CHANGES

Improvements and API changes in trunk

1. Modified SourceLocator so that it will automatically load a SootClass
   from Soot's own JAR file if the class cannot be found on the
   soot-classpath, but ONLY if the class is in package soot.rtlib.
   See package documentation for soot.rtlib for more details.
2. The option -process-dir how has an alias -process-path, which makes
   sense because Soot can also process JAR files. Also, -keep-line-numbers
   now defaults to true. This is because so many people ask about how to
   access line numbers.   
3. Removed "org.apache" package from the default excludes list because
   that, by default, leads to Soot not analyzing anything of tomcat, for
   instance. This change follows the principle of least surprise.
4. Added option no-bodies-for-excluded that allows soot This
   option causes Soot to not load any method bodies of classes from the
   "exclude" packages (see -exclude option), even in whole-program mode,
   unless the class is explicitly as a "basic class".
   -no-bodies-for-excluded implies -allow-phantom-refs, as it uses the
   phantom-refs mechanism to model classes that are not loaded.
   The intent of this flag is to allow whole-program analyses to execute
   quickly on little toy examples, disregarding the Java runtime. Of
   course, such an analysis may be unsound - you get what you are asking
   for.
5. Added new packs "wjpp" and "wspp" for adding pre-processors. 
6. Thanks to Matthias Perner, added support for handling invokedynamic
   instructions. The instructions are handled as follows:
   - On the Jimple, Grimp and Baf level, invokedynamic instructions
     appear as DynamicInvokeExpr with a static SootMethodRef that
     has class java.dyn.InvokeDynamic as target class. This is in
     accordance with the internal handling in JDK 7.
   - SootMethodRefs which have java.dyn.InvokeDynamic as target class
     may not be resolved: calling resolve() will throw an exception
   - On the Jasmin level, invokedynamic instruction appear in the
   	 form "invokedynamic <methodName>(<methodParams>)<methodRetType>".
   	 In particular, there is no target class given.
   - Jasmin translates this into an invokedynamic instruction with
     two arguments:
     a) a NameAndType attribute referring to
        <methodName>(<methodParams>)<methodRetType> and
     b) the constant "0" (two zero bytes, i.e., a short);
        according to the VM spec this is reserved for future use
7. ReflectiveCallsInliner now supports field set/get through reflection.        
8. Integrated a novel context-sensitive points-to analysis that uses
   a memory-efficient geometric encoding. The analysis was presented
   in the following paper:
   Xiao Xiao and Charles Zhang. Geometric Encoding: Forging high
   performance context sensitive points-to analysis for Java.
   In ISSTA 2011: International Symposium on Software Testing and
   Analysis, Toronto, Canada, 2011
   The code was contributed to Soot by Richard (Xiao) Xiao. Thanks for
   this! You can find the code in soot.jimple.spark.geom. There are
   new phase options in cg.spark to enable and configure this
   analysis.   
9. Improved support for custom entry points in the points-to analysis.
10.Added option allowing putiry analysis to add a "Pure" bytecode
   attribute for pure methods.
        
Improvements and API changes in version 2.4.0

1. Deprecated class StronglyConnectedComponents because it is
   inefficient. There is a new class StronglyConnectedComponentsFast
   that implements Tarjan's algorithm and is very efficient.
   
2. Changed type parameter of PatchingChain to be a subtype of Unit
   because this is what the implementation actually requires.

3. Added Hossein Sadat-Mohtasham's implementation of program dependency
   graphs.
   
4. Added new phase option "reflection-log" to support TamiFlex
   reflection log files. This option allows Spark to construct sound
   call graphs and points-to sets even for programs that use reflection.
   More information is available at:
   http://code.google.com/p/tamiflex/wiki/DaCapoAndSoot

Improvements and API changes in version 2.3.0

1. Integrated new JastAdd-based frontend with support for Java 5
   source code. Thanks to Torbjorn Ekman for contributing his
   frontend and his implementation of Java2Jimple!
   If you like to use the old polyglot-based frontend then
   just give the -polyglot command line option (Java 1.4 only).
   
