| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
|
|
|
|
|
|
|
| |
initialization
Follow-up of commit 1eadaf928f4f61aae4de1c8bf33c5b77bdfa882f
- Refine MethodBinding detection for 'requiresStaticInitialization',
i.e. return type is a "compound type" and not a pointer
- JVMUtil_NewDirectByteBufferCopy: Throw FatalError if initializeImpl() has not been called
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
initialization
Follow-up of commit 9843e983a4fc71a64eb3de9cb364a1f4ffa56b3a
Only add static initialization code (java, native)
and hence native code 'JVMUtil_NewDirectByteBufferCopy(..)'
if either required _or_ forced via configuration.
This shall reduce possible sideffects
and dead code.
Add JavaConfiguration:
/**
* Returns true if the static initialization java code calling <code>initializeImpl()</code>
* for the given class will be manually implemented by the end user
* as requested via configuration directive <code>ManualStaticInitCall 'class-name'</code>.
*/
public boolean manualStaticInitCall(String clazzName);
/**
* Returns true if the static initialization java code implementing <code>initializeImpl()</code>
* and the native code implementing:
* <pre>
* static jobject JVMUtil_NewDirectByteBufferCopy(JNIEnv *env, void * source_address, jlong capacity);
* </pre>
* for the given class will be included in the generated code, always,
* as requested via configuration directive <code>ForceStaticInitCode 'class-name'</code>.
* <p>
* If case above code has been generated, static class initialization is generated
* to call <code>initializeImpl()</code>, see {@link #manualStaticInitCall(String)}.
* </p>
*/
public boolean forceStaticInitCode(String clazzName);
|
|
|
|
| |
call-by-value' JavaMethodBindingEmitter, array may still require NIO handling, also consider NIO array length.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Completing commit c3054a01990e55ab35756ea23ab7d7c05f24dd37
by allowing passing compound arrays via 'call-by-value.
- Creating linear temp heap, copying NIO values into it
and passing to C function.
Copy-back if not 'const', see below.
- Respect 'const' qualifier to skip write-back of
temp heap passed to C function
- See tag: // FIXME: Compound and Compound-Arrays
for code changes and validation of completeness
- triggers for compound arrays are:
- javaType.isArrayOfCompoundTypeWrappers()
- type.isArray()
- simplified const query by c-type: FunctionEmitter.isBaseTypeConst(ctype)
+++
Tests: Added call-by-value to test1.[ch] binding test!
|
|\ |
|
| |
| |
| |
| |
| |
| | |
functions list
Signed-off-by: Harvey Harrison <[email protected]>
|
| |
| |
| |
| | |
Signed-off-by: Harvey Harrison <[email protected]>
|
| |
| |
| |
| |
| |
| |
| |
| | |
/src/main/antlr/..
This gets the .g files unmixed from the non-generated java files for the cgram/jgram packages.
Signed-off-by: Harvey Harrison <[email protected]>
|
| |
| |
| |
| |
| |
| |
| | |
The annotations are all commented out, uncomment them and reduce the type warnings
during the build.
Signed-off-by: Harvey Harrison <[email protected]>
|
| |
| |
| |
| | |
Signed-off-by: Harvey Harrison <[email protected]>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
constructor
clone() can't be used without warnigns and is no more efficient that the copy constructor,
replace the call and kill the suppressed warning annotation. While here, mark the
modifier list as private final.
Signed-off-by: Harvey Harrison <[email protected]>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
returning struct instance
'compound call-by-value' is not efficient.
However, to allow mapping APIs utilizing passing small structs
as arguments and return values - this feature has been added.
+++
To return the struct value, native code needs to
allocate a NIO ByteBuffer and copy the data.
The stack return value can be dismissed afterwards and the
NIO buffer is returned.
We include this functionality for all generated [impl] classes,
native method:
'static jobject JVMUtil_NewDirectByteBufferCopy(JNIEnv *env, void * source_address, jlong capacity)'
(See: 'JavaEmitter.initClassAccessCode')
Since this code requires knowledge of java classes and methods,
for which a reference needs to be acquired, a static initialization method
has been introduced for all generated [impl] classes:
'private static native boolean initializeImpl();'
Per default the this method will be called in the new
static initializer block of the class,
which can be supressed via the configuration element:
'ManualStaticInit <class-name>'
'ManualStaticInit' can be used to issue the 'initializeImpl()'
call in a custom static initializer written by the user.
