CHANGES
Java(tm) Development Kit
JDK(tm) 1.1.3
-----------------------------------------------------------------------
CONTENTS
-----------------------------------------------------------------------
This document describes the changes between the various JDK
releases of the Java Platform 1.1. It is divided into several
sections:
- Changes from JDK 1.1.2 to JDK 1.1.3
- Changes from JDK 1.1.1 to JDK 1.1.2
- Changes from JDK 1.1_Final to JDK 1.1.1_Final
- Changes from JDK 1.1beta3 to JDK 1.1_Final
- Changes from JDK 1.1beta2 to JDK 1.1beta3
- Changes from the original JDK 1.1beta to JDK 1.1beta2
Changes within a section are in this order:
- VERSION NUMBER
- BUG FIXES
- TEXT CHANGES
- AWT CHANGES
- BEANS CHANGES
- SECURITY CHANGES
- JAR CHANGES
- LANG CHANGES
- IO CHANGES
- JAVA NATIVE INTERFACE CHANGES
- TOOL CHANGES
---------------------------------------------------------
NOTE: To find the Java version number, execute:
java -version
---------------------------------------------------------
NOTE: Another significant list of changes is our
important bug fixes, available from our JavaSoft
website:
http://java.sun.com/products/jdk/1.1/fixedbugs/
---------------------------------------------------------
NOTE: Known bugs are listed at:
http://java.sun.com/products/jdk/1.1/knownbugs/
---------------------------------------------------------
=======================================================================
Changes from JDK 1.1.2 to JDK 1.1.3
-----------------------------------------------------------------------
JDK 1.1.3 is a bug-fix release to correct a localization problem in
the Solaris JDK 1.1.2. The following list summarizes changes in
JDK 1.1.3.
______________
VERSION NUMBER - The version number for this release is "JDK1.1.3".
________________
LOCALIZATION BUG - In Solaris versions of JDK 1.1.2, the mechanism
for loading the ByteToCharConverter classes for non-8859_1 encodings
was broken. Characters in text areas of windows did not display
properly for non-English-language locales and non-Western European
locales. This problem did not exist in JDK 1.1.1. This bug has been
fixed in JDK 1.1.3. [Bugs 4049223, 4055084]
___________________________
VERIFIER IMPLEMENTATION BUG - A bug in the JDK 1.1.2 verifier would
allow a type-unsafe applet to search and locate strings stored in the
browser's address space. This bug is not known to be present in
JDK 1.0.x. This bug has been fixed in JDK 1.1.3. [Bug 4059597]
______________
JAVA BEANS BUG - The method java.beans.BeanInfo.getMethodDescriptors()
did not return MethodDescriptors for all externally visible methods
in some situations. getMethodDescriptors() did not report methods
having no arguments when those methods were defined later in a class
than similarly named methods having arguments. This problem did not
exist in JDK 1.1.1. This bug has been fixed in JDK 1.1.3. [Bug 4056837]
_______________
AWT REPAINT BUG - Synchronization problems with graphics in the
Win32 version of JDK 1.1.2 would cause graphics deadlocks or
crashes of the Java virtual machine in some situations. This problem
did not exist in JDK 1.1.1. This bug has been fixed in JDK 1.1.3.
[Bugs 4049421, 4051303]
________________
java_g -prof BUG - The profiling option of the Java interpreter
did not work in the Win32 version of JDK 1.1.2. Using a command
such as C:\> java_g -prof MyClass would cause the Java
virtual machine to crash. This problem did not exist in JDK 1.1.1.
This bug has been fixed in JDK 1.1.3. [Bug 4056944]
_______________________
AWT FONTMETRICS SPEEDUP - Improvements have been made to AWT font
metrics in JDK 1.1.3, resulting in significant speedups for programs
such as HotJava that rely heavily on text rendering.
[Bug 4046795]
_________________
AWT COMPONENT BUG - In the Windows version of JDK 1.1.2, a race
condition sometimes caused the HotJava browser to crash when disposing
components that were in the process of being shown. This bug has
been fixed in JDK 1.1.3. [Bug 4051487]
________________
AWT GRAPHICS BUG - In the Windows version of JDK 1.1.2, multi-
threaded access to Graphics objects sometimes resulted in dangling
pointers due to a race condition in awt_Graphics.cpp. This bug has
been fixed in JDK 1.1.3. [Bug 4049421]
________________
INPUT STREAM BUG - An InputStreamReader using the UTF8 encoding
produced an erroneous character stream with missing characters.
This bug has been fixed in JDK 1.1.3 [Bug 4059684]
=======================================================================
Changes from JDK 1.1.1 to JDK 1.1.2
-----------------------------------------------------------------------
JDK 1.1.2 is a bug-fix release. The following is a list of important
bug fixes and other changes in JDK 1.1.2.
______________
VERSION NUMBER - The version number for this release is "JDK1.1.2".
____________
VERIFIER BUG - The JDK 1.1.1 bytecode verifier did not check to
determine whether the number of arguments passed into a method is less
than the amount of memory allocated to local variables for that
method. There was no known security attack based on this bug. This bug
has been corrected in JDK 1.1.2.
___________
SIGNING BUG - In JDK 1.1.1, digitally signed code could be
manipulated to impersonate another digital signature from the list of
signers that are recorded in the Java runtime. This bug has been
fixed in JDK 1.1.2. [Bug 4048143]
_______
DSA BUG - In JDK 1.1.1, the same default seed was used for every
invocation of DSA. This bug has been fixed in JDK 1.1.2.
[Bug 4050406]
________________
AWT MENU BUG FIX - Trying to add a pull-right cascade menu to a
PopupMenu in JDK 1.1.1 led to an exception. This problem did not
exist in JDK 1.1. This bug has been corrected in JDK 1.1.2.
[Bug 4039089]
_________________________________
AWT LIGHTWEIGHT COMPONENT BUG FIX - Mouse tracking events for
lightweight components were not being delivered in some situations.
This bug has been fixed in JDK 1.1.2. [Bug 4038897]
___________________
AWT REPAINT BUG FIX - Some components were not being repainted
properly in some situations in JDK 1.1.1. This problem did not
exist in JDK 1.1. This bug has been fixed in JDK 1.1.2.
[Bug 4040638]
__________________
AWT DIALOG BUG FIX - Closing nested modal dialogs out of order would
cause parent to remain disabled in some cases. This bug has been fixed
in JDK1.1.2. [Bug 4045610]
_________________
AWT FOCUS BUG FIX - Setting default focus caused applets to hang in
some situations. This problem did not exist in JDK 1.1. This bug
has been fixed in JDK 1.1.2. [Bug 4038896]
____________________
APPLETVIEWER BUG FIX - Trying to save an applet from the Appletviewer
led to an exception. This bug has been fixed in JDK 1.1.2.
[Bug 4036537]
________________
DEBUGGER BUG FIX - The debugger "dump" command converted double
values to float values before displaying them. This bug has been fixed
in JDK 1.1.2. [Bug 4046775]
_______________
JAVADOC BUG FIX - The javadoc html output for interfaces repeated
the word "interface" in the signature. This bug has been fixed in
JDK 1.1.2. [Bug 4041579]
__________________________
JNI INVOCATION API CHANGES - To better support the JRE, the
Invocation API is extended in JDK 1.1.2 in a few minor ways.
The changes do not break any existing code. The JNI Native
Method Interface has not been changed.
1. The reserved0 field in the JDK1_1InitArgs structure has been
renamed to "version." The JDK1_1InitArgs structure holds the
initialization arguments to JNI_CreateJavaVM. Callers of
JNI_CreateJavaVM must set the version field to 0x00010001.
JNI_GetDefaultJavaVMInitArgs has been changed to return a "jint"
indicating whether the requested version is supported.
2. The reserved1 field in the JDK1_1InitArgs structure has been
renamed to "properties." This is a NULL-terminated array of
strings. Each string has the format:
name=value
indicating a system property. (This facility corresponds to the -D
option in the java command line.)
3. In JDK 1.1.1, the thread calling DestroyJavaVM must be the only
user thread in the VM. JDK 1.1.2 has lifted this restriction. If
DestroyJavaVM is called when there is more than one user thread,
the VM waits until the current thread is the only user thread, and
then tries to destroy itself.
_________________________
NEW jre COMMAND-LINE TOOL - The jre command-line tool is similar to
the java command-line tool, but is intended primarily for end users
who do not require the development-related options available with
the java tool. For more information on the jre tool, see
http://java.sun.com/products/jdk/1.1/docs/tooldocs/solaris/jre.html
or http://java.sun.com/products/jdk/1.1/docs/tooldocs/win32/jre.html.
Source code for the jre tool can be found in the Windows JDK 1.1.2
directory tree in the jdk1.1.2\demo\jre\win32 folder. In the
Solaris JDK 1.1.2, jre source files can be found in the
jdk1.1.2/demo/jre/solaris directory.
