Project

General

Profile

Actions

Bug #11826

closed

After prepending a module, rewrite Hash#[] takes no effect for calls like Hash.new[:a]

Added by cichol (Renxiang Cai) almost 9 years ago. Updated almost 9 years ago.

Status:
Closed
Target version:
-
ruby -v:
ruby 2.2.2p95 (2015-04-13 revision 50295) [x64-mingw32]
[ruby-core:72186]

Description

module M
  def self.prepended(clz)
    clz.class_eval do
      def [](k)
        p 1
      end
    end
  end
end

module N
  def self.included(clz)
    clz.class_eval do 
      def []=(k, v)
        p 2
      end
    end
  end
end

class Hash
  prepend M
  include N
end

Hash.new[1]
Hash.new[1]=1

Running this example gives no output, which is expected to be '1 2'.
I am using 'ruby 2.2.2p95 (2015-04-13 revision 50295) [x64-mingw32]'.
Thanks in advanced.

Updated by cichol (Renxiang Cai) almost 9 years ago

  • Subject changed from In 'prepended', 'class_eval' rewriting a method take no effect to After prepending a module, rewrite Hash#[] takes no effect for calls like Hash.new[:a]
  • ruby -v set to ruby 2.2.2p95 (2015-04-13 revision 50295) [x64-mingw32]
module M
end

class Hash
  prepend M
end

class Hash
  def [](k)
    p 3
  end

  def map
    p 4
  end
end

h = {a: 1}
h[:a] #=> return 1 and without print 3
h.send(:[], :a) #=> print 3
h.map # => print 4

A shorter snippet for the issue.
It seems like an issue only for [] method, maybe something about syntax.
But if use 'include' instead of 'prepend', the 'h[:a]' can give an output.
Please help me check if it is a bug, sorry for bothering.

Updated by shugo (Shugo Maeda) almost 9 years ago

  • Status changed from Open to Assigned
  • Assignee set to ko1 (Koichi Sasada)

cichol tsai wrote:

It seems like an issue only for [] method, maybe something about syntax.
But if use 'include' instead of 'prepend', the 'h[:a]' can give an output.
Please help me check if it is a bug, sorry for bothering.

It seems that redefinition check of optimized methods doesn't work properly for prepended classes.
Is the following patch correct, ko1?

diff --git a/vm.c b/vm.c
index a03e068..cf528e8 100644
--- a/vm.c
+++ b/vm.c
@@ -1400,6 +1400,9 @@ static void
 rb_vm_check_redefinition_opt_method(const rb_method_entry_t *me, VALUE klass)
 {
     st_data_t bop;
+    if (RB_TYPE_P(klass, T_ICLASS) && FL_TEST(klass, RICLASS_IS_ORIGIN)) {
+	klass = RBASIC_CLASS(klass);
+    }
     if (me->def->type == VM_METHOD_TYPE_CFUNC) {
 	if (st_lookup(vm_opt_method_table, (st_data_t)me, &bop)) {
 	    int flag = vm_redefinition_check_flag(klass);
Actions #3

Updated by shugo (Shugo Maeda) almost 9 years ago

  • Status changed from Assigned to Closed

Applied in changeset r53173.


  • vm.c (rb_vm_check_redefinition_opt_method): should check the real
    class instead of the origin iclass.
    [ruby-core:72188] [Bug #11826]

Updated by shugo (Shugo Maeda) almost 9 years ago

  • Backport changed from 2.0.0: UNKNOWN, 2.1: UNKNOWN, 2.2: UNKNOWN to 2.0.0: REQUIRED, 2.1: REQUIRED, 2.2: REQUIRED

Shugo Maeda wrote:

It seems that redefinition check of optimized methods doesn't work properly for prepended classes.
Is the following patch correct, ko1?

Committed it with a test in r53173 after checking that make exam succeeded.
Revert it if it's wrong.

Actions

Also available in: Atom PDF

Like0
Like0Like0Like0Like0