However, at the time 'initializeImpl()' gets called
the JNI native library must have been loaded, of course!
+++
- See tag: // FIXME: Compound call-by-value
for code changes and validation of completeness
Trigger for compond call-by-value in CMethodBindingEmitter is:
!cArgType.isPointer() && javaArgType.isCompoundTypeWrapper()
Trigger for compond call-by-value in JavaEmitter is:
t.isCompound()
+++
Further more we do tolerate 'javaType.isCPrimitivePointerType()',
i.e. adding comments for offset/size and field entries,
which are all NOP.
This allows to utilize the remaining fields of the native structure.
+++
Tests: Added call-by-value to test1.[ch] binding test!
|
|/
|
|
| |
JavaType.getDumpString(): Add more info ..
|
|
|
|
|
|
|
|
| |
reuse
Reuse a single stringbuilder to build all the comment strings and emit them if required.
Signed-off-by: Harvey Harrison <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
generated_class_user due to Java8 issues.
Java8's annotation processor as embedded within javac does not allow
referencing not-yet existing generated classes in a class source code
which will produce these generated classes via the annotation process.
Example:
+++
import com.jogamp.gluegen.structgen.CStruct;
public class Lala {
@CStruct(name="RenderingConfig", header="TestStruct01.h")
public RenderingConfig config;
}
+++
Above example illustrates that the type 'RenderingConfig'
does not exist at time of processing the annotation.
The type will be created via the annotation process itself.
Even though we pass '-proc:only', i.e. skip java compilation,
Java8's annotation processing via javac fails in such case.
We see this as a bug within javac's annotation processing itself!
+++
This workaround splits the annotation process and using the class as
generated by the former.
To allow this to work, CStruct receives a new field 'jname'
allowing to enforce the java-name of the structure
using a dummy field type like boolean.
@CStruct(name="RenderingConfig", jname="RenderingConfig", header="TestStruct01.h")
public boolean dummy;
Further more CStruct can be annotated on the package, i.e. 'package-info.java',
avoiding the dependency problem altogether.
To support multiple header files and types,
'CStructs' has been introduced using an array of 'CStruct'.
@CStructs({@CStruct(name="RenderingConfig", header="TestStruct01.h"), @CStruct(name="Pixel", header="TestStruct02.h")})
package com.jogamp.gluegen.test.junit.structgen;
Tests:
- Build w/ Java7 and Java8
- Validated 'major version 50' (Java 6) class files (OK)
|
|
|
|
| |
This reverts commit fccb2460832aedd1b51372aa1e7881770cb638c9.
|
| |
|
|\
| |
| |
| | |
'remotes/wwalker/bug_994_fix_cstruct_annotation_warning'
|
| |
| |
| |
| |
| | |
Updated the Java source version the annotation processor supports
to stop it from throwing a warning during the build process.
|
|\ \
| | |
| | |
| | | |
'remotes/wwalker/bug_992_allow_ignore_of_unnamed_structs'
|
| |/
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Augments the behavior of the Ignore keyword to
ignore when gluegen skips the emission of unnamed
structs. These structs are usually unnamed because
we've created opaque types for them, so we know
they're not going to be emitted and want to be
able to suppress the warning on a case-by-case
basis.
|
| | |
|
|/
|
|
|
|
|
|
| |
Fix 1: Only emit "int * _offsetHandle = NULL" if it will be used, to
avoid unused variable warning. Fix 2: Add "unsigned" to typecasts in C
function calls when needed to avoid implicit typecast warning. This
commit also adds a unit test for a method that uses an "unsigned char
**" parameter, to mimic the JOCL clCreateProgramWithBinary() function
that caused the typecast warnings.
|
|
|
|
| |
element, allowing to add custom API doc lines per method for the JavaMethodBindingEmitter
|
|
|
|
|
|
| |
Just directly call the copy-constructor and avoid using clone.
Signed-off-by: Harvey Harrison <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Changes emitted code from:
if (__addr_ == 0) {
throw new Exception("Method \"" + "$methodName" + "\" not available");
to:
if (__addr == 0) {
throw new Exception(String.format("Method \"%s\" not available", "$methodName"));
This removes all the redundant error message parts for each error string and only
stores the unique method name in a string.