=======================================================================
Changes from JDK 1.1_Final to JDK 1.1.1
-----------------------------------------------------------------------
The following changes and bug fixes are in the JDK 1.1.1 final release.
This is a bug fix release.
______________
VERSION NUMBER - The version number for this release is "JDK1.1.1".
_________
BUG FIXES - This version contains the bug fixes listed at the
web page mentioned above.
____________
UTIL BUG FIX - Removed getMenu and getMenuBar methods from
util.ResourceBundle class. [Bug 4036076]
This bug fix is mentioned here because it fixes an API bug
in JDK 1.1.
These two "convenience" methods were part of the beta release
and were mistakenly left in the final JDK 1.1 after we
discovered the design weakness they presented and decided to
remove them. It is a bug that they were included in the 1.1
release, and you should not call these methods. We have
removed them in 1.1.1. Being convenience methods, they are not
essential, and alternate coding is simple.
To be more specific, the ResourceBundle class in JDK 1.1
included the two methods:
public final Menu getMenu(String key)
throws MissingResourceException {
return (Menu) getObject(key);
}
public final MenuBar getMenuBar(String key)
throws MissingResourceException {
return (MenuBar) getObject(key);
}
These "convenience" methods saved the user from having to
explicitly cast objects in a ResourceBundle that happened to
be of type Menu or MenuBar, such as:
(Menu)rb.getObject(key)
The side effect was that by returning types Menu and MenuBar, this
class referred to the awt package. ResourceBundle is a relatively
low level piece of the core however while AWT is a relatively high
level piece. This dependency caused a number of problems. Most
notably it did not allow a Java runtime environment to be created
that omitted the AWT package (such as in a server).
Removing this dependency on AWT required removing these APIs
completely from ResourceBundle. This change, of course, breaks
code that calls these methods. The upside is that these two APIs
were of very little marginal value. They simply saved the user
the effort of typing an explicit cast. Furthermore, since these
two methods were new at 1.1, they have not been available for
very long.
_____________________
AWT COMPONENT BUG FIX - Changed the field java.awt.Component.locale
from protected to private.
The locale field is accessible through getLocale() and
setLocale() methods, and all code should be using those.
This change is necessary to support future development of
the lightweight framework.
________________________
AWT DATATRANSFER BUG FIX - Corrected the MIME type for DataFlavor.
[Bug 4037854]
In JDK 1.1, java.awt.datatransfer.DataFlavor used the wrong MIME
type. DataFlavor claimed that the MIME type for a java
serialized object was:
application/x-javaserializedobject
This name is inconsistent with other existing names and common
conventions. The name has been corrected by adding hypens to it,
as follows:
application/x-java-serialized-object
This makes it consistent with:
application/java-vm
application/x-java-vm
application/x-java-archive
____________
FONT CHANGES - Times, Helvetica, and Courier are now mapped to
Latin1 characters.
Times, Helvetica, and Courier fonts are no longer mapped to
non-Latin1 classes of fonts, such as Symbol or ZapfDingabats.
These three fonts are now mapped to Latin1 characters, just like
they were in version 1.0.
If you want to have non-Latin1 characters, you must map fonts
such as Symbol and ZapfDingabats to the Java virtual fonts. These
virtual font classes are Serif, Sans-serif, Monospaced, Dialog,
and DialogInput.
===================================================================
Changes from JDK 1.1beta3 to JDK 1.1_Final
-------------------------------------------------------------------
The following changes and bug fixes are in the JDK 1.1 final release.
______________
VERSION NUMBER - The version number for this release is "JDK1.1_Final".
Throughout the documentation we call it simply "JDK1.1"
____________
TEXT CHANGES - Extensive surface changes to Text package
The following extensive changes in the Text package are the
result of a design review recently carried out to
simplify, rationalize and properly abstract the
internationalization API. We realized this would
make a significant improvement in learning and using
internationalization in Java.
The changes involve mostly surface changes, such as
renaming and moving API, and changing the order of arguments,
to better conform with conventions established with the rest
of Java.
____________
TEXT CHANGES - CharacterIterator
Removed getText method. CharacterIterators are intended to allow character
at a time access to text without exposing how the text is actually stored.
It is expected that some CharacterIterators will operate on text that can
not easily or efficiently be stored as a String. Accessing the text as a
whole should be the responsibility of the text object.
Rename startIndex to getBeginIndex and rename endIndex to
getEndIndex. This is to follow established naming patterns for getters and
setters as well as java.lang.String's convention for identifying ranges in
text.
____________
TEXT CHANGES - ChoiceFormat
Changed the constructor to take an array of Strings rather than an array
of Objects. String were the only type that made sense in this context.
Changed the setChoices method to take an array of Strings rather than an
array of Objects. String were the only type that made sense in this
context.
Added applyPattern and toPattern methods similar to the methods on other
Format classes.
Changed the format methods to take a FieldPosition rather than a
FormatStatus argument. See the change description for FormatStatus.
Changed the parse method to take a ParsePosition rather than a ParseStatus
argument. See the change description for ParseStatus.
Added clone, hashCode and equals methods. These are standard overrides
and do not change the semantics.
____________
TEXT CHANGES - Collation
Renamed class "Collator". It was not clear from the old name whether this
class actually did collation or simply represented a set of collation
rules. The new name makes it clear that the class performs string
comparisons.
Removed the constants LESS, EQUAL and GREATER. Changed the compare methods
to return the int value -1, 0 or 1 instead of LESS, EQUAL or GREATER. This
follows the pattern used by the java.lang.String.compareTo method.
Removed the greater and greaterOrEqual methods. The same information is
provided by the compare method. Having only a compare and and equals
method mimics the comparison API provided by java.lang.String.
Removed the compare(String, int, int, String, int, int) version of the
compare method and the getSortKey(String, int, int) version of the
getSortKey method. A recent change to String.substring makes the substring
operation efficient. Users can use String.substring to specify a substring
rather than requiring methods to provide substring variants. Slimming the
API makes it easier to learn.
Changed the type of public constants from byte to int. Changed the
getStrength and getDecomposition methods to return int instead of byte.
Changed the setStrength and setDecomposition methods to accept int
arguments instead of byte. Although the byte type provides a small amount
of type safety, using ints for constants is more in keeping with the rest
of the JDK. The affected constants are:
byte PRIMARY = 0; ==> int PRIMARY = 0;
byte SECONDARY = 1; ==> int SECONDARY = 1;
byte TERTIARY = 2; ==> int TERTIARY = 2;
byte IDENTICAL = 3; ==> int IDENTICAL = 3;
byte NO_DECOMPOSITION = 0; ==> int NO_DECOMPOSITION = 0;
byte CANONICAL_DECOMPOSITION = 1; ==> int CANONICAL_DECOMPOSITION = 1;
byte FULL_DECOMPOSITION = 2; ==> int FULL_DECOMPOSITION = 2;
Removed the getDisplayName methods. These returned the same result as
the corresponding Locale.getDisplayName method calls. Until we have more
useful data, its better not to have these methods in the API.
Provided a concrete equals method. This allows sub-classes to correctly
implement the equals method.
Renamed getSortKey method to getCollationKey. See the change description
for getSortKey.
Renamed getDefault methods to getInstance. This follows new naming pattern
for factory methods.
____________
TEXT CHANGES - CollatedString
Removed this class. Its purpose was to associate a String with its
CollationKey. CollationKey has been changed retain a reference to the
String it was generated from. This allows us to reduce the size of the
API.
____________
TEXT CHANGES - CollationElementIterator
Removed the public constructor. CollationElementIterators are created by
RuleBasedCollator objects. This was expressed by handing a
RuleBasedCollator object to the constructor. This connection is more
obvious if the CollationElementIterator can only be created by a factory
method on RuleBasedCollator.
Made the values returned by the next method directly comparable. This is
not an API change but a semantic change for the next method.
____________
TEXT CHANGES - DateFormat
Made the constructor protected. DateFormat is an abstract class so it
doesn't need a public constructor.
Removed the getDisplayName methods. These returned the same result as
the corresponding Locale.getDisplayName method calls. Until we have more
useful data, its better not to have these methods in the API.
Added a style called MEDIUM that does what DEFAULT does now. For this
release, MEDIUM and DEFAULT have the same effect. In future, the meaning
of DEFAULT will be determined from the locale data. This is a more
logical organization of the style choices.
Changed the format methods to take a FieldPosition rather than a
FormatStatus argument. See the change description for FormatStatus.
Changed the parse and parseObject methods to take a ParsePosition rather
than a ParseStatus argument. See the change description for ParseStatus.
Renamed the getTimeFormat methods to getTimeInstance. Renamed the
getDateFormat methods to getDateInstance. Renamed the getDateTimeFormat
methods to getDateTimeInstance. This follows new naming pattern
for factory methods.
