Name: bsC130419 Date: 06/21/2001


java version "1.4.0-beta"
Java(TM) 2 Runtime Environment, Standard Edition (build 1.4.0-beta-b62)
Java HotSpot(TM) Client VM (build 1.4.0-beta-b62, mixed mode)

The implementation of java.io.File.toURL appears to be buggy in that it does not
consider that the filename it is converting to a URL may already have a variety
of special characters which need to be escaped for the URL to be valid. At least
as of 1.3, it simply calls the URL constructor with the file path more or less
unmodified. Either File.toURL or the URL constructor needs to do some escaping
here; not clear which it is, but as the java.net.URL documentation says nothing
about the meaning or interpretation of escape characters that I can see, I would
guess it is the responsibility of every protocol to deal with escapes in an
appropriate way; so file: specifically should do something.

The specific case which I know of being problematic is when the filename
contains a '#' character. This is not so unusual; various backup file often have
this; and there is at least one popular version-control system which names its
versioned directories using a '#' character, meaning every file controlled by
this system is affected by this bug! The problem with '#' of course is that if
unescaped, it indicates the fragment in the URL, and the end of the path part.

The following Java program demonstrates this (on 1.4 beta on Linux):

------%<------
import java.io.*;
import java.net.*;
public class TestHashMarkInFileToURL {
    public static void main (String args[]) throws IOException {
        testName ("testinghashmark");
        testName ("testing#hashmark");
        testURL (new URL ("file:/tmp/testing%23hashmark/test.txt"));
    }
    private static void testName (String n) throws IOException {
        File f = new File ("/tmp/" + n);
        f.mkdirs ();
        File f2 = new File (f, "test.txt");
        f2.createNewFile ();
        System.err.println("file: " + f2);
        System.err.println("file exists: " + (f2.isFile () && f2.canRead ()));
        URL u = f2.toURL ();
        testURL (u);
    }
    private static void testURL (URL u) {
        System.err.println("url=" + u);
        System.err.println("file part: " + u.getPath ());
        try {
            URLConnection c = u.openConnection ();
            c.connect ();
            System.err.println("opened connection; content length: " +
c.getContentLength ());
        } catch (IOException e) {
            System.err.println("connection failed: " + e);
        }
    }
}
------%<------

I get:

------%<------
file: /tmp/testinghashmark/test.txt
file exists: true
url=file:/tmp/testinghashmark/test.txt
file part: /tmp/testinghashmark/test.txt
opened connection; content length: 0
file: /tmp/testing#hashmark/test.txt
file exists: true
url=file:/tmp/testing#hashmark/test.txt
file part: /tmp/testing
connection failed: java.io.FileNotFoundException: /tmp/testing (No such file or
directory)
url=file:/tmp/testing%23hashmark/test.txt
file part: /tmp/testing%23hashmark/test.txt
opened connection; content length: 0
------%<------

Note that the use of File.toURL on /tmp/testing#hashmark/test.txt produced an
invalid URL that could not be opened. The explicitly formed URL with the correct
"%23" escape syntax worked.
(Review ID: 127084)
======================================================================