] Promote bundled gems at Ruby 3.3 — "hsbt (Hiroshi SHIBATA) via ruby-core" <ruby-core@...>

Issue #19351 has been reported by hsbt (Hiroshi SHIBATA).

26 messages 2023/01/18

[#111890] [Ruby master Bug#19352] Feature #17391 was not such a good idea because now Ruby 3.2 can not install Rails v5 or v6 if they use webpacker. — "Milella@... (Scott Milella) via ruby-core" <ruby-core@...>

Issue #19352 has been reported by Milella@Hotmail.com (Scott Milella).

16 messages 2023/01/19

[#111953] [Ruby master Bug#19362] #dup on Proc doesn't call initialize_dup — "zverok (Victor Shepelev) via ruby-core" <ruby-core@...>

Issue #19362 has been reported by zverok (Victor Shepelev).

8 messages 2023/01/21

[#111956] [Ruby master Bug#19363] Fix rb_transient_heap_mark: wrong header (T_STRUCT) segfault — "bkuhlmann (Brooke Kuhlmann) via ruby-core" <ruby-core@...>

SXNzdWUgIzE5MzYzIGhhcyBiZWVuIHJlcG9ydGVkIGJ5IGJrdWhsbWFubiAoQnJvb2tlIEt1aGxt

9 messages 2023/01/21

[#111988] [Ruby master Feature#19370] Anonymous parameters for blocks? — "zverok (Victor Shepelev) via ruby-core" <ruby-core@...>

Issue #19370 has been reported by zverok (Victor Shepelev).

10 messages 2023/01/23

[#112041] [Ruby master Feature#19377] Rename Fiber#storage to Fiber.storage — "zverok (Victor Shepelev) via ruby-core" <ruby-core@...>

Issue #19377 has been reported by zverok (Victor Shepelev).

8 messages 2023/01/25

[#112045] [Ruby master Bug#19378] Windows: Use less syscalls for faster require of big gems — "aidog (Andi Idogawa) via ruby-core" <ruby-core@...>

SXNzdWUgIzE5Mzc4IGhhcyBiZWVuIHJlcG9ydGVkIGJ5IGFpZG9nIChBbmRpIElkb2dhd2EpLg0N

7 messages 2023/01/26

[#112048] [Ruby master Bug#19379] Regex: "end pattern with unmatched parenthesis" with Ruby 3.2 and interpolation — "renchap (Renaud Chaput) via ruby-core" <ruby-core@...>

Issue #19379 has been reported by renchap (Renaud Chaput).

8 messages 2023/01/26

[#112058] [Ruby master Bug#19383] Time.now.zone encoding for German display language in Windows is incorrect — "stringsn88keys (Thomas Powell) via ruby-core" <ruby-core@...>

SXNzdWUgIzE5MzgzIGhhcyBiZWVuIHJlcG9ydGVkIGJ5IHN0cmluZ3NuODhrZXlzIChUaG9tYXMg

11 messages 2023/01/26

[#112072] [Ruby master Bug#19386] `test_hmac.rb` of openssl is timeout on RHEL9 — "hsbt (Hiroshi SHIBATA) via ruby-core" <ruby-core@...>

Issue #19386 has been reported by hsbt (Hiroshi SHIBATA).

14 messages 2023/01/27

[#112091] [Ruby master Bug#19387] Issue with ObjectSpace.each_objects not returning IO objects after starting a ractor — "luke-gru (Luke Gruber) via ruby-core" <ruby-core@...>

Issue #19387 has been reported by luke-gru (Luke Gruber).

9 messages 2023/01/27

[#112119] [Ruby master Bug#19392] Endless method vs and/or — "zverok (Victor Shepelev) via ruby-core" <ruby-core@...>

Issue #19392 has been reported by zverok (Victor Shepelev).