Added a getInstance() factory method that does the same thing as
getDateTimeFormat(SHORT,SHORT) previously did. This follows the pattern
for factories which provides for a getInstance() that returns a default
value to be always available.
Renamed getValidationMode to isLenient. Renamed setValidationMode to
setLenient. These names are more descriptive.
Change the type of public constants from byte to int. Although the byte
type provides a small amount of type safety, using ints for constants is
more in keeping with the rest of the JDK. Also rename constants to follow
the naming convention uses by JDK. The results are:
byte ERA_FIELD = 0; ==> int ERA_FIELD = 0;
byte YEAR_FIELD = 1; ==> int YEAR_FIELD = 1;
byte MONTH_FIELD = 2; ==> int MONTH_FIELD = 2;
byte DATE_FIELD = 3; ==> int DATE_FIELD = 3;
byte HOUROFDAY1_FIELD = 4; ==> int HOUR_OF_DAY1_FIELD = 4;
byte HOUROFDAY0_FIELD = 5; ==> int HOUR_OF_DAY0_FIELD = 5;
byte MINUTE_FIELD = 6; ==> int MINUTE_FIELD = 6;
byte SECOND_FIELD = 7; ==> int SECOND_FIELD = 7;
byte MILLISECOND_FIELD = 8; ==> int MILLISECOND_FIELD = 8;
byte DAYOFWEEK_FIELD = 9; ==> int DAY_OF_WEEK_FIELD = 9;
byte DAYOFYEAR_FIELD = 10; ==> int DAY_OF_YEAR_FIELD = 10;
byte DAYOFWEEKINMONTH_FIELD = 11; ==> int DAY_OF_WEEK_IN_MONTH_FIELD = 11;
byte WEEKOFYEAR_FIELD = 12; ==> int WEEK_OF_YEAR_FIELD = 12;
byte WEEKOFMONTH_FIELD = 13; ==> int WEEK_OF_MONTH_FIELD = 13;
byte AMPM_FIELD = 14; ==> int AM_PM_FIELD = 14;
byte HOUR1_FIELD = 15; ==> int HOUR1_FIELD = 15;
byte HOUR0_FIELD = 16; ==> int HOUR0_FIELD = 16;
byte TIMEZONE_FIELD = 17; ==> int TIMEZONE_FIELD = 17;
____________
TEXT CHANGES - DateFormatData
Renamed class to DateFormatSymbols. This makes its purpose as a set of
strings and symbols to be used in formatting clearer.
Removed the millisPerHour protected field. This was not an appropriate
field in a collection of formatting symbols.
Renamed getAmpms to getAmPmStrings. Renamed setAmpms to setAmPmStrings.
The new names follow the capitalization standard and are clearer.
Removed methods useLocalizedPattern and setPatternLocalized. This
functionality can be obtained through the toLocalizedPattern and
applyLocalizedPattern methods of SimpleDateFormat. These methods were not
appropriate in a collection of formatting symbols.
____________
TEXT CHANGES - DecimalFormat
Renamed getThousandsInterval and setThousandsInterval methods to
getGroupingSize and setGroupingSize. The new names are clearer.
Renamed getFactor and setFactor methods to getMultiplier and setMultiplier.
The new names are clearer.
Changed the constructor that takes a NumberFormatData to use a
DecimalFormatSymbols object instead. See the change description for
NumberFormatData.
Added a constructor that takes just a String and figures out the correct
DecimalFormatSymbols for the default locale. This fills in the telescope
pattern for constructors.
Renamed getPattern method to toPattern. Also removed the boolean argument
and instead added a toLocalizedPattern method. The new name better
reflects the fact that there is no "pattern" property, but rather a
pattern string is constructed that matches the current state of the
DecimalFormat object. Splitting the method into two allows many users to
ignore the existence of toLocalizedPattern.
Renamed setPattern method to applyPattern. Also removed the boolean
argument and instead added an applyLocalizedPattern method. The new
name better reflects the fact that there is no "pattern" property,
but rather a pattern string is used to set the state of the DecimalFormat
object. Splitting the method into two allows many users to ignore the
existence of applyLocalizedPattern.
Removed the "*" and "_" characters from the pattern syntax. This is not
strictly an api change but it is a semantic change. These characters were
of limited use the way they were implemented.
Changed the format methods to take a FieldPosition rather than a
FormatStatus argument. See the change description for FormatStatus.
Changed the parse method to take a ParsePosition rather than a
ParseStatus argument. See the change description for ParseStatus.
Renamed getNumberFormatData to getDecimalFormatSymbols and changed its
return type to DecimalFormatSymbols. Renamed setNumberFormatData to
setDecimalFormatSymbols and changed its argument type to
DecimalFormatSymbols. See the change description for NumberFormatData.
____________
TEXT CHANGES - Format
Changed the format method to take a FieldPosition rather than a
FormatStatus argument. See the change description for FormatStatus.
Changed the parseObject method to take a ParsePosition rather than a
ParseStatus argument. See the change description for ParseStatus.
____________
TEXT CHANGES - FormatException
Renamed class to ParseException. This better reflects its purpose.
Removed the no argument constructor. This constructor was not useful.
____________
TEXT CHANGES - FormatStatus
Renamed class to FieldPosition. This better reflects its purpose as an
object which describes where sub-fields are located in a formatted string.
Replaced public integer fields with getter methods. Adjusted names to
match pattern for indicating ranges in text. The result:
int alignField ==> int getField()
int alignStart ==> int getBeginIndex()
int alignEnd ==> int getEndIndex()
Replaced no argument constructor with a constructor that sets the field.
A field value must always be set so this prevents invalid objects.
It is also more convenient than doing it in two steps.
____________
TEXT CHANGES - MessageFormat
Redesigned the syntax for message patterns to allow inlining of other
pattern strings, such as date patterns, number patterns, choice patterns
etc. This is not itself an API change but is a big semantic difference.
This change allows complete message formats to be created in fewer steps.
Removed the constructor that took a String and a Format array. This was
not needed with the new inline pattern syntax.
Added a setFormat method which allows a single format to be set at a time
rather than requiring all formats to be specified like the setFormats
method. This is more convenient given the inline pattern syntax.
Removed the format method that took a String, a Format array and an Object
array. It was not necessary with the new inline pattern syntax.
Changed the parse method that takes a single String argument to return an
array of Objects rather than a single Object. This makes more sense given
the fact that a message may have many objects embedded in it.
Changed the format methods to take a FieldPosition rather than a
FormatStatus argument. See the change description for FormatStatus.
Changed the parse method to take a ParsePosition rather than a ParseStatus
argument. See the change description for ParseStatus.
Renamed getPattern method to toPattern. The new name better reflects the
fact that there is no "pattern" property. Rather a pattern string is
constructed that matches the current state of the MessageFormat object.
Renamed setPattern method to applyPattern. The new name better reflects
the fact that there is no "pattern" property. Rather a pattern string is
used to set the state of the MessageFormat object.
Added setLocale and getLocale methods.
____________
TEXT CHANGES - NumberFormat
Renamed getDefaultCurrency to getCurrencyInstance. Renamed
getDefaultPercent to getPercentInstance. Added a getNumberInstance
factory. Rename getDefault to getInstance. This follows the naming
pattern for factories.
Removed getCurrencySymbol and getIntlCurrencySymbol methods. These will be
added to a more appropriate class in a future release.
In the methods called {get|set}{Minimum|Maximum}{Integer|Decimal}Count,
rename the word Count to Digits and the word Decimal to Fraction. These
names are clearer.
Remove the methods isDecimalUsedWithInteger and
setDecimalUsedWithInteger. These methods are now found in DecimalFormat
under the names isDecimalSeparatorAlwaysShown
setDecimalSeparatorAlwaysShown.
Renamed setIntegerOnly to setParseIntegerOnly. Renamed isIntegerOnly to
isParseIntegerOnly.
Renamed {is|set}ThousandsUsed to {is|set}GroupingUsed.
Removed the getDisplayName methods. These returned the same result as
the corresponding Locale.getDisplayName method calls. Until we have more
useful data, its better not to have these methods in the API.
Renamed DECIMAL_FIELD to FRACTION_FIELD. Also changed its type from byte
to int. This name is more general. Use of ints as constants is standard
practice.
Removed NUMERATOR_FIELD, DENOMINATOR_FIELD and EXPONENT_FIELD. This
functionality was not supported in this release.
____________
TEXT CHANGES - NumberFormatData
Renamed class DecimalFormatSymbols. This makes its purpose as a set of
strings and symbols to be used in formatting clearer.
Added a default constructor which constructs an object suitable for the
default locale. Added a constructor which takes a locale. These follow
the pattern set by DateFormatSymbols.
Removed starDigit, spaceDigit and exponential fields. This functionality
is not being supported by DecimalFormat.