Signed-off-by: Harvey Harrison <[email protected]>
|
|
|
|
| |
Signed-off-by: Harvey Harrison <[email protected]>
|
|
|
|
| |
Signed-off-by: Harvey Harrison <[email protected]>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
calling 'PropertyAccess.isPropertyDefined(propName, default)' through Debug class.
Calling 'Debug.isPropertyDefined(propName, default)' may be 'optimized' to
'PropertyAccess.isPropertyDefined(propName, default)', which would skip the modules Debug's class initialization.
Iff that happens, an AccessControlException may happen, due to requesting an insecure property,
since modules own Debug class has not been added it's trusted prefixes from within it's init block yet.
This seems to be a bug of the JVM .. to me, however .. the above description is the only
able to explain the issue at hand.
+++
Fix calls Debug class own static methods, either Debug.initSingleton() or Debug.debug(),
before calling 'isPropertyDefined(propName, default)'.
+++
Also mark Debug class static methods final!
+++
|
|
|
|
|
|
|
| |
- call append on the StringBuilder we are using rather than using String concatenation, which
uses StringBuffer.
Signed-off-by: Harvey Harrison <[email protected]>
|
|
|
|
|
|
|
| |
- use copy-constructor rather than clone to suppress type warnings
- annotate a Class as Class<?>
Signed-off-by: Harvey Harrison <[email protected]>
|
|
|
|
| |
Signed-off-by: Harvey Harrison <[email protected]>
|
|
|
|
| |
non overloading of c-funcs.
|
| |
|
|
|
|
| |
incl. access check; Utilize 'AccessibleObject.setAccessible(Field[], true)' for performance.
|
|
|
|
| |
instead of 'checkAllPermissions' if accessing the cached function handles.
|
|
|
|
| |
getDeclaredField() and setAccessible(true) due to package private handle fields.
|
|
|
|
|
|
|
|
| |
PROCADDRESS_VAR_PREFIX instance, add checkAllPermissions() for reset() and initEntry(..)
- Generated ProcAddressTable's function handles are all package private
- Generated ProcAddressTable's visibility can be set via 'AccessControl' config, default: public.
- ProcAddressTable's reset() and initEntry(..) perform checkAllPermissions() 1st.
|
|
|
|
| |
pointer' _must_ be private!
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
'gluegen' C Structs on-the-fly (2-pass actually).
Convenient annotation processing (APT) hooked to 'javac' (1.6) via gluegen.jar META-INF
service provider 'javax.annotation.processing.Processor' -> 'com.jogamp.gluegen.structgen.CStructAnnotationProcessor'.
Am implicit APT / JAVAC would be possible, however - to have a proper process
the unit test utilizes an explicit 2 pass run:
<!-- Annotation Processor Only - First -->
<javac destdir="${build_t.java}">
<classpath refid="junit.compile.classpath"/>
<compilerarg value="-proc:only"/>
<compilerarg value="-J-Djogamp.gluegen.structgen.debug"/>
<compilerarg value="-J-Djogamp.gluegen.structgen.output=${build_t.gen}/classes"/>
<src path="${test.base.dir}/com/jogamp/gluegen/test/junit/structgen"/>
</javac>
<!-- Javac Only - Second -->
<javac destdir="${build_t.java}">
<classpath refid="junit.compile.classpath"/>
<compilerarg value="-proc:none"/>
<src path="${test.base.dir}"/>
<src path="${build_t.gen}" />
</javac>
Original code from Michael Bien's 'superglue' git://github.com/mbien/superglue.git,
finally merged to GlueGen (as once intended).
Note: The APT javac pass requires to use 'gluegen.jar' instead of 'gluegen-rt.jar' !
The 2-pass process also alows using the runtime gluegen-rt.jar and hence ensures
clean namespace check at compilation.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
and privileged access)
This review focuses on how we perform permission checks,
or better - do we circumvent some assuming full privileges ?
Some native methods do need extra permission validation, i.e. loading native libraries.
Further more AccessController.doPrivileged(..) shall not cover generic code
exposing a critical feature to the user.
Further more .. we should rely on the SecuritManager, i.e. AccessControlContext's
'checkPermission(Permission)' code to comply w/ fine grained permission access.
It is also possible to have full permission w/o having any certificates (-> policy file).
+++
We remove implicit AccessController.doPrivileged(..) from within our trusted code
for generic methods, like Property access, temp. files.
