From: TAKAHASHI Masayoshi <maki@...>
Date: 2002-10-01T08:00:04+09:00
Subject: ruby-dev summary 18320-18406

Hi all,

This is a summary of ruby-dev ML in these days.

---- ruby-dev #18320-18406 (2002-09-21 ... 2002-09-28) ----

  [ruby-dev:18354] parser reentrancy

Nobu Nakada pointed out that allocation is called in
Ruby 1.7 when an instance of pre-defined class is created,
so thread switching may happen during parsing.
It breaks parser's reentrancy.

  class << String
    alias_method(:__allocate__, :allocate)
    def allocate(*args)
      s = __allocate__(*args)
      c = @allocate_hook
      @allocate_hook = nil
      c.call if c
      s
    end
    def allocate_hook(&block)
      @allocate_hook = block
    end
  end
  
  x = 0
  s = "p x"
  String.allocate_hook do
    p __FILE__
    # eval("x = 1")     # delete this comment, then...
  end
  eval(s)               # raise syntax error.


Matz said that allocation would be canceled and
other API would be added.


  [ruby-dev:18361] compile parse.y with -Wall

Nobu Nakada (also) pointed out that there were many warnings
when he compiled parse.y using gcc with -Wall option.

Some warnings were fixed, but Matz wanted to ignore
"assignment in condition" warnings because of 'taste'.

Akinori MUSHA suggested that double parentheses around
assignment supress warnings:

	if ((a = b(c))) /*...*/;

and also suggested that sometimes someone checked
these warnings and added parentheses.


  [ruby-dev:18391] pstore.rb can make a broken store

YANAGAWA Kazuhisa pointed out that Pstore can break
stored data because pstore has no locking system.

He posted a patch, but Tanaka Akira noticed that
the patch is not enough.
Matz told that pstore.rb is just 'proof of concept'
(at least when he wrote it) and he would change
its maintainer of pstore if someone will want to
improve it.



TAKAHASHI 'Maki' Masayoshi     E-mail: maki@rubycolor.org