Changed public fields to getter/setter methods. Adjusted the names of the
methods for clarity. The results are:
char zeroDigit; ==> getZeroDigit(), setZeroDigit()
char thousandsSign; ==> getGroupingSeparator(), setGroupingSeparator()
char decimalSign; ==> getDecimalSeparator(), setDecimalSeparator()
char millePercent; ==> getPerMill(), setPerMill()
char percent; ==> getPercent(), setPercent()
char digit; ==> getDigit(), setDigit()
char separator; ==> getPatternSeparator(), setPatternSeparator()
java.lang.String infinity; ==> getInfinity(), setInfinity()
java.lang.String nan; ==> getNaN(), setNaN()
char minusSign; ==> getMinusSign(), setMinusSign()
Removed the static final char fields. These were not appropriate for a
collection of locale specific number formatting fields. They were also
of little use outside the implementation of DecimalFormat. The removed
fields are:
static final char patternZeroDigit = 48;
static final char patternThousandsSign =
static final char patternDecimalSign = 4
static final char patternMillePercent =
static final char patternPercent = 37;
static final char patternDigit = 35;
static final char patternStarDigit = 42;
static final char patternSpaceDigit = 95
static final char patternSeparator = 59;
____________
TEXT CHANGES - ParseStatus
Renamed class ParsePosition. This better reflects its purpose as an object
which shows where parsing starts and stops.
Removed no argument constructor. This was redundant and of no real
convenience.
Changed public field startAt to a getter/setter method pair. Also
adjusted names of the methods to suit the naming pattern for indicating
characters in a String. The results:
startAt ==> getIndex(), setIndex()
____________
TEXT CHANGES - SimpleDateFormat
Changed the format method to take a FieldPosition rather than a
FormatStatus argument. See the change description for FormatStatus.
Changed the parse method to take a ParsePosition rather than a ParseStatus
argument. See the change description for ParseStatus.
Changed the constructor which took a DateFormatData object to use a
DateFormatSymbols object instead. See the change description for
DateFormatData.
Renamed getDateFormatData and setDateFormatData to getDateFormatSymbols
and setDateFormatSymbols. See the change description for DateFormatData.
Renamed getPattern method to toPattern. Also removed the boolean argument
and instead added a toLocalizedPattern method. The new name better
reflects the fact that there is no "pattern" property, but rather a
pattern string is constructed that matches the current state of the
DecimalFormat object. Splitting the method into two allows many users
to ignore the existence of toLocalizedPattern.
Renamed setPattern method to applyPattern. Also removed the boolean
argument and instead added an applyLocalizedPattern method. The new
name better reflects the fact that there is no "pattern" property, but
rather a pattern string is used to set the state of the DecimalFormat
object. Splitting the method into two allows many users to ignore the
existence of applyLocalizedPattern.
Added a convenience no argument constructor.
____________
TEXT CHANGES - SortKey
Renamed class CollationKey. This keeps the terminology consistent with
the other collation classes.
Added getSourceString method. This returns the string that this
CollationKey represents. This allows the removal of the CollatedString
class.
Added a toByteArray method that returns a byte[] containing the key.
Changed compareTo method to return int rather than byte. This follows the
pattern set by java.lang.String.compareTo.
____________
TEXT CHANGES - StringCharacterIterator
Added the following convenience constructor:
public StringCharacterIterator(String text, int begin, int end, int pos)
Renamed startIndex method to getBeginIndex. Renamed endIndex method to
getEndIndex. These names are consistent with the pattern for indicating
ranges in strings.
Removed the getText method. See the change description for
CharacterIterator.
____________
TEXT CHANGES - TableCollation
Renamed class RuleBasedCollator. It was not clear from the old name
whether this class actually did collation or simply represented a set
of collation rules. The new name makes it clear that the class performs
string comparisons. It also avoids the impression that this class
collates tables.
Renamed getSortKey method to getCollationKey. See the change description
for SortKey.
Removed the getSortKey method that operated on a substring of a given
String. A recent change to String.substring makes the substring
operation efficient. Users can use String.substring to specify a
substring rather than requiring methods to provide substring variants.
Slimming the API makes it easier to learn.
Removed the compare method that operated on a substring of a given String.
A recent change to String.substring makes the substring operation
efficient. Users can use String.substring to specify a substring rather
than requiring methods to provide substring variants. Slimming the API
makes it easier to learn.
Changed compare to return an int value instead of a byte value. See
the change description for Collation.
____________
TEXT CHANGES - TextBoundary
Renamed class to BreakIterator. This more clearly reflects its purpose.
Renamed method nthFromCurrent to next(int). This is a less awkward name.
Renamed method nextAfter to following. This is a less awkward name.
Changed the return type of getText to CharacterIterator. Since
BreakIterator works in terms of a CharacterIterator it didn't make sense
to return a String.
Renamed getWorkBreak methods to getWordInstance. Renamed getLineBreak
methods to getLineInstance. Renamed getCharacterBreak methods to
getCharacterInstance. Renamed getSentenceBreak methods to
getSentenceInstance. This follows the naming pattern for factory methods.
____________
UTIL CHANGES - Calendar
Changed the type of the public constants from byte to int. Changed the
corresponding methods to accept and return ints instead of bytes. and
getDecomposition methods to return int instead of byte. Although the byte
type provides a small amount of type safety, using ints for constants is
more in keeping with the rest of the JDK. The results are:
byte ERA = 0; ==> int ERA = 0;
byte YEAR = 1; ==> int YEAR = 1;
byte MONTH = 2; ==> int MONTH = 2;
byte WEEKOFYEAR = 3; ==> int WEEK_OF_YEAR = 3;
byte WEEKOFMONTH = 4; ==> int WEEK_OF_MONTH = 4;
byte DATE = 5; ==> int DATE = 5;
byte DAYOFMONTH = 5; ==> int DAY_OF_MONTH = 5;
byte DAYOFYEAR = 6; ==> int DAY_OF_YEAR = 6;
byte DAYOFWEEK = 7; ==> int DAY_OF_WEEK = 7;
byte DAYOFWEEKINMONTH = 8; ==> int DAY_OF_WEEK_IN_MONTH = 8;
byte AMPM = 9; ==> int AM_PM = 9;
byte HOUR = 10; ==> int HOUR = 10;
byte HOUROFDAY = 11; ==> int HOUR_OF_DAY = 11;
byte MINUTE = 12; ==> int MINUTE = 12;
byte SECOND = 13; ==> int SECOND = 13;
byte MILLISECOND = 14; ==> int MILLISECOND = 14;
byte ZONEOFFSET = 15; ==> int ZONE_OFFSET = 15;
byte DSTOFFSET = 16; ==> int DST_OFFSET = 16;
byte FIELDCOUNT = 17; ==> int FIELD_COUNT = 17;
int get(byte); ==> int get(int);
set(byte,int); ==> void set(int,int);
void clear(byte); ==> void clear(int);
void add(byte,int); ==> void add(int,int);
void roll(byte,boolean); ==> void roll(int,boolean);
void setFirstDayOfWeek(byte); ==> void setFirstDayOfWeek(int);
void setMinimalDaysInFirstWeek(byte);
==> void setMinimalDaysInFirstWeek(int);
int getMinimum(byte); ==> int getMinimum(int);
int getMaximum(byte); ==> int getMaximum(int);
int getGreatestMinimum(byte); ==> int getGreatestMinimum(int);
int getLeastMaximum(byte); ==> int getLeastMaximum(int);
Renamed getValidationMode to isLenient. Renamed setValidationMode to
setLenient. These names are more descriptive.
Renamed the getDefault methods to getInstance. This is in keeping with the
pattern for naming factory methods.
____________
UTIL CHANGES - GregorianCalendar
Changed the type of the public constants from byte to int. Changed the
corresponding methods to accept and return ints instead of bytes. and
getDecomposition methods to return int instead of byte. Although the byte
type provides a small amount of type safety, using ints for constants
is more in keeping with the rest of the JDK. The results are:
final byte AD = 0; ==> final int AD = 0;
final byte BC = 1; ==> final int BC = 1;
void add(byte,int); ==> void add(int,int);
void roll(byte,boolean); ==> void roll(int,boolean);
int getMinimum(byte); ==> int getMinimum(int);
int getMaximum(byte); ==> int getMaximum(int);
int getGreatestMinimum(byte);==> int getGreatestMinimum(int);
int getLeastMaximum(byte); ==> int getLeastMaximum(int);
____________
UTIL CHANGES - ResourceBundle
Renamed getResourceBundle methods to getBundle. The name
getResourceBundle was too long and overly redundant. The new name is not
in keeping with the naming pattern for factory methods. This is because
the name "getInstance" is being reserved for a future addition to the
resource bundle classes.
__________
AWT CHANGE - Compiling programs that use old AWT API now produces
deprecation warnings
In the beta versions of 1.1, the deprecated methods in the AWT were
marked with simple "DEPRECATED" strings in the method's javadoc
comment. For the final release we have added the appropriate
"@deprecated" tags to those methods so that you can use the
appropriate compiler option to generate warnings that make it
easier for you to convert your programs when you desire to do so.
