[PATCH] subclassing IO works properly for pipe/popen

[Eric Wong]

Hi all,

I noticed subclassing IO didn't work as expected for pipe/popen and
pushed the following patch out to the "io-subclass-fix" branch of
git://git.bogomips.org/rubinius.git

I'm not sure about the spec location/style, but the fix itself is
trivially correct to me. Let me know if you have any
questions/comments.

From 69e7a1cc0a7b2187f13245930980171b46afcc6f Mon Sep 17 00:00:00 2001
From: Eric Wong <e...@yhbt.net>
Date: Sat, 25 Sep 2010 04:31:47 +0000
Subject: [PATCH] subclassing IO works properly for pipe/popen

Subclasses of IO should return objects of the
subclass, not the core IO object.
---
kernel/common/io.rb | 8 ++--
spec/ruby/core/io/subclass_singleton_spec.rb | 41 ++++++++++++++++++++++++++
2 files changed, 45 insertions(+), 4 deletions(-)
create mode 100644 spec/ruby/core/io/subclass_singleton_spec.rb

diff --git a/kernel/common/io.rb b/kernel/common/io.rb
index 8907631..67c7bee 100644
--- a/kernel/common/io.rb
+++ b/kernel/common/io.rb
@@ -340,8 +340,8 @@ class IO
# Sending message to parent
# Parent got: <Hi Dad>
def self.pipe
- lhs = IO.allocate
- rhs = IO.allocate
+ lhs = allocate
+ rhs = allocate
connect_pipe(lhs, rhs)
lhs.sync = true
rhs.sync = true
@@ -401,8 +401,8 @@ class IO
readable = true
end

- pa_read, ch_write = IO.pipe if readable
- ch_read, pa_write = IO.pipe if writable
+ pa_read, ch_write = pipe if readable
+ ch_read, pa_write = pipe if writable

pid = Process.fork

diff --git a/spec/ruby/core/io/subclass_singleton_spec.rb b/spec/ruby/core/io/subclass_singleton_spec.rb
new file mode 100644
index 0000000..d7ac05b
--- /dev/null
+++ b/spec/ruby/core/io/subclass_singleton_spec.rb
@@ -0,0 +1,41 @@
+require File.expand_path('../../../spec_helper', __FILE__)
+
+class SubIO < IO
+end
+
+describe "subclassed IO singleton methods exposes subclassed objects" do
+
+ it "lets subclassed popen yield an object of subclass" do
+ SubIO.popen("true", "r") do |io|
+ io.should be_an_instance_of(SubIO)
+ end
+ end
+
+ it "lets subclassed popen return an object of subclass" do
+ io = SubIO.popen("true", "r")
+ io.should be_an_instance_of(SubIO)
+ io.close
+ end
+
+ it "returns a pipe belonging to subclasses" do
+ r, w = SubIO.pipe
+ r.should be_an_instance_of(SubIO)
+ w.should be_an_instance_of(SubIO)
+ w.close
+ r.close
+ end
+
+ it "creates objects belonging to subclasses from file descriptors" do
+ r, w = IO.pipe
+ r.should be_an_instance_of(IO)
+ w.should be_an_instance_of(IO)
+ sr = SubIO.for_fd(r.fileno)
+ sr.should be_an_instance_of(SubIO)
+ sr.close
+ begin
+ r.close
+ rescue Errno::EBADF
+ end
+ w.close
+ end
+end
--
Eric Wong

[Dirkjan Bussink]

Hi Eric,

Could you create an issue on github issues at http://github.com/evanphx/rubinius/issues with all this information? That is actively used and tracked for issues so it's probably a better place to handle it.

--
Regards,

Dirkjan Bussink

> --
> --- !ruby/object:MailingList
> name: rubinius-dev
> view: http://groups.google.com/group/rubinius-dev?hl=en
> post: rubini...@googlegroups.com
> unsubscribe: rubinius-dev...@googlegroups.com

[Eric Wong]

Dirkjan Bussink <d.bu...@gmail.com> wrote:
> Hi Eric,
>
> Could you create an issue on github issues at
> http://github.com/evanphx/rubinius/issues with all this information?
> That is actively used and tracked for issues so it's probably a better
> place to handle it.

Created http://github.com/evanphx/rubinius/issues/500

I wasn't sure about how/where to deal with the spec, so I was hoping
to get some feedback there. I also much prefer to stick with mutt
than any web browser :>

--
Eric Wong