Fix exception caused by non-ascii paths

doc/CHANGELOG.rst

@@ -19,6 +19,9 @@ Latest Changes:
     For editable installs, ``python setup.py --enable-tracing develop``
     must now be done with ``python setup.py develop --enable-tracing``.
 
+  - Fixed a problem that caused ``dir(jpype.JPackage("mypackage"))`` to fail if
+    the class path contained non-ascii characters. See issue #1194.
+
 - **1.5.0 - 2023-04-03**
 
   - Support for Python 3.12

native/java/org/jpype/pkg/JPypePackageManager.java

@@ -464,7 +464,21 @@ private static URI toURI(Path path)
     URI uri = path.toUri();
     if (uri.getScheme().equals("jar") && uri.toString().contains("%2520"))
       uri = URI.create("jar:" + uri.getRawSchemeSpecificPart().replaceAll("%25", "%"));
-    return uri;
+
+    // `toASCIIString` ensures the URI is URL encoded with only ascii
+    // characters. This avoids issues in `sun.nio.fs.UnixUriUtils.fromUri` that
+    // naively uses `uri.getRawPath()` despite the possibility that it contains
+    // non-ascii characters that will cause errors. By using `toASCIIString` and
+    // re-wrapping it in a URI object we ensure that the URI is properly
+    // encoded. See: https://github.com/jpype-project/jpype/issues/1194
+    try {
+      return new URI(uri.toASCIIString());
+    } catch (Exception e) {
+      // This exception *should* never occur as we are re-encoding a valid URI.
+      // Throwing a runtime exception avoids java exception handling boilerplate
+      // for a situation that *should* never occur.
+      throw new RuntimeException("Failed to encode URI: " + uri, e);
+    }
   }
 //</editor-fold>
 }

test/jpypetest/test_non_ascii_paths.py

@@ -0,0 +1,45 @@
+"""
+Regression test for https://github.com/jpype-project/jpype/issues/1194
+"""
+
+import unittest
+import jpype
+from pathlib import Path
+import shutil
+from multiprocessing import get_context, Queue, Process
+
+
+class TestNonAsciiPaths(unittest.TestCase):
+
+    def setUp(self):
+        # Create a copy of the mrjar.jar file with a non-ASCII path.
+        mrjar_path = Path("test/jar/mrjar.jar")
+        self.mrjar_non_ascii = Path("test/jar/à_non_ascii/mrjar.jar")
+        self.mrjar_non_ascii.parent.mkdir(exist_ok=True)
+        shutil.copy(mrjar_path, self.mrjar_non_ascii)
+
+    def tearDown(self):
+        # Clean up the non-ASCII path.
+        self.mrjar_non_ascii.unlink()
+        self.mrjar_non_ascii.parent.rmdir()
+
+    @staticmethod
+    def _subproc_test_non_ascii_paths(path: str, queue: Queue):
+        # Can `dir` be called on a package with a non-ASCII path?
+        jpype.addClassPath(path)
+        jpype.startJVM(jpype.getDefaultJVMPath())
+        queue.put(dir(jpype.JPackage("org.jpype.mrjar")))
+        jpype.shutdownJVM()
+
+    def test_non_ascii_paths(self):
+        # Uses a separate process to avoid polluting the main process with
+        # a JVM instance, since restarting the JVM is not supported.
+        ctx = get_context("spawn")
+        queue: Queue = ctx.Queue()
+        path = self.mrjar_non_ascii.absolute().as_posix()
+        proc: Process = ctx.Process(
+            target=self._subproc_test_non_ascii_paths, args=(path, queue)
+        )
+        proc.start()
+        proc.join(5)
+        assert queue.get() == ["A", "B", "sub"]