The document at the following URL describes how to convert to the
1.1 AWT API. It also links to a document that lists every deprecated
method and its 1.1 substitute.
http://java.sun.com/products/JDK/1.1/docs/guide/awt/HowToUpgrade.html
RATIONALE:
The bulk of the AWT deprecations are a result of migrating the AWT
towards JavaBeans compliance. In particular, the two areas which
required a significant number of deprecations were:
- Properties (location, size, visibility, etc.)
In order for introspection to be able to programmatically extract
properties from AWT components, it was necessary to change the
names of various methods to the JavaBeans getFoo/setFoo pattern.
- New event model
JavaBeans and the AWT have defined a new delegation-based event
model for 1.1 which required significant changes to the event
handling API.
Finally, a small set of deprecated methods were changed to enhance
the learnability and consistency of the toolkit API.
__________
AWT CHANGE - Z-ordering changed back to 1.0.2 order
For beta3, the Z-ordering for children in Container instances was
defined in the documentation to be "back to front". Because this
was contrary to the default Z-ordering which existed in 1.0.2 (and
the beta implementation), for FCS we have corrected the Z-ordering
specification in containers to be "front to back".
__________
AWT CHANGE - Changed names of newly-added APIs:
The following APIs have been renamed ( from old name => to new name ):
AverageScaleFilter => AreaAveragingScaleFilter
createScaledImage(...) => getScaledInstance(...)
SCALE_AVERAGE => SCALE_AREA_AVERAGING
RATIONALE:
The method name was changed to "getScaledInstance(...)" to be
consistent with the overall philosophy within the Java API where
factories and derivative constructors should follow the naming
convention of
get<Flavor>Instance(...arguments...)
The term "Average" was not descriptive of the algorithm applied by
the default smooth scaling filter, so the name of this filter and
its algorithm were changed to "AreaAveraging" to indicate what
quantity was being averaged.
__________
AWT CHANGE - Changed dispatchEvent() method and added new method
dispatchEventImpl()
The codepath in dispatchEvent() really must be executed for event
mechanics to work properly in the AWT (hence we made it
package-private). We moved the implementation to a new
package-private method (dispatchEventImpl()) and then
changed dispatchEvent() to be "public final" and have it
call dispatchEventImpl() internally.
RATIONALE:
We have deprecated the 1.0 postEvent()/deliverEvent() methods,
however the method that replaces those (dispatchEvent()) is
currently package-private. This left no options for folks
porting to the new event model.
__________
AWT CHANGE - Changed capitalization of getId() to getID() in
java.awt.AWTEvent
Changed the method name to getID() and provided a deprecated version
for compatibility with the beta versions.
RATIONALE:
java.awt.AWTEvent.getID() is named in a manner inconsistent with
the equivalent methods in other classes:
java.awt.MediaTracker.getID()
java.util.TimeZone.getID()
sun.rmi.registry.RegistryImpl.getID()
sun.rmi.transport.DGCImpl.getID()
Developers have had trouble because they expected the naming to
be consistent. This fixes bug 4027793.
__________
AWT CHANGE - Changed EventQueue to be non-system-specific
The following changes were made to EventQueue:
1) Moved the static getEventQueue() method out of EventQueue
and instead put it in Toolkit with the name "getSystemEventQueue()".
Left the security check in that method, so that applets
cannot get access to the system queue instance (see also #3).
2) Added a public constructor to the EventQueue class such that
multiple instances of it can be created. This will allow programs
to create and use the queue generically, as well as enable browser
vendors to implement a queue/dispatch-thread-per-applet model.
3) Removed the overriding checkAwtEventQueueAccess() method in the
AppletSecurityManager such that it will always default to
invoking lang.SecurityManager.checkAwtEventQueueAccess()
and will *always* throw a security exception if an applet tries
to get a handle on the system event queue.
4) Modified the EventQueue documentation so that it's clear that
the class is very generic. All references to a "system
EventQueue" have been eliminated.
RATIONALE:
This allows the implementation of one dispatch thread per applet.
It also provides design flexibility for the future.
__________
AWT CHANGE - Removed EventSource class
Removed this class because it has no relation to the new event model.
__________
AWT CHANGE - Changed API for delivery of window-open and
window-closed events
Moved the window event API (addWindowListener,removeWindowListener,
etc.) from Dialog and Frame to Window (their superclass) so
that the code is there in one place.
RATIONALE:
Centralizing the code eliminates duplication and makes it easier
to maintain.
__________
AWT CHANGE - Changed return type from Transferable.getTransferData
Transferable.getTransferData returns class BufferedInputStream.
__________
AWT CHANGE - MenuShortcut class no longer extends java.awt.Event.
__________
AWT CHANGE - Changed button modifier mask values in InputEvent:
BUTTON2_MASK is now equivalent to ALT_MASK
BUTTON3_MASK is now equivalent to META_MASK
RATIONALE:
A couple of mask constants were incorrectly defined in the
InputEvent class. These event masks are new in JDK1.1.
This change will ensure that programs that depend on seeing the
BUTTON1_MASK and BUTTON2_MASK will continue to work properly on
platforms that have 1 and 2 button mice (because Alt/Meta can be
used to generate these masks on those systems).
This fixes bug 4029159: Button2/Button3 masks should have same
value as Alt/Meta in InputEvent
__________
AWT CHANGE - Change to serialization of AWT components
Made a systematic change to the way JDK1.1 AWT does serialization.
RATIONALE:
This change corrects bug #4027305, and provides more carefully
thought-out and rational behaviour for AWT serialization in general.
These changes enable customers to use serialization to store
Java Beans (in general) and to enable AWT implementations
to correctly interoperate between vendors.
Implications for developers:
- Saving graphs of AWT objects will now work, even if an object
has more than one listener.
- If an AWT object has listeners that are marked serializable, they
will be saved and restored automatically, such as beans
interconnected with and saved to a file with the BeanBox.
Note that the BeanBox always code-generates serializable
listener implementations.
- If an AWT object has listeners that aren't marked serializable
they will be dropped at writeObject() time.
- Developers will need, as always, to consider the implications of
making an object serializable. One idiom to watch out for is
this:
import java.awt.*;
import java.awt.event.*;
import java.io.Serializable;
class MyApp implements ActionListener, Serializable
{
BigObjectThatShouldNotBeSeralizedWithAButton bigOne;
Button aButton = new Button();
MyApp()
{
// Oops, now aButton has a listener with a
// reference to bigOne!
aButton.addActionListener(this);
}
public void actionPerformed(ActionEvent e)
{
System.out.println("Hello There");
}
}
In this example, serializing aButton by itself will cause MyApp and
everything it refers to to be serialized as well. The problem
is that the listener is serializable by coincidence, not by design.
To separate the decisions about MyApp and the ActionListener being
serializable one can use an inner (nested) class, for example:
import java.awt.*;
import java.awt.event.*;
import java.io.Serializable;
class MyApp java.io.Serializable
{
BigObjectThatShouldNotBeSeralizedWithAButton bigOne;
Button aButton = new Button();
class MyActionListener implements ActionListener
{
public void actionPerformed(ActionEvent e)
{
System.out.println("Hello There");
}
}
MyApp()
{
aButton.addActionListener(new MyActionListener());
}
}
__________
AWT CHANGE - Added setKeyChar() method to KeyEvent
KeyEvent had a setKeyCode() already. This new method was needed
to copy any changes a 1.0 style program may have made
to an Event "key". This fix ensures that all chars work
properly in TextFields, like hitting <return> in HotJava's URL field.
__________
AWT CHANGE - Changed PaintEvent class to use a rectangle for damaged area
Replaced PaintEvent's "Graphics" property with a Rectangle
representing a damaged area instead.
RATIONALE:
Although the PaintEvent class is public, its use in AWT
is completely private -- it is used to trigger calls to paint(),
but programs never explicitly see the event object at all.
__________
AWT CHANGE - Defined a new constant in AWTEvent:
RESERVED_ID_MAX
This new constant is the maximum value of any AWT-defined event ID.
Programs that need to create their own event types should use
an ID greater than this max ID, so that the event will be properly
passed through the system.
RATIONALE:
This is part of the fix for a problem that prevented anyone from
being able to create their own event type and send it to a
component.
This fixes bug 4028353: eventEnabled returns false for all but
AWT events; can't send custom event types.
____________
BEANS CHANGE - Improved handling of beans that are applets
Changed Beans.instantiate to check if a newly created or deserialized
bean is an Applet. If so, it is given a minimal AppletStub and
AppletContext and its init method is called.
After developers have added a newly instantiated bean to an AWT
container, they should check if it is an Applet and if so call
Applet.start.
Also, developers need to test "wanna-be" applet/beans against
Beans.instantiate.