+++
SecurityUtil:
- Remove 'getCommonAccessControlContext(Class<?> clz)',
which returned a local AccessControlContext for later restriction
if the passed class contains all certificates as the 'trusted' GlueGen class has.
- Simply expose convenient permission check methods relying on
SecurityManager / AccessControlContext.
PropertyAccess:
- 'protected static void addTrustedPrefix(..)' requires AllPermissions if SecurityManager is installed.
- Remove implicit doPrivileged(..) triggered by passed AccessControlContext instance,
only leave it for trusted prefixes.
IOUtil:
- Remove all doPrivileged(..) - Elevation shall be performed by caller.
DynamicLinker:
- 'public long openLibraryLocal(..)' and 'public long openLibraryGlobal(..)'
may throw SecurityException, if a SecurityManager is installed and the dyn. link permission
is not granted in the calling code.
Implemented in their respective Unix, OSX and Windows manifestation.
Caller has to elevate privileges via 'doPrivileged(..) {}' !
+++
Tests:
- Property access
- File access
- Native library loading
Manual Applet test (unsigned, but w/ SecurityManager and policy file):
> gluegen/test/applet
Applet has been tested w/ signed JAR w/ Firefox and Java7 on GNU/Linux as well.
Manual Application test (unsigned, but w/ SecurityManager and policy file):
com.jogamp.junit.sec.TestSecIOUtil01
- Run w/ SecurityManager and policy file:
- gluegen/scripts/runtest-secmgr.sh
- Run w/o SecurityManager:
- gluegen/scripts/runtest.sh
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
GetPrimitiveArrayCritical(..)
The 'carray' pointer returned from GetPrimitiveArrayCritical(..) was moved about the array offset
and used in ReleasePrimitiveArrayCritical(..) to release the pinpointed memory.
Even though this 'is' a bug by violating the _sparse_ specification, Hotspot impl. doesn't use the value at all (NOP)
and hence this code didn't produce an error since .. (Same w/ Dalvik).
Now the array offset is added while passing the carray pointer to the native function call
and hence is no more modified and the orig. value is passed to ReleasePrimitiveArrayCritical(..).
Tested w/ GlueGen unit tests and all JOGL unit tests (on Linux x64 w/ 'a' hotspot VM).
|
|
|
|
| |
Signed-off-by: Harvey Harrison <[email protected]>
|
|
|
|
|
|
| |
PointerBuffer, since referenced memory-size is arch dependent
Added extensive PointerBuffer unit tests w/ new mapping in generated test class.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Semantics Change:
ExtendedInterfaceSymbolsOnly was used for implementation generation only,
which is considered a bug!
- ExtendedInterfaceSymbolsIgnore C.java
- Ignore symbols in C.java for interface generation
- ExtendedInterfaceSymbolsOnly C.java
- Only use symbols in C.java for interface generation
- ExtendedImplementationSymbolsIgnore C.java
- Ignore symbols in C.java for implementation generation
- ExtendedImplementationSymbolsOnly C.java
- Only use symbols in C.java for implementation generation
- ExtendedIntfAndImplSymbolsIgnore C.java
- Ignore symbols in C.java for interface and implementation generation
- ExtendedIntfAndImplSymbolsOnly C.java
- Only use symbols in C.java for interface and implementation generation
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
d7fb6a7bcfbd2d5ac452abdcdd31fb4d71441c70"
This reverts commit 73e8828566404e864170688dfb4fd530a83f8add.
Convinced after discussing semantics w/ Harvey Harrison:
"the copy-constructor is way for predictable (as in branch-predictable)
and has better cache behavior, it can issue almost all the writes in
parallel, and has no exception catching.
So, the copy-constructor actually ends up being more efficient, and
you get typechecking to boot."
|
|
|
|
|
|
| |
d7fb6a7bcfbd2d5ac452abdcdd31fb4d71441c70
clone of ArrayList is more shallow than it's clone operation (?)
|
|
|
|
|
|
|
|
| |
ignores and unimplemented are both Set<Pattern>, they cannot contain a String,
immediately after these 'fast paths' we loop over the patterns calling .matches()
which makes sense, so remove the misleading tests.
Signed-off-by: Harvey Harrison <[email protected]>
|
|
|
|
|
|
|
| |
- Use the copy constructor rather than ArrayList.clone()
- constrain listsEqual so the compiler will warn about comparing lists of different types
Signed-off-by: Harvey Harrison <[email protected]>
|