spoonlabs / nopol Goto Github PK

In main/java/fr/inria/lille/repair/nopol/synth/ConstraintModelBuilder.java:103:

the assignment

int nbSecondSuccess = firstResult.getRunCount() - firstResult.getFailureCount();

should be

int nbSecondSuccess = secondResult.getRunCount() - secondResult.getFailureCount();

rename the two interfaces "Synthesizer"

rename BrutSynthesizer in DynamothNopol

rename DefaultSynthesizer in SMTNopol

rename everywhere (code+doc) brutpol in dynamoth

It should be interesting to have information about angelic values like:

• Location
• Type (conditionnal or statement)
• Map of value per test

 if (x!=1 || foo && X) // x!=1 || foo is original code

goal: angelic value of X and synthesis of X only

see
Math-88 and Math-85
Lang-51 and Lang-58

Ariel:
I know NOPOL collects the values of all local variables, method parameters, and class fields that are typed with a basic primitive type, but what about public fields of locally defined objects?

class MyClass{
  public int value;
}

foo(){
  MyClass bar = new MyClass();
  bar.value = <any value from anywhere>
  //int nopolHack = bar.value;
  ...
}

the value of bar.value wouldn't be collected unless I uncomment the last assignment. How can I make NOPOL collect this data?

I can't re-open #16, but the bug is the same

In src/main/java/fr/inria/lille/repair/nopol/spoon/ConditionalLoggingInstrumenter.java:69 the problem persists.

first put the corresponding dynamoth test in SMT mode

to identify pure methods, we can leverage purity analysis using
https://www.st.cs.uni-saarland.de/models/jpure/
or
https://github.com/proganalysis/type-inference/

from Ariel

what are expected to be the inputs listed after the following line on NOPOL's output?

10:02:33.611 [main] DEBUG f.i.l.c.s.ConstraintBasedSynthesis - Specifications:
input: {.....}
input: {.....}
...

maybe the constraints imposed by passing tests which execute the statement being patched?

and set it to default (instead of IF)

4 modes pre_then_if, condition, precondition, loop

in Dynamoth, check at startup one JDI class is available (eg Class.forName(com.sun.jdi.value)) and if not display an error message "add tools.jar in your classpath"

In main/java/fr/inria/lille/repair/nopol/spoon/brutpol/ConditionalInstrumenter.java:34 and
main/java/fr/inria/lille/repair/nopol/spoon/ConditionalLoggingInstrumenter.java:39

CtCodeSnippetStatement defaultValueEvaluation = getFactory().Code().createCodeSnippetStatement("try{spoonDefaultValue=" + subprocessor().getDefaultValue() + ";}catch(" + Exception.class.getCanonicalName() + " e){}");

Nopol uses the default name "e" for catch statements, which generates uncompilable code.

and put tests in @ignore

median/90a14c1ac8f56395389fade6f67872a9684e61f83099e634ac675eae04f391f3cc2b6f6ebe966f2488ce7e00a2cabb218f3b1372d4161b3c05d134b1b7f296d2/000

The part of the project https://github.com/SpoonLabs/nopol/tree/master/nopol/src/evaluation/ does not compile since the change of @danglotb

and set it to default. should yield less overfitting.

Nopol can synthesize the following expression

0 < cond ? 1 : 0

which java interprets as

(0 < cond) ? 1 : 0

according its operators precedence https://docs.oracle.com/javase/tutorial/java/nutsandbolts/operators.html

causing uncompilable code ("<" between int and boolean)

but Nopol meant

0 < (cond ? 1 : 0)

this parenthesis are missing in
main/java/fr/inria/lille/commons/synthesis/OperationCodeLine.java:43:

return String.format("(%s)%s(%s)%s(%s)", subContent(0), operator.firstSymbol(), subContent(1), operator.secondSymbol(), subContent(2));

it should be

return String.format("((%s)%s(%s)%s(%s))", subContent(0), operator.firstSymbol(), subContent(1), operator.secondSymbol(), subContent(2));

(reported by Ariel, running in SMT mode.

any idea? looks like there is a mix of SMT and Dynamoth.

