From: Laurent Sansonetti <laurent@...>
Date: 2003-05-16T16:48:18+09:00
Subject: Re: PTY: still problems (+patch)

Hi Nobu,

nobu.nokada@softhome.net wrote:
> I noticed a bit different problem on Linux, EIO occurs.
> 

I can test the code on Linux, and help you to fix this problem.

The most difficult part here is to steal my girlfriend's laptop ;-)

>>-    rb_thread_kill(info->thread);
>>+    rb_funcall(info->thread, rb_intern("join"), 0);
> 
> 
> Here, it might be better to use "value".
> 

You're right, rb_thread_value() is meaningful here.

What do you think of the following patch?

pinux@natalie ~/ruby> cvs diff -u eval.c intern.h ext/pty/pty.c
Index: eval.c
===================================================================
RCS file: /src/ruby/eval.c,v
retrieving revision 1.430
diff -u -r1.430 eval.c
--- eval.c      13 May 2003 05:53:08 -0000      1.430
+++ eval.c      16 May 2003 07:43:23 -0000
@@ -9178,7 +9178,7 @@
      return rb_thread_start_0(rb_thread_yield, args, 
rb_thread_alloc(klass));
  }

-static VALUE
+VALUE
  rb_thread_value(thread)
      VALUE thread;
  {
Index: intern.h
===================================================================
RCS file: /src/ruby/intern.h,v
retrieving revision 1.119
diff -u -r1.119 intern.h
--- intern.h    13 May 2003 05:53:08 -0000      1.119
+++ intern.h    16 May 2003 07:43:24 -0000
@@ -212,6 +212,7 @@
  VALUE rb_thread_local_aref _((VALUE, ID));
  VALUE rb_thread_local_aset _((VALUE, ID, VALUE));
  void rb_thread_atfork _((void));
+VALUE rb_thread_value _((VALUE));
  /* file.c */
  int eaccess _((const char*, int));
  VALUE rb_file_s_expand_path _((int, VALUE *));
Index: ext/pty/pty.c
===================================================================
RCS file: /src/ruby/ext/pty/pty.c,v
retrieving revision 1.15
diff -u -r1.15 pty.c
--- ext/pty/pty.c       10 Mar 2003 15:05:18 -0000      1.15
+++ ext/pty/pty.c       16 May 2003 07:43:26 -0000
@@ -299,7 +299,7 @@
  pty_finalize_syswait(info)
      struct pty_info *info;
  {
-    rb_thread_kill(info->thread);
+    rb_thread_value(info->thread);
      rb_detach_process(info->child_pid);
      return Qnil;
  }

It works nicely here.

> I agree to export it (and rb_thread_value maybe), but could you
> explain why 'limit' is needed here?
> 

After some reflection, there is no need to 'limit' the blocking.  I was 
a bit tired yesterday ;-)

--
Laurent