2004-10-22 Michael Koch <konqueror@gmx.de>
* java/lang/Math.java * java/lang/StackTraceElement.java * java/nio/Buffer.java * java/nio/ByteBuffer.java * java/nio/CharBuffer.java * java/nio/DoubleBuffer.java * java/nio/FloatBuffer.java * java/nio/IntBuffer.java * java/nio/LongBuffer.java * java/nio/ShortBuffer.java * java/nio/charset/Charset.java * java/rmi/server/RMIClassLoader.java * java/rmi/server/RMISocketFactory.java * java/security/Policy.java * java/text/ChoiceFormat.java * java/text/CollationElementIterator.java * java/text/DateFormat.java * java/text/DecimalFormat.java * java/text/DecimalFormatSymbols.java * java/text/MessageFormat.java * java/text/NumberFormat.java * java/text/RuleBasedCollator.java * java/text/SimpleDateFormat.java * java/util/BitSet.java * java/util/Calendar.java * java/util/Collections.java * java/util/IdentityHashMap.java * java/util/Locale.java * java/util/TreeMap.java * java/util/logging/LogRecord.java * java/util/logging/XMLFormatter.java * java/util/prefs/AbstractPreferences.java * java/util/prefs/Preferences.java * javax/crypto/interfaces/DHPrivateKey.java * javax/crypto/interfaces/DHPublicKey.java * javax/crypto/interfaces/PBEKey.java * javax/net/ssl/HandshakeCompletedEvent.java * javax/security/auth/Subject.java From-SVN: r89454
This commit is contained in:
parent
471854f82a
commit
0460023ee2
39 changed files with 201 additions and 180 deletions
|
@ -298,7 +298,7 @@ public final class Math
|
|||
* @param a the angle (in radians)
|
||||
* @return sin(a)
|
||||
*/
|
||||
public native static double sin(double a);
|
||||
public static native double sin(double a);
|
||||
|
||||
/**
|
||||
* The trigonometric function <em>cos</em>. The cosine of NaN or infinity is
|
||||
|
@ -307,7 +307,7 @@ public final class Math
|
|||
* @param a the angle (in radians)
|
||||
* @return cos(a)
|
||||
*/
|
||||
public native static double cos(double a);
|
||||
public static native double cos(double a);
|
||||
|
||||
/**
|
||||
* The trigonometric function <em>tan</em>. The tangent of NaN or infinity
|
||||
|
@ -317,7 +317,7 @@ public final class Math
|
|||
* @param a the angle (in radians)
|
||||
* @return tan(a)
|
||||
*/
|
||||
public native static double tan(double a);
|
||||
public static native double tan(double a);
|
||||
|
||||
/**
|
||||
* The trigonometric function <em>arcsin</em>. The range of angles returned
|
||||
|
@ -328,7 +328,7 @@ public final class Math
|
|||
* @param a the sin to turn back into an angle
|
||||
* @return arcsin(a)
|
||||
*/
|
||||
public native static double asin(double a);
|
||||
public static native double asin(double a);
|
||||
|
||||
/**
|
||||
* The trigonometric function <em>arccos</em>. The range of angles returned
|
||||
|
@ -339,7 +339,7 @@ public final class Math
|
|||
* @param a the cos to turn back into an angle
|
||||
* @return arccos(a)
|
||||
*/
|
||||
public native static double acos(double a);
|
||||
public static native double acos(double a);
|
||||
|
||||
/**
|
||||
* The trigonometric function <em>arcsin</em>. The range of angles returned
|
||||
|
@ -351,7 +351,7 @@ public final class Math
|
|||
* @return arcsin(a)
|
||||
* @see #atan2(double, double)
|
||||
*/
|
||||
public native static double atan(double a);
|
||||
public static native double atan(double a);
|
||||
|
||||
/**
|
||||
* A special version of the trigonometric function <em>arctan</em>, for
|
||||
|
@ -400,7 +400,7 @@ public final class Math
|
|||
* @return <em>theta</em> in the conversion of (x, y) to (r, theta)
|
||||
* @see #atan(double)
|
||||
*/
|
||||
public native static double atan2(double y, double x);
|
||||
public static native double atan2(double y, double x);
|
||||
|
||||
/**
|
||||
* Take <em>e</em><sup>a</sup>. The opposite of <code>log()</code>. If the
|
||||
|
@ -414,7 +414,7 @@ public final class Math
|
|||
* @see #log(double)
|
||||
* @see #pow(double, double)
|
||||
*/
|
||||
public native static double exp(double a);
|
||||
public static native double exp(double a);
|
||||
|
||||
/**
|
||||
* Take ln(a) (the natural log). The opposite of <code>exp()</code>. If the
|
||||
|
@ -430,7 +430,7 @@ public final class Math
|
|||
* @return the natural log of <code>a</code>
|
||||
* @see #exp(double)
|
||||
*/
|
||||
public native static double log(double a);
|
||||
public static native double log(double a);
|
||||
|
||||
/**
|
||||
* Take a square root. If the argument is NaN or negative, the result is
|
||||
|
@ -444,7 +444,7 @@ public final class Math
|
|||
* @return the square root of the argument
|
||||
* @see #pow(double, double)
|
||||
*/
|
||||
public native static double sqrt(double a);
|
||||
public static native double sqrt(double a);
|
||||
|
||||
/**
|
||||
* Raise a number to a power. Special cases:<ul>
|
||||
|
@ -514,7 +514,7 @@ public final class Math
|
|||
* @param b the power to raise it to
|
||||
* @return a<sup>b</sup>
|
||||
*/
|
||||
public native static double pow(double a, double b);
|
||||
public static native double pow(double a, double b);
|
||||
|
||||
/**
|
||||
* Get the IEEE 754 floating point remainder on two numbers. This is the
|
||||
|
@ -530,7 +530,7 @@ public final class Math
|
|||
* @return the IEEE 754-defined floating point remainder of x/y
|
||||
* @see #rint(double)
|
||||
*/
|
||||
public native static double IEEEremainder(double x, double y);
|
||||
public static native double IEEEremainder(double x, double y);
|
||||
|
||||
/**
|
||||
* Take the nearest integer that is that is greater than or equal to the
|
||||
|
@ -541,7 +541,7 @@ public final class Math
|
|||
* @param a the value to act upon
|
||||
* @return the nearest integer >= <code>a</code>
|
||||
*/
|
||||
public native static double ceil(double a);
|
||||
public static native double ceil(double a);
|
||||
|
||||
/**
|
||||
* Take the nearest integer that is that is less than or equal to the
|
||||
|
@ -551,7 +551,7 @@ public final class Math
|
|||
* @param a the value to act upon
|
||||
* @return the nearest integer <= <code>a</code>
|
||||
*/
|
||||
public native static double floor(double a);
|
||||
public static native double floor(double a);
|
||||
|
||||
/**
|
||||
* Take the nearest integer to the argument. If it is exactly between
|
||||
|
@ -561,7 +561,7 @@ public final class Math
|
|||
* @param a the value to act upon
|
||||
* @return the nearest integer to <code>a</code>
|
||||
*/
|
||||
public native static double rint(double a);
|
||||
public static native double rint(double a);
|
||||
|
||||
/**
|
||||
* Take the nearest integer to the argument. This is equivalent to
|
||||
|
|
Loading…
Add table
Add a link
Reference in a new issue