2. At the very least if -validate is on, Soot now checks that @this
   is only assigned in the first statement of an instance method.
   This has always been an implicit assumption for Jimple code
   but up until now was not enforced. Code transformations written
   by clients should make sure to restore this invariant before
   transforming Jimple into another intermediate representation
   or into bytecode. (Eric Bodden)  
   
3. Integrated full support for preserving annotations in Jimple
   and writing them to bytecode. SOURCE level annotations are
   preserved if code is read from source and the flag
   -p jb preserve-source-annotations is given. (does not hold for
   package and local variable annotations) SOURCE level annotations
   are not written into bytecode by Soot.
   Annotations can easily be added to SootClass, SootMethod and
   SootField instances using the utility class
   soot.jimple.toolkits.annotation.j5anno.AnnotationGenerator.
   (Thanks to Will Benton!)   
 
4. Java 5 Annotations for fields, classes, parameters and methods
   are now fully preserved in bytecode. Jasmin was changed to
   implement this support. (Eric Bodden)   
   
5. Ben Bellamy contributed a new implementation for the type assigner.
   This is to be published under the title 'Efficient Local Type
   Inference' at OOPSLA '08. The new type assigner is usually
   faster (it is applied when processing bytecode), and sometimes
   more precise. In particular it should always infer the most
   narrow type possible.   
   Ben has tested this code, and verified that it provides a typing
   that is at least as tight as the original algorithm (tighter in
   2914 methods out of 295598) on a number of benchmarks. These are:
   abc-complete.jar, BlueJ, CSO (Scala code), Gant, Groovy, havoc.jar,
   Java 3D, jEdit, Java Grande Forum, Jigsaw, Jython, Kawa, rt.jar,
   Kawa, Scala and tools.jar.
   The mean execution time improvement is around 10 times,
   but for the longest methods (abc parser methods and havoc with
   >9000 statements) the improvement is between 200 and 500 times.
   There are new phase options in jb.tr that allow to switch
   back to the old type assigner or to compare both implementations.
   The command line option -use-old-type-assigner was removed.
   It used to enabled an even older version of the type assigner
   (based on bit vectors) that was commented out any way.
   The appropriate code was flagged as deprecated and may be removed
   in future versions of Soot.   

Improvements and API changes in version 2.2.5 

1. Soot now has an option to prepend the classpath given on the
   command line to the default classpath.  See the documentation on 
   the option -pp for details.

2. Included Manu Sridharan's demand-driven points-to analysis.  See
   http://doi.acm.org/10.1145/1133255.1134027 for details.  It can be
   activated via: -p cg.spark cs-demand:true, and also has two other
   phase options to control the effort with which the analysis
   tries to build context information.

3. Added local must-alias and not-may alias analyses. See package
   soot.jimple.toolkits.pointer.

4. Updated NullnessAnalysis for better performance. Thanks to Julian
   Tibble!

5. The flow analysis framework is now generic. One can explicitly
   declare the type of the abstraction and that is used for the fixed
   point iteration and the node type for the graph that is iterated over.
   This avoids a lot of casting one had to do in the past.

6. Between versions 2.2.3 and 2.2.4, a bug was introduced to Spark:
   PointsToSetInternal now implements equals/hashCode which confused the
   propagator algorithm because it was relying on comparison by identity.
   This lead to a five-fold slowdown on the points-to analyses and
   potentially wrong results.  The bug has now been fixed by modifying
   the propagator algorithm. Original speed was restored.

7. Richard Halpert's lock allocator as published at PACT'07, see:
   http://www.sable.mcgill.ca/publications/papers/#pact2007
   It is accessible via the wjtn (whole-jimple transaction) pack.
   
8. Added option: empties-as-allocs (default:true)
   When this option is set to true, Spark treats references to EMPTY_SET,
   EMPTY_MAP, and EMPTY_LIST as allocation sites for HashSet, HashMap and
   LinkedList objects, respectively, and references to
   Hashtable.emptyIterator as allocation sites for
   Hashtable$EmptyIterator. This enables subsequent analyses to
   differentiate different uses of Java's immutable empty collections.
        