____________
BEANS CHANGE - Serializable listener fix for Beans classes
The java.beans package contains two utility classes,
PropertyChangeSupport and VetoableChangeSupport, that manage
a list of event listeners. The change was to make the way
these classes are serialized consistent with the AWT.
RATIONALE
In beta3 these two classes were inconsistent with the AWT,
which meant there was no support for selectively saving and
restoring listeners marked serializable. The changes are
all class private.
_______________
SECURITY CHANGE - Key is now an interface
Made Key an interface and updated relevant clients. Removed all
implementation from Key to make it an interface. Moved its
implementation into sun.security.
RATIONALE:
Key as an interface is more flexible and easier to implement than
a mix of interfaces and classes.
_______________
SECURITY CHANGE - Changed names of factory methods
Changed factory names from <Engine>.get<Engine> to get<Engine>.
RATIONALE:
Factory method names were MessageDigest.getMessageDigest,
Signature.getSignature, etc. A better naming scheme is Signature.get,
MessageDigest.get, etc. This is important when you start casting the
results to some fairly long interface names, for example:
DSAKeyGen keyGen = (DSAKeyGen)KeyGenerator.getKeyGenerator("DSA");
_______________
SECURITY CHANGE - Changed Provider SPI for java.security.Signature
Initialization
The SPI is the Service Provider Interface, which Providers write to.
There currently are two providers: SUN and JSAFE.
a) Changed the arguments to two initialization method in Signature
from (byte[], String) to (Key)
b) Made one method in key public.
c) Added a small interface (no implementation): DSAPrivateKey
Changed java.security.Signature and sun.security.provider.DSA.
_______________
SECURITY CHANGE- API changes to Signature, MessageDigest, and
KeyPairGenerator
Signature
public Signature get(String algorithm) ->
public Signature getInstance(String algorithm)
public Signature get(String algorithm, String provider) ->
public Signature getInstance(String algorithm, String provider)
MessageDigest
public MessageDigest get(String algorithm) ->
public MessageDigest getInstance(String algorithm)
public MessageDigest get(String algorithm, String provider) ->
public MessageDigest getInstance(String algorithm, String provider)
KeyPairGenerator
public KeyPairGenerator get(String algorithm) ->
public KeyPairGenerator getInstance(String algorithm)
public KeyPairGenerator get(String algorithm, String provider) ->
public KeyPairGenerator getInstance(String algorithm, String provider)
_______________
SECURITY CHANGE - Changed key generation API
The change consisted of moving a set of existing methods from
Signature to a new KeyPairGenerator class.
API change:
a) Moved key generation calls from signature to a new
KeyPairGenerator class
b) Added a small DSAKeyPairGenerator interface, which specializes
behavior for DSA key generation.
The change consisted of moving 3 abstract API methods to a new class.
- defined a KeyGenerator factory method (identical to
Signature's).
- added one line to ths sun.security.provider.DSA class.
- added one line to ths sun.security.provider.Sun class.
- updated API clients in JDK (there are two of them).
RATIONALE:
Customer feedback makes a strong case for moving the
Signature class key generation calls into its own class. Key
generation is currently implemented as part of the Signature
class. While it provides a uniform, algorithm-independent
key-generation semantics, it is hard to understand (and therefore to
use). It has confused users both internally and externally.
__________
JAR CHANGE - Can now handle compressed JAR files in the CLASSPATH
java.util.zip.ZipFile was enhanced so that it can now handle
compressed ZIP/JAR file entries. No changes to the API were made.
With this change, compressed JAR/ZIP files can now be handled
by javac.
Additionally, the runtime has also changed so that it can now
handle compressed ZIP/JAR files that are specified in the CLASSPATH.
___________
LANG CHANGE - Removed a constructor and two methods from java.lang.Throwable
- Removed the Throwable(String, Object[]) constructor
- Removed the setArguments() method
- Removed the getArguments() method
RATIONALE:
The arguments property was originally added to java.lang.Throwable for
use in the generation of localized messages. This was deemed premature,
and so the property was removed in order to avoid tempting programmers
to use it for some other purpose.
_________
IO CHANGE - Deprecated constructor StreamTokenizer(InputStream)
This fixed bug 4025998.
RATIONALE:
The StreamTokenizer(InputStream) constructor was not binary-compatible
with JDK 1.0.2. This constructor was changed in JDK1.1 to convert
automatically from bytes to characters using the platform's default
character encoding, but that change required buffering input data,
which breaks binary compatibility. This change has been removed.
____________________________
JAVA NATIVE INTERFACE CHANGE - Changed the names of VM initialization
and thread-attach structures
Changed the JDK initialization structure name from JavaVMInitArgs to
JDK1_1InitArgs to make it clear that it is JDK1.1 specific.
Similarly, ThreadAttachArgs has changed its name to JDK1_1AttachArgs.
RATIONALE:
When a native application uses the invocation API to start up
the VM, it passes the VM an initialization structure. Since in
general different VM implementations require different
initialization information (min/max heap size, native stack size,
enable debugging, etc.), the JNI spec explicitly says
that the contents of the structure vary among VM implementations.
However, the JDK initialization structure, JavaVMInitArgs, gives
the wrong impression that it should work with other VMs. This has
caused confusion on how to implement the JNI without conflicting
with JDK's naming scheme.
____________________________
JAVA NATIVE INTERFACE CHANGE - Changed NewStringUTF parameters
Removed the length argument, strlen(s, from NewStringUTF(). The
signature for this method is now:
jstring NewStringUTF(JNIEnv *env, const char *s);
RATIONALE:
NewStringUTF constructs a Java string from a UTF string. Although
UTF strings are already 0-terminated, the function required an
unnecessary length argument, forcing the programmer to write:
NewStringUTF(env, s, strlen(s))
instead of just:
NewStringUTF(env, s)
____________________________
JAVA NATIVE INTERFACE CHANGE - Changed DeleteLocalRef and DeleteGlobalRef
Now programmers can pass local or global references to DeleteLocalRef
and DeleteGlobalRef directly, instead of having to pass the address
of these references.
RATIONALE:
DeleteLocalRef and DeleteGlobalRef delete local or global references
created by the JNI. They used to take the address of the reference,
forcing programmers to write:
DeleteLocalRef(&ref)
now references are passed as follows:
DeleteLocalRef(ref)
____________________________
JAVA NATIVE INTERFACE CHANGE - Changed the "IsSubclassOf" function name
to "IsAssignableFrom."
RATIONALE:
The following JNI (Java Native Interface) function was misnamed:
jboolean IsSubclassOf(JNIEnv *env, jclass clazz1,
jclass clazz2);
It did not match the Java Language Specification's definition
of the "subclass" relationship.
The Reflection API has a similar method in java.lang.Class:
boolean isAssignableFrom(Class class);
This is a much more suitable name. The JNI function therefore changed
its name to:
jboolean IsAssignableFrom(JNIEnv *env, jclass clazz1,
jclass clazz2);
____________________________
JAVA NATIVE INTERFACE CHANGE - Change RegisterNatives so that it
takes all the information in one array.
RATIONALE:
In beta3, the RegisterNatives function took three arrays
containing native method names, signatures and function entry
pointers, respectively:
jint RegisterNatives(JNIEnv env, jclass clazz,
const char *names[],
const char *signatures[];
void *nativeProcs[]);
These arrays had to be of the same length, and had to be NULL-
terminated. It's better to put all the info in a single array,
and pass the length of the array as another argument:
struct JNINativeMethod {
char* name;
char* signature;
void* fnPtr;
};
typedef struct JNINativeMethod JNINativeMethod;
void RegisterNatives(JNIEnv* env, jclass classID,
const JNINativeMethod methods[], jsize count);
Here's how code that uses this looks in FCS:
JNINativeMethodSpec methods[] = {
{ "sin", "(D)D", &::sin },
{ "cos", "(D)D", &::cos },
{ "tan", "(D)D", &::tan }
};
env->RegisterNatives(env->FindClass("java/lang/Math"), methods, 3);
Here's how the same code used to look in beta3:
const char* names[] = { "sin", "cos", "tan", NULL };
const char* signatures[] = { "(D)D", "(D)D", "(D)D", NULL };
void* procs[] = { &::sin, &::cos, &::tan, NULL };
env->RegisterNatives(env->FindClass("java/lang/Math"),
names, signatures, proc);
The former approach is clearer.
____________________________
JAVA NATIVE INTERFACE CHANGE - Changed the typing scheme of references
to Java objects.
