Re: Regression in GCJ 3.3 in parsing property files

[Tom Tromey <tromey at redhat dot com>]

>>>>> "Ranjit" == Ranjit Mathew <rmathew at hotmail dot com> writes:

Ranjit> Actually, I have found more problems with parseURL( ), so it's OK
Ranjit> if this patch is not applied right away - not that it's wrong, but
Ranjit> I might as well submit a fuller patch.

Ok, I'll drop this patch.  If you change your mind, resend it.

Ranjit> What I'm referring to is the fact that this method gets confused
Ranjit> by other colons in the path of the URL when the URL is something
Ranjit> like "file:/c:/temp/foo.txt". And I think this case is valid
Ranjit> only for "file" URLs.

Stuff like this deserves a Mauve test, though I guess it would have to
be host-specific.

Ranjit> Can someone with Sun's or IBM's JRE on Linux please tell me
Ranjit> what it is for a path like "/foo/bar" - is it
Ranjit> "file:///foo/bar" or just "file:/foo/bar"?

Here you go:

fleche. cat foo.java 
import java.io.*;

class foo
{
  public static void main(String[] args) throws Throwable
  { 
    System.out.println(new File("/tmp/foo.java").toURL());
  }
}
fleche. javac foo.java 
fleche. java foo
file:/tmp/foo.java


This is weird, though.  Isn't this invalid according to the URL standard?
(Which I haven't looked at since 1996... so don't think I know what
I'm talking about.)

Tom