9. Added instance keys (see class InstanceKey). An instance key is a
   static abstraction of one or multiple runtime objects. It's identity
   is defined via an intraprocedural flow-sensitive must-alias analysis.
   Further an instance key can be tested for must-not-aliasing with
   other keys. This functionality is provided via both an intraprocedural
   flow-sensitive must-not-alias analysis and the normal interprocedural
   flow-insensitive points-to analysis (whichever present). See the
   Technical Report SABLE-TR-2007-8 for details, at:
   http://www.sable.mcgill.ca/publications/techreports/#report2007-8

10. Building against polyglot version 1.3.5 (used to be 1.3.4).

11.Multiple changes to the Soot Eclipse plugin:
 - The SootOutput view is now deprecated. Instead Soot writes to a proper
   console that provides better formatting and other options.
 - Soot extenders do not longer have to put their code into Soot's plugin
   directory. Instead one can put the code into a normal Eclipse Java
   project and configure this project as "main project" in the Soot run
   configuration. The Main class is then automatically loaded from that
   project.
 - Examples: Hit New -> Examples -> Soot to learn Soot by example.
   Currently there are four example projects provided, from our well-known
   tutorials.
 - Soot classpath variables: The Soot example projects include
   sootclasses.jar via a platform-independent build path variable
   SOOTCLASSES. This variable always points to sootclasses.jar, and
   programmers extending Soot can include Soot on their build path by just
   using the variable, avoiding downloading Soot a second time.              

Improvements and API changes in version 2.2.4 (Fri, Apr 27, 2007)

1. Soot is now able to infer a main-class if no main-class attribute
   is given.

2. soot.jimple.toolkits.annotation.nullcheck.BranchedRefVarsAnalysis
   has been deprecated and replaced by 
   soot.jimple.toolkits.annotation.nullcheck.NullnessAnalysis.

3. Added EXPERIMENTAL transactions/lock allocation code, including
	- may-happen-in-parallel analysis
	- interprocedural information flow analysis
	- thread-local objects analysis
	- thread-based side effect analysis

Improvements in version 2.2.3 (Sat, Mar 18, 2006)

1. Dava:
    - Handling invocation of parent constructor and other constructors
      within a constructor
    - DeInlining final static fields.
    - Ensuring initialization of all final fields
    - Using copy propagation to remove un-necessary copy stmts
    
2. many bug fixes and improvements for Dava

3. many bug fixes and improvements for Java To Jimple 

Improvements in version 2.2.2

Improvements in version 2.2.1 (Mon, Jan 3, 2005)

1. update Soot - Eclipse plugin to work with Eclipse versions 3.x

Improvements in version 2.2 (Fri, Dec 17, 2004)

1. changes to class resolution see: 
   http://svn.sable.mcgill.ca/wiki/index.php/ClassResolver
2. major reorganization of control flow graphs and extended support for 
   exception control flow see: 
   http://svn.sable.mcgill.ca/wiki/index.php/CFG%20Reorganization
3. complete and stable JavaToJimple module
4. Java 1.4 sytle handling of method and field references see:
   http://svn.sable.mcgill.ca/wiki/index.php/MethodAndFieldReferences
5. minor fixes and cleanup of Shimple
   * made dominator analysis and structure classes public
   * Shimple factory added for finer grained control
6. updates to Soot - Eclipse Plugin
   * improved visual attributes in source
   * attributes legend
   * attribute management (for turning sets of attributes on/off)
   * interactive cfgs see:
   * growable graphical callgraph (experimental)
7. jimple validator does partial type checking (-validate switch)
8. testsuite
9. minor bug fixes
10. extended set of options

Improvements in version 2.1 (Wed, 17 Dec 2003)
1. bug fixes
2. switched from make to ant
3. many improvements to Shimple and Eclipse plugin
4. first release of Java-to-Jimple compiler
5. initial support for context-sensitive call graph and points-to analysis

Improvements in version 2.0.1 (Fri, 18 Jun 2003)
1. bug fixes
2. minor improvements and cleanup of Shimple, including:
 a) Significantly improved constant propagator analysis.
3. ant build file
4. new framework for adding Soot system tests

Improvements in version 2.0 (Fri, 6 Jun 2003)
1. a true whole-program mode
2. rewrite of the call graph code to provide more information (particularly
   about implicit calls) with a cleaner interface
3. first release of Eclipse plugin
4. first release of Shimple (SSA-Jimple)
5. refactoring of tag aggregators to make it easier to introduce new
   annotations for use by JITs
6. new, better organized option parsing code, accessible to other code
   (so you can set an option with a method call, rather than having to
   pass in a command line string)
7. elimination of most global variables; the remaining ones have been
   isolated so that Soot is restartable, and could be easily modified to
   allow concurrent Soot sessions on different threads.
8. bug fixes

