Re: [ ruby-Bugs-2131 ] ruby ( v183) bcc32: using Socket.new with timeout -> files not closed

From: H.Yamamoto (ocean@...)


>
> Can you try this code on Win98? On my machine, this outputs "OK".
>


Good news? -- Mine too gives "OK"

>
> And special adhok patch. I looked into borland RTL source code and found that
> this magic number tells _close() to skip CloseHandle. But I cannot call this "fix"...
>


I deleted these last five lines of the patch - is that OK ?
 (It's replacing 12 lines with 8 ... but early EOF reached)

> @@ -3555,12 +3561,8 @@ unixtime_to_filetime(time_t time, FILETI
>      return 0;
>  }
>
>
> ////////////////////////////////////////////////////
>

After applying the patch, our script now gives:

["Net::HTTP", "1.126", "1.1"]
D:/RUBY/SRC_CVSINST/lib/ruby/1.9/net/http.rb:507:in `initialize': getaddrinfo: no address associated with hostname. (SocketError)
 from D:/RUBY/SRC_CVSINST/lib/ruby/1.9/net/http.rb:507:in `connect'
 from D:/RUBY/SRC_CVSINST/lib/ruby/1.9/net/http.rb:507:in `timeout'
 from D:/RUBY/SRC_CVSINST/lib/ruby/1.9/timeout.rb:63:in `timeout'
 from D:/RUBY/SRC_CVSINST/lib/ruby/1.9/net/http.rb:507:in `connect'
 from D:/RUBY/SRC_CVSINST/lib/ruby/1.9/net/http.rb:500:in `do_start'
 from D:/RUBY/SRC_CVSINST/lib/ruby/1.9/net/http.rb:489:in `start'
 from D:/RUBY/SRC_CVSINST/lib/ruby/1.9/net/http.rb:982:in `request'
 from D:/RUBY/SRC_CVSINST/lib/ruby/1.9/net/http.rb:716:in `get'
 from C:/TEMP/rbF223.TMP:5


Well, it's a much better error message than I've had since Nov. 2003  ;-)  ...
http://blade.nagaokaut.ac.jp/cgi-bin/scat.rb/ruby/ruby-talk/85522


>
> I think the real problem is, ruby assumes file descripter and socket decripter
> are equivarent. This is true on unix system, but on some platforms. cannot call
> read(2) for socket, close(2) for socket, select(2) for file....
>

Yes.
It's only Windows people, like myself, who learn that sockets
are 'awkward'.  BSD socket design is beautifully simple and I
believe that *nix implements that design faithfully.


> So I think ideal IO system should be something like this.
>
>  [...]
>
> It's just an idea though.
>


Anything that works on all supported platforms is good for me ;-)

Thanks, Yamamoto-san.  IMHO, you are *very* near to a fix.


daz