Exception in thread "main" java.lang.NoClassDefFoundError: com/sun/jdi/Value
at fr.inria.lille.spirals.repair.expression.factory.AccessFactory.literal(AccessFactory.java:25)
at fr.inria.lille.spirals.repair.synthesizer.collect.spoon.ConstantCollector.process(ConstantCollector.java:78)
at fr.inria.lille.spirals.repair.synthesizer.collect.spoon.ConstantCollector.process(ConstantCollector.java:18)
at spoon.support.visitor.ProcessingVisitor.scan(ProcessingVisitor.java:92)
at spoon.reflect.visitor.CtScanner.visitCtBinaryOperator(CtScanner.java:266)
at spoon.support.reflect.code.CtBinaryOperatorImpl.accept(CtBinaryOperatorImpl.java:35)
at spoon.reflect.visitor.CtScanner.scan(CtScanner.java:145)
at spoon.support.visitor.ProcessingVisitor.scan(ProcessingVisitor.java:88)
at spoon.reflect.visitor.CtScanner.visitCtFor(CtScanner.java:417)
at spoon.support.reflect.code.CtForImpl.accept(CtForImpl.java:44)
at spoon.reflect.visitor.CtScanner.scan(CtScanner.java:145)
at spoon.support.visitor.ProcessingVisitor.scan(ProcessingVisitor.java:88)
at spoon.support.visitor.ProcessingVisitor.scan(ProcessingVisitor.java:64)
at spoon.reflect.visitor.CtScanner.visitCtBlock(CtScanner.java:273)
at spoon.support.reflect.code.CtBlockImpl.accept(CtBlockImpl.java:46)
at spoon.reflect.visitor.CtScanner.scan(CtScanner.java:145)
at spoon.support.visitor.ProcessingVisitor.scan(ProcessingVisitor.java:88)
at spoon.reflect.visitor.CtScanner.visitCtFor(CtScanner.java:419)
at spoon.support.reflect.code.CtForImpl.accept(CtForImpl.java:44)
at spoon.reflect.visitor.CtScanner.scan(CtScanner.java:145)
at spoon.support.visitor.ProcessingVisitor.scan(ProcessingVisitor.java:88)
at spoon.support.visitor.ProcessingVisitor.scan(ProcessingVisitor.java:64)
at spoon.reflect.visitor.CtScanner.visitCtBlock(CtScanner.java:273)
at spoon.support.reflect.code.CtBlockImpl.accept(CtBlockImpl.java:46)
at spoon.reflect.visitor.CtScanner.scan(CtScanner.java:145)
at spoon.support.visitor.ProcessingVisitor.scan(ProcessingVisitor.java:88)
at spoon.reflect.visitor.CtScanner.visitCtMethod(CtScanner.java:508)
at spoon.support.reflect.declaration.CtMethodImpl.accept(CtMethodImpl.java:58)
at spoon.reflect.visitor.CtScanner.scan(CtScanner.java:145)
at spoon.support.visitor.ProcessingVisitor.scan(ProcessingVisitor.java:88)
at spoon.support.visitor.ProcessingVisitor.scan(ProcessingVisitor.java:64)
at spoon.reflect.visitor.CtScanner.visitCtClass(CtScanner.java:309)
at spoon.support.reflect.declaration.CtClassImpl.accept(CtClassImpl.java:58)
at spoon.reflect.visitor.CtScanner.scan(CtScanner.java:145)
at spoon.support.visitor.ProcessingVisitor.scan(ProcessingVisitor.java:88)
at spoon.support.RuntimeProcessingManager.process(RuntimeProcessingManager.java:145)
at spoon.support.RuntimeProcessingManager.process(RuntimeProcessingManager.java:133)
at fr.inria.lille.commons.spoon.SpoonedFile.processModelledClasses(SpoonedFile.java:136)
at fr.inria.lille.commons.spoon.SpoonedFile.process(SpoonedFile.java:128)
at fr.inria.lille.commons.spoon.SpoonedFile.process(SpoonedFile.java:124)
at fr.inria.lille.repair.nopol.synth.DefaultSynthesizer.buildPatch(DefaultSynthesizer.java:94)
at fr.inria.lille.repair.nopol.NoPol.solveWithMultipleBuild(NoPol.java:183)
at fr.inria.lille.repair.nopol.NoPol.build(NoPol.java:117)
at fr.inria.lille.repair.nopol.NoPolLauncher.launch(NoPolLauncher.java:82)
at fr.inria.lille.repair.Main.main(Main.java:63)

TSE dataset: https://github.com/SpoonLabs/nopol-experiments/

EmSE dataset: Defects4J

SolverFactory seems not to use Config when using Nopol directly through the Launcher.

On src/main/java/fr/inria/lille/commons/trace/SpecificationTestCasesListener.java:80/84

The field inconsistentInputs is accessed through the self reference this, but it should be accessed using the corresponding getter, like all accesses to consistentInputs, which use consistentInputs().

from Ariel

Analysing introclassJava.median_90a14c1a_000{exec()V[69:69 0.5773502691896258
Failing test(s): SourceLocation introclassJava.median_90a14c1a_000:69
[WB 3]
First set:
[WB 3]
Second set:
[WB 4, WB 3, WB 1, BB 7]
1 input values set(s). There are not enough tests for SourceLocation introclassJava.median_90a14c1a_000:69 otherwise the patch would be "true" or "false"
Could not find a patch in introclassJava.median_90a14c1a_000{exec()V[69:69 0.5773502691896258

this statement (69) evaluates to false for all failing tests, and making it true would work for 3 out of 4 failing tests, so I think this case should be tagged as "Changing only this statement does not solve the bug" (the test WB 3 fails no matter the angelic value used on statement 69)

are there really 4 executions of this line?

with default to 10 minutes

Whats does the parameter --maxTime is expected to limit? the whole execution time of NOPOL or the execution time per test? what unit?

This should be said in Javadoc and Readme.

nopol should return Unix code

• 0 if patch found
• -1 if not found

Just obtained several errors like this using Nopol:

14:06:56.977 [pool-10-thread-1] DEBUG fr.inria.lille.repair.nopol.NoPol - Analysing SourceLocation com.rarchives.ripme.ripper.rippers.InstagramRipper:124 which is executed by 1 tests
14:06:58.787 [pool-345-thread-1] DEBUG x.java.compiler.DynamicClassCompiler - [Compilation errors]
InstagramRipper.java:120: error: variable imageURL might not have been initialized
            imageURL = imageURL.replaceAll("scontent.cdninstagram.com/hphotos-", "igcdn-photos-d-a.akamaihd.net/hphotos-ak-");