Improvements in version 1.2.5 (Fri, 14 Feb 2003)
1. Spark improvements:
 a) now supports truly on-the-fly call graph construction, resulting
    in much smaller call graphs.
 b) alias edge algorithm tweaked to be faster.
 c) bug fixes.
2. General bug fixes.
3. Some performance improvements, plugging of memory leaks.
4. Reinstate support for JDK versions 1.2 through 1.4.

Improvements in version 1.2.4 (Fri, 22 Nov 2002)

1. main change is the inclusion of development version of Spark: Soot Pointer
   Analysis Research Kit. Available documentation for Spark includes the
   Spark poster (http://www.sable.mcgill.ca/publications/#poster-2002-1)
   and the Spark options 
   (http://www.sable.mcgill.ca/soot/tutorial/phase-options/spark.ps).
   More detailed documentation in the form of a master's thesis is being
   written. (olhotak)
2. Soot should now correctly handle classes compiled with 1.4.x versions
   of javac (olhotak)
3. new FastHierarchy class designed to eventually replace inefficient and
   confusing Hierarchy class (olhotak)
4. fix for missing control-flow graph edges from predecessors of statements
   in nested try blocks to some exception handlers (jjorge1)
5. make virtual call resolution respect access modifiers (olhotak)
6. other bug fixes

Improvements in version 1.2.3 (Wed, 08 May 2002)

1. fix to SourceLocator for the case where there are no actual
     paths in the soot classpath. (plam)
2. fix to HashMutableDirectedGraph to remove wayward extra
     preds and succs edges to a removed node. (fabien)
3. fix for Windows classpaths which contain less than two letters.
     Formerly, this crashed in charAt(1).  (plam, reported by Mary Payne)
4. add an option '-A LineNumber' to keep line number attributes in
   source class files. (feng)
5. make SootClass.containsBafBody() public (requested by David Airapetyan);
     avoid creating duplicate names in CommonSubexpressionEliminator
         (requested by Florian Mateoc).  Both changes by plam.
6. added changes by Michael Pan (pan@math.tau.ac.il) which make Soot 
     more parsimonious with respect to memory usage:
     it may use up to 20% less maximum heap space.
     Changes were: - sharing of coffi CONSTANT_Utf8_info objects
                   - SootMethod exceptions list created on demand
                   - get[Sub]Signature() storage hack useful for Sun JDK
7. minor changes in HashChain (remove useless null checks, states increments
   and modified insertBefore(List) ) (fabien)
8. fix to AbstractHost to avoid creation of thousands of empty lists (fabien)
9. added patch by Stephen Andrew Neuendorffer <neuendor@eecs.berkeley.edu>
     to make DeadAssignmentEliminator more aggressive: we no longer declare
     InstanceFieldRefs which read from 'this' in a non-static method to be
     'isEssential'. (plam)
10. Added some Timer.end() calls to JimpleBody.java
    and CoffiMethodSource.java, so that the '--time' option no
    longer causes runtime exceptions.  While this lets you
    specify '--time' without crashing soot, the accuracy of the
    reported times remains questionable. (jjorge1)
11. Improved Psuedotopological ordering, replaced deprecated code 
    (Vectors and Stacks), fixed line numbering, 
    fixed a minor UnitGraph bug (berndl)
12. Kludged together a bug fix for Alex Warshavsky (walex@il.ibm.com) where
    use-original-names sometimes produced Jimple Locals with #'s 
    in their names. (plam)
13. Fixed bug reported by Stephen Andrew Neuendorffer where CopyPropagator
     doesn't quite fail in the expected way when there's a variable used
     without definition (plam)
14. Added basic XML support in the form of an XMLStmtPrinter. Using
     the --xml command option will generate an XML version of the 
     class based on the Jimple statements and extended with static
     program data and transformation/analysis results. (deng)
15. Added entryInitialFlow method in FlowAnalysis, which allows a 
    new analysis just return flow object for entry points. 
	The customizeFlowGraph is still called. Currently it is 
	compatible with old analyses, but new analysis should override
	entryInitialFlow method rather than customizeFlowGraph.
    Forward and backward flow analyses are changed respectively.
	                                                (fqian)
16. Fixed local variable name problem in coffi.LocalVariable....
    reported by Nathan Kitchen. See fix 12 also.    (fqian)
17. Small improvement in ArraySparseSet and BackwardFlowAnalysis pointed
    out by Florian Loitsch.                         (fqian)
18. Added several utility classes for drawing generate graphs. 
    UnitGraph, BlockGraph, and Hierarchy can produce "dot" file
	for generating postscript files. 
	See soot/util/DotGraph*.java
	(More function will be added on.)               (fqian)
19. Added one flag in soot.Main.java to indicate the compilation for
    J2ME library which does not have Cloneable and Serializable interfaces.
	This affects the typing system which handcodes both interfaces.
	Has not been tested on J2ME. But when the flag is set to false, 
	original function is not affected.
	Patch provided by Stephen Cheng.                (fqian)
20. Fixed some obvious bugs in CFGViewer, added simple options. (fqian)
21. Added a new implementation of PRE by Florian Loitsch, merging 
    was done by Ondrej. The old implementation of PRE was obsoleted.
    The new implementation to be tested.                (fqian)