20 messages 2023/01/30

[#112146] [Ruby master Bug#19394] cvars in instance of cloned class point to source class's cvars even after class_variable_set on clone — "jamescdavis (James Davis) via ruby-core" <ruby-core@...>

SXNzdWUgIzE5Mzk0IGhhcyBiZWVuIHJlcG9ydGVkIGJ5IGphbWVzY2RhdmlzIChKYW1lcyBEYXZp

8 messages 2023/01/31

[ruby-core:111831] [Ruby master Feature#19333] Setting (Fiber Local|Thread Local|Fiber Storage) to nil should delete value in order to avoid memory leaks.

From: "Eregon (Benoit Daloze) via ruby-core" <ruby-core@...>
Date: 2023-01-16 11:53:41 UTC
List: ruby-core #111831
Issue #19333 has been updated by Eregon (Benoit Daloze).


The current solution/workaround to both of these memory leaks is to use concurrent-ruby's [ThreadLocalVar](https://ruby-concurrency.github.io/concurrent-ruby/master/Concurrent/ThreadLocalVar.html), which is the reason that exists.

The idea to remove on assigning `nil` looks elegant at first sight but has multiple problems:
* The user might use `nil` as a "valid value" and with this change it will be more expensive to remove the fiber local than to just assign, also more expensive if later that fiber local is set again to non-nil (delete+insert vs set+set).
* This is problematic when implementing fiber locals with Shapes, which TruffleRuby currently does, because it causes a lot more Shape polymorphism if fiber locals can be removed, so makes fiber locals slower.
* It doesn't solve the leak itself, it would need a finalizer per fiber local, which is pretty heavy for this feature AND we can't put a finlizer on the Symbol since the Symbol is never collected. So this seems unusable for fiber/thread locals, it would need an extra wrapping object for every usage, or explicitly clearing the fiber local in a begin/ensure/end pattern.

For those reasons I'm rather against this.
I think it's a rare enough problem that it's OK to rely on concurrent-ruby for this feature.
Many fiber/thread local usages are fine because they don't use a random part in the Symbol and so it's a small number of them.
It is a problem when there are many fiber/thread locals, IMO those should concurrent-ruby's ThreadLocalVar/FiberLocalVar.

> That being said, the other option is just not to encourage people to have long running threads that accumulate cruft over time.

That's a hard sell, at least the not having long running threads part. On all current Rubies the time to create a thread is significant and so reusing threads e.g. via a thread pool is a necessity for performance (e.g., in threaded web servers like Puma). (maybe CRuby can do M-N threading in the future, but this is not possible on JVM AFAIK because e.g. ReentrantLock relies on java.lang.Thread identity and not RubyThread/RubyFiber identity)

----------------------------------------
Feature #19333: Setting (Fiber Local|Thread Local|Fiber Storage) to nil should delete value in order to avoid memory leaks.
https://bugs.ruby-lang.org/issues/19333#change-101240

* Author: ioquatix (Samuel Williams)
* Status: Open
* Priority: Normal
----------------------------------------
As it stands, Fiber Locals, Thread Locals and Fiber Storage have no way of deleting key-value associations.

```ruby
100.times do |i|
  name = :"variable-#{i}"
  Thread.current[name] = 10
end
```

Because of this, dynamically generated associations can leak over time. This is worse for things like Threads that might be pooled (or maybe an argument against user-space pooling).

In any case, having a way to delete those associations would allow application code to at least delete the associations when they no longer make sense.

I propose that assigning `nil` to "locals" or "storage" should effectively delete them.

e.g.

```ruby
100.times do |i|
  name = :"variable-#{i}"
  Thread.current[name] = 10
  Thread.current[name] = nil # delete association
end
```

A more invasive alternative would be to define new interfaces like `Thread::Local`, `Fiber::Local` and `Fiber::Storage::Local` (or something) which correctly clean up on GC.



-- 
https://bugs.ruby-lang.org/
 ______________________________________________
 ruby-core mailing list -- ruby-core@ml.ruby-lang.org
 To unsubscribe send an email to ruby-core-leave@ml.ruby-lang.org
 ruby-core info -- https://ml.ruby-lang.org/mailman3/postorius/lists/ruby-core.ml.ruby-lang.org/

In This Thread