RE: Patch: File.deleteOnExit()

[Jeroen Frijters]

Hi,

Thanks for the explanation! In that case I would suggest (and this hopefully 
also addresses Mark's points):

+package java.io;
+
+final class DeleteFileHelper extends Thread
+{
+  private static ArrayList filesToDelete;
+
+  static synchronized void add(File file)
+  {
+    if (filesToDelete == null)
+    {
+      filesToDelete = new ArrayList();
+      AccessController.doPrivileged(new PrivilegedAction()
+      {
+        public Object run()
+        {
+          try
+          {
+            Runtime.getRuntime().addShutdownHook(new DeleteFileHelper());
+          }
+          catch (IllegalStateException x)
+          {
+            // shutdown is already in progress, so we can't register ours
+          }
+          return null;
+        }
+      });
+    }
+    filesToDelete.add(file);
+  }
+
+  private static synchronized deleteFiles()
+  {
+    Iterator it = filesToDelete.iterator();
+       
+    while (it.hasNext())
+      {
+       try
+         {
+           File file = (File) it.next();
+           file.delete();
+         }
+       catch (Exception e)
+         {
+           // Do nothing here.
+         }
+      }
+  }
+
+  private DeleteFileHelper()
+  {
+  }
+
+  public void run()
+  {
+    deleteFiles();
+  }
+}

Regards,
Jeroen

> -----Original Message-----
> From: Ingo Prötel [mailto:address@hidden 
> Sent: Tuesday, May 18, 2004 08:55
> To: Jeroen Frijters
> Cc: Michael Koch; address@hidden
> Subject: Re: Patch: File.deleteOnExit()
> 
> Hi,
> 
> static initializers in our case (a real-time vm) run at 
> startup. Otherwise we would need to add all 
> static initializers to the worst case execution time for 
> everything. For real-time purposes we need 
> to have all classes loaded and initialized before the 
> application starts.
> If we have the creation in a method that will never gets 
> executed we can remove the code.
> So in general we prefere lazy initialization over static 
> initializers. If we need good response we 
> can force an early initialization but we do not need to 
> commit resources that might never be used.
> 
> Cheers,
> ingo
> 
> Jeroen Frijters wrote:
> > Hi,
> > 
> > The static initializer only runs when add() is called for 
> the first time.
> > 
> > Regards,
> > Jeroen 
> > 
> > 
> >>-----Original Message-----
> >>From: Ingo Prötel [mailto:address@hidden 
> >>Sent: Monday, May 17, 2004 19:27
> >>To: Jeroen Frijters
> >>Cc: Michael Koch; address@hidden
> >>Subject: Re: Patch: File.deleteOnExit()
> >>
> >>Hi,
> >>
> >>I would vote for Michaels solution. Having a static 
> >>initializer that creates a Thread which might 
> >>never be used seems wasteful.
> >>
> >>Cheers,
> >>ingo
> >>
> >>Jeroen Frijters wrote:
> >>
> >>>Michael Koch wrote:
> >>>
> >>>
> >>>>Am Montag, 17. Mai 2004 12:40 schrieb Jeroen Frijters:
> >>>>
> >>>>
> >>>>>Michael Koch wrote:
> >>>>>
> >>>>>
> >>>>>>The attached patch written by Guilhem Lavaux implements
> >>>>>>File.deleteOnExit() in an easy way.
> >>>>>>
> >>>>>>Ok to commit ?
> >>>>>
> >>>>>It doesn't make me happy to preallocate all these objects (and in
> >>>>>general, the less static initializers the better). Shouldn't we
> >>>>>just move DeleteFileHelper to gnu.java.io (or whatever) 
> and make it
> >>>>>a proper singleton? That way the singleton will be 
> allocated lazily
> >>>>>upon first use of File.deleteOnExit().
> >>>>
> >>>>Here is a revised patch for review:
> >>>
> >>>
> >>>Thanks! I think it would be slightly better to do:
> >>>
> >>>+public final class DeleteFileHelper extends Thread
> >>>+{
> >>>+  private static LinkedList filesToDelete = new LinkedList();
> >>>+
> >>>+  static
> >>>+  {
> >>>+    Runtime.getRuntime().addShutdownHook(new DeleteFileHelper());
> >>>+  }
> >>>+
> >>>+  public static void add(File file)
> >>>+  {
> >>>+    filesToDelete.add(file);
> >>>+  }
> >>>+
> >>>+  private DeleteFileHelper()
> >>>+  {
> >>>+  }
> >>>
> >>>Since we already had a static initializer (to initialize the static
> >>>filesToDelete field), it's better to leverage it to do the call to
> >>>addShutdownHook as well. I also made the constructor private.
> >>>
> >>>Regards,
> >>>Jeroen
> >>>
> >>>
> >>>_______________________________________________
> >>>Commit-classpath mailing list
> >>>address@hidden
> >>>http://mail.gnu.org/mailman/listinfo/commit-classpath
> >>>
> >>
> >>-- 
> >>Ingo Prötel                                          
> address@hidden
> >>aicas GmbH                                        
> http://www.aicas.com
> >>Haid-und-Neu-Str. 18                        phone   +49 721 
> 663 968-32
> >>76131 Karlsruhe                             fax     +49 721 
> 663 968-93
> >>Germany
> > 
> > 
> > 
> 
> -- 
> Ingo Prötel                                          address@hidden
> aicas GmbH                                        http://www.aicas.com
> Haid-und-Neu-Str. 18                        phone   +49 721 663 968-32
> 76131 Karlsruhe                             fax     +49 721 663 968-93
> Germany
>