Improvements in version 1.2.2 (Thu Mar 29 11:45:13 EST 2001)

1. added \t case for soot.util.StringTools.getQuotedStringOf.
    (reported by Thomas VanDrunen)
2. CFG.java is improved to handle nested JSR/RETs
3. improved some exception message reported by Olivier Zendra.
4. The Jasmin package is also updated to 1.06-s10, the bug in grammar
   for attributes is fixed.
5. replaced 'getNodes().contains()' by 'containsNode()' in 
   "jimple/toolkits/invoke/MethodCallGraph.java", which speeds up
   the call graph construction significantly (by John Jorgensen.)

Improvements in version 1.2.1 (Fri Jan 12 15:22:59 EST 2001)

1. permissions changes and fix to Baf/JasminClass for BooleanType.
2. removal of a UnitGraph hack that shouldn't have been there.
3. internal changes to cmdline parsing, watch for bigger changes coming
4. fix by patrice to work around stupid jar bug
5. minor changes to jar bug workaround
6. fix for grimp precedence
7. improved error messages in the event of missing .dat files.
8. new build system (via the SootCompiler class).


Improvements in version 1.2.0 (Thu Sep 28 21:41:16 EDT 2000)

1. An annotation framework is included in this release.
2. A clean implementation of VTA is finally included in Soot.
3. New implementation of typing system included
     - contains nice integer typing system.
4. HashChain has a fast 'contains' method.
     (this was announced on the soot mailing list)
5. A branched flow analysis framework is present.
     - This allows different results to be propagated on
        different edges of a branch.
6. Implemented an Array Bounds Check and a Null Pointer Check algorithm.
   The new option '-A' annotates class files with analysis results.
      - See the new annotation tutorials.
7. Other minor bugs were zapped.
 
(release by Feng Qian)

Improvements in version 1.0.1 (not released)

1. Eliminated useless Directed interface.

2. Fixed bug in HashChain where null could wheedle its way into a HashChain.  Bad stuff!

3. Added attribute support and null pointer checks / array bounds checks.

4. Added common subexpression elimination.

Improvements in Release 1 (version 1.0.0)

Some of these changes occurred earlier but are only mentioned now.

1. Changed the version numbering scheme to follow the Linux
version numbering scheme.

2. Added warning for unconsumed phase options.

3. Fixed bug with original name recovery.

4. Added documentation.

5. Fixed another aggregation bug where multiple field refs between
def, use were not correctly handled.

6. Added static class synchronizer.

7. Changed format of Jimple source again, now removing the dots
for keywords.

8. Fixed issue with absolute paths in dos/windows

Improvements in 021400

1. Fixed the SootResolver so that it displays the name of
the missing classfile before quitting.

2. Merged in partial code for Partial Redundancy Elimination
(not working yet).

3. Added more documentation comments.  (Thanks patrice!)

4. Added EquivTo interface, made all Jimple Values implement 
EquivTo, and provided EquivalentValue wrapper objects.  Using an
EquivalentValue allows Collections to see structurally equal 
Values as satisfying .equals().  This is not universally useful;
for instance, the Units chain should not use EquivalentValues!

