[PATCH] tempfile.rb (was Re: Bug in open-uri under win32 (?))

On Sun, May 16, 2004 at 05:00:33PM +0900, Mauricio Fern疣dez wrote:
> I'm told that this doesn't download the file correctly under windows
> (only a few KB are dled). On visual inspection, the associated Tempfile
> is not complete either, so the same behavior is observed when using
> is.gets(nil) or a loop with eof?.

The bug has been traced back to Tempfile, which doesn't use the binary
mode for the underlying file. IMHO it should since the tempfile isn't
meant to be seen by other processes, so CRLF translation (and other
oddities such as ETB) makes no sense.


--- tempfile.rb 2004-04-17 20:41:57.000000000 +0200
+++ tempfile.rb.new     2004-05-16 13:16:08.000000000 +0200
@@ -52,7 +52,9 @@
     @clean_proc = Tempfile.callback(@data)
     ObjectSpace.define_finalizer(self, @clean_proc)

-    @tmpfile = File.open(tmpname, File::RDWR|File::CREAT|File::EXCL, 0600)
+    @tmpfile = File.open(tmpname,
+                         File::RDWR|File::CREAT|File::EXCL|
+                         (File::BINARY rescue 0), 0600)
     @tmpname = tmpname
     @@cleanlist << @tmpname
     @data[1] = @tmpfile
@@ -66,10 +68,10 @@
     Dir.rmdir(lock)
   end

-  # Opens or reopens the file with mode "r+".
+  # Opens or reopens the file with mode "rb+".
   def open
     @tmpfile.close if @tmpfile
-    @tmpfile = File.open(@tmpname, 'r+')
+    @tmpfile = File.open(@tmpname, 'rb+')
     @data[1] = @tmpfile
     __setobj__(@tmpfile)
   end

-- 
Running Debian GNU/Linux Sid (unstable)
batsman dot geo at yahoo dot com

Are Linux users lemmings collectively jumping off of the cliff of
reliable, well-engineered commercial software?
	-- Matt Welsh