langtools/test/com/sun/javadoc/WindowTitles/WindowTitles.java
changeset 10 06bc494ca11e
child 5520 86e4b9a9da40
--- /dev/null	Thu Jan 01 00:00:00 1970 +0000
+++ b/langtools/test/com/sun/javadoc/WindowTitles/WindowTitles.java	Sat Dec 01 00:00:00 2007 +0000
@@ -0,0 +1,221 @@
+/*
+ * Copyright 2002-2005 Sun Microsystems, Inc.  All Rights Reserved.
+ * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
+ *
+ * This code is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 only, as
+ * published by the Free Software Foundation.
+ *
+ * This code is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
+ * version 2 for more details (a copy is included in the LICENSE file that
+ * accompanied this code).
+ *
+ * You should have received a copy of the GNU General Public License version
+ * 2 along with this work; if not, write to the Free Software Foundation,
+ * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
+ *
+ * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
+ * CA 95054 USA or visit www.sun.com if you need additional information or
+ * have any questions.
+ */
+
+/*
+ * @test
+ * @bug 4530730
+ * @summary stddoclet: With frames off, window titles have "()" appended
+ * @author dkramer
+ * @run main WindowTitles
+ */
+
+
+import com.sun.javadoc.*;
+import java.util.*;
+import java.io.*;
+
+// If needing regular expression pattern matching,
+// see /java/pubs/dev/linkfix/src/LinkFix.java
+
+/**
+ * Runs javadoc and runs regression tests on the resulting HTML.
+ * It reads each file, complete with newlines, into a string to easily
+ * find strings that contain newlines.
+ */
+public class WindowTitles
+{
+    private static final String BUGID = "4530730";
+    private static final String BUGNAME = "WindowTitles";
+    private static final String FS = System.getProperty("file.separator");
+    private static final String PS = System.getProperty("path.separator");
+    private static final String LS = System.getProperty("line.separator");
+    private static final String TMPDIR_STRING1 = "." + FS + "docs1" + FS;
+    private static final String TMPDIR_STRING2 = "." + FS + "docs2" + FS;
+
+    // Subtest number.  Needed because runResultsOnHTML is run twice, and subtestNum
+    // should increment across subtest runs.
+    public static int subtestNum = 0;
+    public static int numSubtestsPassed = 0;
+
+    // Entry point
+    public static void main(String[] args) {
+
+        // Directory that contains source files that javadoc runs on
+        String srcdir = System.getProperty("test.src", ".");
+
+        // Test for all cases except the split index page
+        runJavadoc(new String[] {"-d", TMPDIR_STRING1,
+                                 "-use",
+                                 "-sourcepath", srcdir,
+                                 "p1", "p2"});
+        runTestsOnHTML(testArray);
+
+        // Test only for the split-index case (and run on only one package)
+        System.out.println("");  // blank line
+        runJavadoc(new String[] {"-d", TMPDIR_STRING2,
+                                 "-splitindex",
+                                 "-sourcepath", System.getProperty("test.src", "."),
+                                 "p1"});
+        runTestsOnHTML(testSplitIndexArray);
+
+        printSummary();
+    }
+
+    /** Run javadoc */
+    public static void runJavadoc(String[] javadocArgs) {
+        if (com.sun.tools.javadoc.Main.execute(javadocArgs) != 0) {
+            throw new Error("Javadoc failed to execute");
+        }
+    }
+
+    /**
+     * Assign value for [ stringToFind, filename ]
+     * NOTE: The standard doclet uses platform-specific line separators (LS)
+     */
+    private static final String[][] testArray = {
+
+            { "<TITLE>" + LS + "Overview" + LS + "</TITLE>",
+                    TMPDIR_STRING1 + "overview-summary.html"                  },
+
+            { "<TITLE>" + LS + "Class Hierarchy" + LS + "</TITLE>",
+                    TMPDIR_STRING1 + "overview-tree.html"                     },
+
+            { "<TITLE>" + LS + "Overview List" + LS + "</TITLE>",
+                    TMPDIR_STRING1 + "overview-frame.html"                    },
+
+            { "<TITLE>" + LS + "p1" + LS + "</TITLE>",
+                    TMPDIR_STRING1 + "p1" + FS + "package-summary.html"       },
+
+            { "<TITLE>" + LS + "p1" + LS + "</TITLE>",
+                    TMPDIR_STRING1 + "p1" + FS + "package-frame.html"         },