5. Added jikes.def and javac.def files.  Invoking "jams foo"
will copy the .dat files from the src/ directories to the classes/
directories.

6. Patrice added some support for a framework allowing Baf peephole
optimizations to be added in a uniform manner.

7. Added soot.PackAdjuster and soot.jimple.StmtPrinter classes.
These classes can be overridden by placing custom classes with the 
same names earlier in the CLASSPATH.  This permits external classes
to be called by Soot at strategic moments in Soot's execution.

Improvements in 1.beta.6

1. A Jimple parser has been added to the Soot framework.  Soot can
now read both .class files and .jimple files (produced by the
"--jimple" option).  Parsing .jimple files is faster than
reading .class files, and these files can be modified easily in a text
editor.

2. Different phases of Sootification can now be passed command-line
options.  For instance, the static inliner (si) in the
whole-jimple optimization pack (wjop) can told that it is to
expand code by no more than a factor of 5, by giving the following
command-line option:  -p wjop.si expansion-factor:5.

3. Some whole-program (excluding library) optimizations are
included, namely a static method binder and a static inliner.  The
inliner is enabled by default when the -W option is given.
It can be disabled with the option  -p wjop.si disabled:true -p
wjop.smb disabled:false.

Improvements in 1.beta.5

1. Baf, a streamlined form of bytecode has been added to Soot.  This is
useful for performing low level bytecode optimizations.  We can produce code
which is as good as javac now with the option "--final-rep baf".
Unfortunately this option creates bytecode which is different than what the Java
Virtual Machine is used to and so causes the JVM to crash.  For this reason, the
option is not enabled by default.

2. Major restructuring of the API has taken place.

3. The entire SpecJVM98 suite can be processed by Soot correctly, as well as
numerous other Java programs.

4. Process times are reasonable: a program can be usually put through Soot
within 2.5 times the execution time of javac on the original program.

Improvements in 1.beta.4

Date: Thu, 18 Mar 1999 01:05:44 -0500
From: Raja Vallee-Rai <rvalleerai@sable.mcgill.ca>
To: soot-list@sable.mcgill.ca
Subject: Soot-1.beta.4 released.

Hello soot-list subscribers,


A new version of Soot has been released!  The notable improvements are:

1. Grimp, an aggregated Jimple has been added to Soot. This is a
medium-high level intermediate representation which
allows you to manipulate statements as trees instead of 3-address code.
This is particularly useful for code generation and
decompilation. 

2. The code produced by Soot from Jimple code (by first converting it to
Grimp code) has been substantially improved. We
are now 2 to 3 percent away from the original classfile in performance.
(this is without any optimizations performed) 

3. A subroutine duplicator has been installed as a temporary solution
for the jsr bytecode. This means that there is no
concept of the subroutines in the framework and that the classfiles
produced by Soot are 'jsrless'. 

4. The speed of the transformations has been improved. 

5. Numerous bug fixes have been made. 


See http://www.sable.mcgill.ca/soot for more details.


Best regards,

Raja Vallee-Rai
Sable Research Group

Improvements in 1.beta.3

Date: Wed, 27 Jan 1999 21:11:49 -0500
From: Raja Vallee-Rai <rvalleerai@sable.mcgill.ca>
To: soot-list@sable.mcgill.ca
Subject: Soot version 1.beta.3 available.

Hello,

A new version of Soot has just been released.  This version fixes
numerous bugs and provides improved performance.  You can acquire the
new version at

http://www.sable.mcgill.ca/software

See the release notes at
http://www.sable.mcgill.ca/soot/releaseNotes.html for more details. 
Make sure you download the latest version of SableUtil to benefit from
the bug fixes and improvements in that package as well.

--------------------------------

For our next version we would like some feedback from our users.  Please
vote on the feature you would like to see implemented next:

A. Fix the handling of the JSR bytecode.
B. Improve the efficiency of the bytecode being generated.
C. Improve the speed of the bytecode -> Jimple transformation.

Only one vote per person, please. :)


Best regards,

Raja Vallee-Rai
Sable Research Group