In JDK Beta 3, reference types are defined as follows:
typedef void *jref; /* generic type for references */
typedef jref jobject; /* Java objects */
typedef jref jclass; /* Java class objects */
typedef jref jstring; /* Java strings */
typedef jref jarray; /* Java arrays */
The distinction between jref and jobject was not clear. In 1.1 FCS,
we got rid of jref, and introduced additional reference types to more
accurately convey the subtyping information. We introduced dummy C++
classes to enforce the subtyping relationships:
#ifdef __cplusplus
class _jobject {};
class _jclass : public _jobject {};
class _jthrowable : public _jobject {};
class _jstring : public _jobject {};
class _jarray : public _jobject {};
class _jbooleanArray : public _jarray {};
class _jbyteArray : public _jarray {};
class _jcharArray : public _jarray {};
class _jshortArray : public _jarray {};
class _jintArray : public _jarray {};
class _jlongArray : public _jarray {};
class _jfloatArray : public _jarray {};
class _jdoubleArray : public _jarray {};
class _jobjectArray : public _jarray {};
typedef _jobject *jobject;
typedef _jclass *jclass;
typedef _jthrowable *jthrowable;
typedef _jstring *jstring;
typedef _jarray *jarray;
typedef _jbooleanArray *jbooleanArray;
typedef _jbyteArray *jbyteArray;
typedef _jcharArray *jcharArray;
typedef _jshortArray *jshortArray;
typedef _jintArray *jintArray;
typedef _jlongArray *jlongArray;
typedef _jfloatArray *jfloatArray;
typedef _jdoubleArray *jdoubleArray;
typedef _jobjectArray *jobjectArray;
#else
struct _jobject;
typedef struct _jobject *jobject;
typedef jobject jclass;
typedef jobject jthrowable;
typedef jobject jstring;
typedef jobject jarray;
typedef jarray jbooleanArray;
typedef jarray jbyteArray;
typedef jarray jcharArray;
typedef jarray jshortArray;
typedef jarray jintArray;
typedef jarray jlongArray;
typedef jarray jfloatArray;
typedef jarray jdoubleArray;
typedef jarray jobjectArray;
#endif
JNI function types have been modified accordingly to accept or return
the newly introduced reference types.
____________________________
JAVA NATIVE INTERFACE CHANGE - Changed the typing of certain primitive
types:
The following types in earlier versions of JDK:
typedef char jboolean;
typedef char jbyte;
typedef unsigned long jsize;
have been changed to:
typedef unsigned char jboolean;
typedef signed char jbyte;
typedef jint jsize;
in JDK 1.1 FCS.
Earlier, jsize was defined to be an unsigned integer type that has
the same number of bits as a native pointer. The idea was to use
jsize to represent array length and size, etc. However, there has
been a great deal of confusion on what operations are available
on jsize, whether it is safe to convert jsize to int, etc. We
changed jsize to be the same as jint because Java arrays and
strings cannot be longer than 2^31 anyway. All standard Java
methods that operate on arrays and strings use "int" as the
length and size types.
___________
TOOL CHANGE - New argument -s added to javap
-s: Print method and field type signature information in the internal
form used by the Java Virtual Machine. This is the form through
which a user of the Java Native Interface (JNI) refers to methods
and fields. Using javap with the -s flag to print method and field
information for a class to be accessed with the JNI avoids the
difficulty of constructing signatures in the obscure internal
format manually.
Examples of use:
% cat foo.java
class foo {
Thread thread;
int i;
public static void main(String args[]) {
}
private void foo() {}
}
% javap -p foo
Compiled from foo.java
class foo extends java.lang.Object
/* ACC_SUPER bit set */
{
java.lang.Thread thread;
int i;
public static void main(java.lang.String []);
private void foo();
foo();
}
% javap -p -s foo
Compiled from foo.java
class foo extends java.lang.Object
/* ACC_SUPER bit set */
{
thread Ljava/lang/Thread;
i I
public static main ([Ljava/lang/String;)V
private foo ()V
<init> ()V
}
___________
TOOL CHANGE - Added -deprecation flag to javac
By default, when compiling a class that contains deprecated APIs,
javac now displays only a brief note, rather than immediately
listing all of the deprecated APIs. For example:
% javac MyClass.java
Note: MyClass.java uses deprecated APIs. Recompile with "-deprecation"
for details.
Then, you can re-compile using the -deprecation flag to get the
name of the deprecated constructor, field, method, class or
interface.
% javac -deprecation MyClass.java
MyClass.java:3: Note: The constructor java.lang.String(byte[],int)
has been deprecated.
new String(new byte[0], 0);
^
Note: wombat.java uses deprecated APIs. Please consult the
documentation for a better alternative.
At this point, please refer to the JDK API Reference documentation
(the javadoc-generated web pages) jdk1.1/docs/api/packages.html.
If you look up the deprecated API, it should give you more details
such as which API replaces it.
To summarize the compiler options:
javac -nowarn => complete silence
javac => a one-line comment about deprecations
javac -deprecation => a full report
RATIONALE:
This can greatly reduce the amount of warning by default, allowing
developers to concentrate on the errors, and to have a graceful
migration towards updating their deprecated code.
===================================================================
Changes from JDK 1.1beta2 to JDK 1.1beta3
-------------------------------------------------------------------
The following changes and bug fixes are in the JDK 1.1beta3 release.
______________
VERSION NUMBER - The version number for this release is "JDK1.1beta3.3".
Versions beta3.1 and beta3.2 were internal candidates that were
never released.
You can get the version number of your particular release by executing:
java -version
_______________
PACKAGE CHANGES - Two new packages in java.*: java.beans and java.math
Changes in these packages are described in more detail below.
________________
JAVABEANS CHANGE - JavaBeans1.0 is now part of JDK1.1
The classes that define the JavaBeans 1.0 API specification
are now part of JDK1.1. These classes are part of BDK1.0
beta1 and beta2; there will be a BDK1.0beta3 that synchronizes
with JDK1.1beta3.
- There has been one change in the JavaBeans APIs: the string
argument to Beans.instantiate() has changed. The argument
used to be the path to a serialized bean, it is now the
name of a bean, which may be serialized, or just a class name.
______________
GENERAL CHANGE - Extended the core API methods that take a PrintStream
argument so they can also take a PrintWriter argument.
Several constructors and methods have been added to make the
new PrintWriter class easier to use.
_________
IO CHANGE - Merged the functionality of the ByteToCharConverter
and CharToByteConverter classes into the InputStreamReader and
OutputStreamWriter classes.
The ByteToCharConverter and CharToByteConverter classes
are no longer public, so various constructors and methods
that previously took converter arguments now accept names
of character encodings. The InputStreamReader and
OutputStreamWriter classes no longer commit to buffering
the underlying byte stream. Details on these and other
changes may be found at:
http://java.sun.com/products/JDK/1.1/docs/guide/io/b3-changes.html
___________
LANG CHANGE - New method: identityHashCode in java.lang.System
Its signature is:
public static native int identityHashCode(Object x);
It returns the same hashcode for the given object as
would be returned by the default method hashCode(),
whether or not the given object's class overrides hashCode().
The hashcode for the null reference is zero.
RATIONALE:
In previous releases, Java lacked a constant-time
identity hash function. This meant there was no way to build
a hash table which keyed on object identity, and for which
access would be a constant-time operation.
Java supplied a constant-time identity comparison operation
(the built-in operator ==) but did not supply a corresponding
constant-time identity hash operation.
This new method is very useful when the object in question does
override equals and hashCode, but the guts of some abstraction
is working with canonicalized instances, such as interned
strings, and needs to use them as keys for fast lookups.
Although this is sometimes done by wrapping the string-like
object inside another object which does not override equals
and hashCode, it's desirable to allow such objects to be used
directly as keys in identity tables, to avoid the overhead
of allocating the wrappers.
When performance is not an issue, the method Object.hashCode()
suffices, even if it is slow (as with strings), but when
performance must be maximized, neither the wrapper technique
nor the slow hashCode() method is workable, and only a fast
identityHashCode() method will be suitable.
___________
LANG CHANGE - getResourceAsName(String) changed to getResource(String)
java.lang.Class and java.lang.ClassLoader used to have methods
with signature:
String getResourceAsName(String)
which would return an external representation to a URL to
the desired resource file. These methods have been replaced
by methods with signature
java.net.URL getResource(String)
which return the actual URL (or null if the resource is not found).
___________
AWT CHANGES - New type of event: ContainerEvent
This change provides the hooks for containers to easily register event
listeners on descendents.
- Added new ContainerEvent class and equivalent ContainerListener
interface to enable notification when components are added/removed
from containers. This makes it relatively easy for containers
to register themselves as listeners for events on all of their
descendents.
- Added the addContainerListener/removeContainerListener methods
to Container.
RATIONALE:
In beta2 the event model provided no reasonable mechanism
to allow containers to get input events which occurred in their
descendents, which turns out to be a fairly common need.
The 1.0 event model allowed this by automatically propagating
events up the containment hierarchy until some component returned
"true" to absorb the event and stop the propagation. While this
seems powerful (and it's easy), the 1.0 propagation model is error
prone and results in poor performance.