+
+            { "<TITLE>" + LS + "p1 Class Hierarchy" + LS + "</TITLE>",
+                    TMPDIR_STRING1 + "p1" + FS + "package-tree.html"          },
+
+            { "<TITLE>" + LS + "Uses of Package p1" + LS + "</TITLE>",
+                    TMPDIR_STRING1 + "p1" + FS + "package-use.html"           },
+
+            { "<TITLE>" + LS + "C1" + LS + "</TITLE>",
+                    TMPDIR_STRING1 + "p1" + FS + "C1.html"                    },
+
+            { "<TITLE>" + LS + "All Classes" + LS + "</TITLE>",
+                    TMPDIR_STRING1 + "allclasses-frame.html"                  },
+
+            { "<TITLE>" + LS + "All Classes" + LS + "</TITLE>",
+                    TMPDIR_STRING1 + "allclasses-noframe.html"                },
+
+            { "<TITLE>" + LS + "Constant Field Values" + LS + "</TITLE>",
+                    TMPDIR_STRING1 + "constant-values.html"                   },
+
+            { "<TITLE>" + LS + "Deprecated List" + LS + "</TITLE>",
+                    TMPDIR_STRING1 + "deprecated-list.html"                   },
+
+            { "<TITLE>" + LS + "Serialized Form" + LS + "</TITLE>",
+                    TMPDIR_STRING1 + "serialized-form.html"                   },
+
+            { "<TITLE>" + LS + "API Help" + LS + "</TITLE>",
+                    TMPDIR_STRING1 + "help-doc.html"                          },
+
+            { "<TITLE>" + LS + "Index" + LS + "</TITLE>",
+                    TMPDIR_STRING1 + "index-all.html"                         },
+
+            { "<TITLE>" + LS + "Uses of Class p1.C1" + LS + "</TITLE>",
+                    TMPDIR_STRING1 + "p1" + FS + "class-use" + FS + "C1.html" },
+        };
+
+    /**
+     * Assign value for [ stringToFind, filename ] for split index page
+     */
+    private static final String[][] testSplitIndexArray = {
+            { "<TITLE>" + LS + "C-Index" + LS + "</TITLE>",
+                    TMPDIR_STRING2 + "index-files" + FS + "index-1.html"       },
+        };
+
+    public static void runTestsOnHTML(String[][] testArray) {
+
+        for (int i = 0; i < testArray.length; i++) {
+
+            subtestNum += 1;
+
+            // Read contents of file into a string
+            String fileString = readFileToString(testArray[i][1]);
+
+            // Get string to find
+            String stringToFind = testArray[i][0];
+
+            // Find string in file's contents
+            if (findString(fileString, stringToFind) == -1) {
+                System.out.println("\nSub-test " + (subtestNum)
+                    + " for bug " + BUGID + " (" + BUGNAME + ") FAILED\n"
+                    + "when searching for:\n"
+                    + stringToFind);
+            } else {
+                numSubtestsPassed += 1;
+                System.out.println("\nSub-test " + (subtestNum) + " passed:\n" + stringToFind);
+            }
+        }
+    }
+
+    public static void printSummary() {
+        if ( numSubtestsPassed == subtestNum ) {
+            System.out.println("\nAll " + numSubtestsPassed + " subtests passed");
+        } else {
+            throw new Error("\n" + (subtestNum - numSubtestsPassed) + " of " + (subtestNum)
+                             + " subtests failed for bug " + BUGID + " (" + BUGNAME + ")\n");
+        }
+    }
+
+    // Read the file into a String
+    public static String readFileToString(String filename) {
+        try {
+            File file = new File(filename);
+            if ( !file.exists() ) {
+                System.out.println("\nFILE DOES NOT EXIST: " + filename);
+            }
+            BufferedReader in = new BufferedReader(new FileReader(file));
+
+            // Create an array of characters the size of the file
+            char[] allChars = new char[(int)file.length()];
+
+            // Read the characters into the allChars array
+            in.read(allChars, 0, (int)file.length());
+            in.close();
+
+            // Convert to a string
+            String allCharsString = new String(allChars);
+
+            return allCharsString;
+
+        } catch (FileNotFoundException e) {
+            System.err.println(e);
+            return "";
+        } catch (IOException e) {
+            System.err.println(e);
+            return "";
+        }
+    }
+
+    public static int findString(String fileString, String stringToFind) {
+        return fileString.indexOf(stringToFind);
+    }
+}