___________
AWT CHANGES - Further Fine-Tuning of AWT Input Event Related API
Responding to customer feedback and testing, we made the following
changes to the API in the java.awt and java.awt.event packages:
- Added a TextEvent class and TextListener interface to enable
programs to track all changes to a text component, including
paste and programmatic modifications. (Bug #4014945)
- Added new windowActivated/windowDeactivated event types to WindowEvent
to enable programs to determine when a window gets/loses focus.
- Added isTemporary method to FocusEvent to enable determining the
difference between when focus is explicitly moved between components
and when focus temporarily changes (such as when a window is
de-activated).
- Added a consume method to the InputEvent class, enabling any object
to consume input events. Previously, only the Component associated
with the input event could consume it.
- Made the AWTEventMulticaster class public, enabling component
subclasses to serve as multicast sources for AWT-defined events.
- Made the java.awt.event Adapter classes abstract so that it's clear
that they need to be extended.
- Removed constructors in java.awt.event Event classes that took
an old 1.0 event as a parameter.
- Changed the event id values of the ComponentEvent types to prevent them
from clashing with the WindowEvent ids.
- Added virtual keycode identifiers to KeyEvent to properly represent
all standard keys on the keyboard. The keyCode property of the key
event now contains one of those identifiers (instead of the ASCII
integer equivalent).
- Added a method called isPopupTrigger to the MouseEvent class,
providing a platform-independent way to determine whether a mouse
event should result in a popup menu being shown. (Bug #4017794)
__________
AWT CHANGE - New method: getTreeLock() to java.awt.Component:
The public field Component.LOCK will NOT be changed for beta3,
to minimize the risk of the change for beta3, but it will be made
non-public (package private) between beta3 and final release.
RATIONALE:
The AWT exposes a public static final field, called Component.LOCK,
for synchronizing operations that affect or depend on
component-tree structure.
Exposing this field for public use unnecessarily constricts future
AWT implementations to providing a single toolkit-wide lock.
It is much better to provide access to the lock object by an
instance method on class Component, such that future
implementations can return a context-sensitive locking object if
they need to.
__________
AWT CHANGE - New Component method: getLocationOnScreen
Added one new method to the Component class:
Point getLocationOnScreen()
RATIONALE:
This method returns the current location of the component in
the screen's coordinate space.
__________
AWT CHANGE - New Choice method: insert
Added one new method to the Choice class:
void insert(String item, int index)
RATIONALE:
This method lets you insert items into a Choice control at the index
you specify. (In Beta2, items were always added at the end.)
__________
AWT CHANGE - ItemSelectable interface change
Removed the following two methods from the ItemSelectable interface:
public int[] getSelectedIndexes();
public String[] getSelectedItems();
They have been replaced by the following method:
public Object[] getSelectedObjects();
RATIONALE:
This change removes the restriction that an item must be representable
by a String. The AWT classes that implement ItemSelectable (List,
Choice, Checkbox, and CheckboxMenuItem) have changed accordingly.
__________
AWT CHANGE - Allow Container and Component classes to be extended
directly to create lightweight components.
1) Added the following method to java.awt.Component
Dimension getMaximumSize();
In beta2 components could specify only their minimum and preferred
sizes and assume that the component was infinitely stretchable.
This assumption is invalid with lightweight components where components
tend to be written at a finer granularity and some components don't
want to be flexible.
The default implementation of this method specifies infinite
size to reflect the current behavior.
2) Added two alignment functions to java.awt.Component.
float getAlignmentX();
float getAlignmentY();
that return a number between 0.0 and 1.0. These values specify how
to align the component relative to other components. 0.0 would be
aligned at the origin, 1.0 the furthest away from the origin,
0.5 is centered, etc. As examples, a text component would have a
y alignment to the baseline, and an icon component would have
an alignment to its hotspot, if defined.
RATIONALE:
Before beta3 the only way of creating new components in the AWT was
to extend Canvas, Panel, and their subclasses. Component and
Container could not be directly extended. All components were
automatically "heavyweight" in that each had an opaque native
window associated with them. The problem with this is that heavyweight
components are expensive (you can't afford to have too many) and
they cannot have any transparent pixels. The new changes create
standard support for lightweight components.
____________
MATH CHANGES - New package: java.math, which initially contains two
classes: BigInteger and BigDecimal.
BigNum class has been removed and replaced by BigInteger and BigDecimal.
BigIntegers are immutable arbitrary-precision integers, which provide
analogs to all of Java's primitive integer operators, and all relevant
static methods from java.lang.Math. Additionally, BigIntegers provide
operations for modular arithmetic, GCD calculation, primality testing,
prime generation, single-bit manipulation, and a few other odds and ends.
BigDecimals are immutable, arbitrary-precision signed decimal numbers,
suitable for monetary calculations. BigDecimals provide operations for
basic arithmetic, scale manipulation, comparison, format conversion
and hashing. The BigDecimal class gives its user complete control
over rounding behavior, forcing the user to explicitly specify a
rounding behavior for operations capable of discarding precision.
________________
SECURITY CHANGES - New method: getProviders(), which provides
a way to query which crypto providers are installed.
This method is added to the Security class and has the form:
public static String[] getProviders();
___________
TOOL CHANGE - javakey is now able to export keys and certs
to files.
This is a change to the javakey tool, a reciprocal to
the existing import facility.
___________
TOOL CHANGE - javac -nowarn flag now suppresses @deprecated messages
This can greatly reduce the amount of messages, allowing you
to see only the errors.
____________
TOOL CHANGES - jar tool is incompatible with earlier JAR files
The beta3 jar tool (more specifically, ZipInputStream and
ZipOutputStream) is incompatible with JAR files created with
the beta2 jar tool. The JAR tool won't understand those files.
There was a bug in beta2 ZipInputStream and ZipOutputStream
such that those classes didn't generate files properly according
to the zip specification (one bit out of place means a lot...)
The bug was fixed and now it generates proper zip files that
other tools will understand.
____________
TOOL CHANGES - Removed ability to generate MIF from javadoc
Removed "-doctype MIF" option from javadoc. This was necessary
in order to remove the HTML parser dependency that HotJava had
on the JDK.
===================================================================
Changes from the original JDK 1.1beta to JDK 1.1beta2
-------------------------------------------------------------------
The following fixes and improvements are in the JDK 1.1beta2 release.
______________
VERSION NUMBER - The version number for this release is "JDK1.1beta2".
You can get the version number of your particular release by executing:
java -version
_______
CHANGE - Source code for public classes included
The source code for the public classes in the java.* package are
now included in this release in the file src.zip. These correspond
to the public classes contained in classes.zip. This src.zip
is equivalent to that shipped previously with JDK 1.0.2
but also includes sources to the new public classes added in JDK 1.1.
_______
BUG FIX - Jar tool manifest-build limitation on number of files on Solaris
DESCRIPTION OF BUG
In some configurations of Solaris, trying to create a JAR file
that has more than 64 entries in it would fail. The command
that would cause this to happen is:
% jar cvf test.jar <somedir>
where <somedir> is a directory containing more than 64 files.
SOLUTION - This bug has been fixed.
_______
BUG FIX - Javakey tool signs jar files with invalid signatures
DESCRIPTION OF BUG
An implementation inconsistency in the javakey security tool causes
jar files to be signed with invalid signatures. This means that
signature checking will always fail, thus all applets will run as
untrusted, with minimal permissions enabled. This makes code signing
feature unusable but it is not a security hole.
SOLUTION - This bug has been fixed.
_______
BUG FIX - 4017054 - Limit to numeric range of case statements
DESCRIPTION OF BUG
If a Java switch statement includes case statements that cover a wide
range of numerical values (although not necessarily a large number of
cases themselves), the Java compiler can run out of memory attempting
to compile that statement. For instance, a switch statement
containing the two cases 0 and 99999999 will cause the compiler to
run out of memory.
SOLUTION - This bug has been fixed.
_______
BUG FIX - 4018832 - Missing class java.io.LineNumberReader
DESCRIPTION OF BUG
In the Win32 release, the class java.io.LineNumberReader is
documented but the class file is absent from classes.zip
SOLUTION - java.io.LineNumberReader has been added to classes.zip
_______
BUG FIX - 4018252 - DecimalNumberFormat methods throw exceptions
DESCRIPTION OF BUG
The class java.text.DecimalNumberFormat throws an exception when any
of its format or parse methods are called. A typical stack trace
looks something like the following:
java.lang.NumberFormatException: 451.0D
at java.lang.Double.<init>(Double.java)
at java.text.DigitList.getRealDouble(DigitList.java)
at java.text.DigitList.set(DigitList.java)
at java.text.DecimalFormat.format(DecimalFormat.java)
at java.text.NumberFormat.format(NumberFormat.java)
Although DecimalNumberFormats are generally not used directly, they
are returned by calls to NumberFormat.getDefault(),
NumberFormat.getDefaultCurrency() and NumberFormat.getDefaultPercent().
SOLUTION - This bug has been fixed.