We use this type of Reflection only in our test framework and not in production. We want add some jar files to the app classloader depending of our test code. Is there any workaround for it? How can we get access to not exported classes?
java.lang.reflect.InaccessibleObjectException: Unable to make member of class jdk.internal.loader.ClassLoaders$AppClassLoader accessible: module java.base does not export jdk.internal.loader to unnamed module #9f73a2
at jdk.internal.reflect.Reflection.throwInaccessibleObjectException(Reflection.java:414)
at java.lang.reflect.AccessibleObject.checkCanSetAccessible(AccessibleObject.java:174)
at java.lang.reflect.Method.checkCanSetAccessible(Method.java:191)
at java.lang.reflect.Method.setAccessible(Method.java:185)
Accessing module-internal classes can not be done from code. This is on purpose. But there is a workaround from the command line - a non-standardized option on java that does what you want:
--add-exports <module>/<package>=<target-module>(,<target-module>)*
updates <module> to export <package> to <target-module>,
regardless of module declaration.
<target-module> can be ALL-UNNAMED to export to all
unnamed modules.
Note
Java 9 is a moving target and the exact syntax has been changed a couple of times - as was this answer. This means (a) some of the comments below may seem outdated and (b) the flag might not work exactly like that. Please leave a comment and it will get fixed.
Related
I'm using Flow to help author a JS project. If I want to provide a libdef file to supplement it do I need to create it manually, or am I able to execute some magic command that I'm not aware of yet which will generate the lib def for me?
Something like $ flow-typed doyourmagic would be nice.
EDIT:
Found this https://stackoverflow.com/a/38906578/192999
Which says:
There's two things:
If the file is owned by you (i.e. not a third party lib inside node_modules or such), then you can create a *.js.flow file next to it that documents its exports.
If the file is not owned by you (i.e. third party lib inside node_modules or such), then you can create a libdef file inside flow-typed/name-of-library.js
For .js.flow files
you write the definitions like this:
// #flow
declare module.exports: { ... }
For libdef files you write the definitions like this:
declare module "my-third-party-library" { declare module.exports: {... } }
For my question I fall into the "is owned by you" camp.
I guess I'm confused as to:
How I write these files.
How/where I publish these files to package it up for another project to reference.
Also, why do I need to create the .js.flow file manually? Can this not be magically generated? Perhaps that's the intention going forward but not implemented yet.
I found a nice guide showing how to package flow code together with the compiled code. So:
You do not have to write your own libdefs, you can use the entire flow source code. If you want a definition with only the type declarations, you can look into flow gen-flow-files, although that is still experimental and might fail.
You can package them as *.js.flow and the flow checker will automatically pick those up when you import your library.
In roboelectric when you write your test class you have declare the #RunWith annotation of which there is RobolectricGradleTestRunner and RobolectricTestRunner.class. What is the difference and which one should we use ? Why is there 2 in the first place. If I use RobolectricTestRunner then it does not work for me it says some weird error like :
"java.lang.IllegalStateException: You need to use a Theme.AppCompat theme (or descendant) with this activity."
However this seems to go away if I use RobolectricGradleTestRunner.
RobolectricTestRunner was first and is for maven based projects.
RobolectricGradleTestRunner was for gradle based projects because some paths have changed.
Since robolectric 3.1.1 this is not more necessary and is now deprecated.
For more details see also http://robolectric.org/getting-started/
Some modules just seem to be invisible to Flow. For example I have react-native-overlay installed via npm into my node_modules directory but I get a whole bunch of errors like this from Flow:
[js/components/DatePickerOverlay.js:18
18: let Overlay = require('react-native-overlay');
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ react-native-overlay. Required module not found
This module doesn't have types so it would be fine if I could just get Flow to ignore it entirely.
Here's my .flowconfig (based on React Native's one):
https://gist.github.com/almost/20c6caf6d18d0e5c689f
As you can see I'm on flow 0.20.1 and I have module.system=haste (as required by React Native)
I tried adding a //$FlowIgnore comment to the import lines but then Flow complains about an unneeded ignore comment! I also tried creating a react-native-flow.js.flow file with a dummy export which seemed to work at first but then after a flow restart stopped working.
Any ideas for how to either help Flow find this module or make it ignore the import line completely?
Looks like you're ignoring it here: https://gist.github.com/almost/20c6caf6d18d0e5c689f#file-flowconfig-L42-L50
If you don't mind manually typing it up, add a react-native-overlay.js to your interfaces and type up a couple signatures.
This is happening because the library doesn't exist in flow-typed.
A simple fix could be creating the following entry in the .flowconfig file:
[ignore]
<PROJECT_ROOT>/libdefs.js
[libs]
./libdefs.js
If using flowtype < 0.60.0 add in libdefs.js
// #flow
declare module "react-native-overlay" {
declare var exports: any;
}
Or if using flowtype > 0.60.0
declare module 'react-native-overlay' {
declare module.exports: any;
}
Note: any is an unsafe type so you can always take advantage of improve the definition of the library
Hope that helps,
So far I have been able to succesfully use boost, cereal and gtest using biicode but I am having troubles with sqlite. I am trying to use it doing the following:
#include <sqlite3.h>
So I edited my biicode.conf to include those lines, including the alising for the header:
[requirements]
sqlite/sqlite:9
[includes]
sqlite.h: sqlite/sqlite/sqlite3/sqlite3.h
But when I try to call bii cpp:build it does the following
WARN: Removing unused reference to "sqlite/sqlite: 9" from myuser/test "requirements"
Then I ended up with the expected:
database_impl.cpp:(.text+0x516): undefined reference to `sqlite3_exec'
Surprisingly, the compilation succedd even though sqlite3.h is obviously not included but that's maybe because the call to sqlite is from a template function.
I have looked at the example but CMakeList.txt does not seem to add any additional includes directories. For example for boost I had to add:
SET(Boost_USE_STATIC_LIBS OFF)
bii_find_boost(COMPONENTS chrono system filesystem log thread REQUIRED)
target_include_directories(${BII_BLOCK_TARGET} INTERFACE ${Boost_INCLUDE_DIRS})
target_link_libraries(${BII_BLOCK_TARGET} INTERFACE ${Boost_LIBRARIES})
But the two examples I found here and here don't seem to add anything to the includes directories, not even a link folder. I suppose sqlite has to be compiled with your sources so how do I make biicode add those files to my projects automatically ?
There're several problems. You just wrote in [includes] section sqlite.h instead of sqlite3.h and you should only write the prefix, sqlite/sqlite/sqlite3, later instead of the full dependency name.
Then, you can solve it so:
[requirements]
sqlite/sqlite: 9
[includes]
sqlite3.h: sqlite/sqlite/sqlite3
Or, you could try the SQLite version uploaded into fenix user:
[requirements]
fenix/sqlite: 0
[includes]
sqlite3.h: fenix/sqlite
Don't worry about a "WARN" message that says biicode is ignoring sqlite3.c file because it's harcoded into the block CMakeLists.txt to catch this file ;)
Note: you should write your external #includes's with double quotes instead of <>, because the last ones are referred to system headers and biicode could be using some system deps and you don't realize it.
I want to load dynamically XML meta-data files that are not in the classpath (the XML meta-data files will be generated after launching my app).
I thought I could call the method
MetaDataManager.registerFile(java.lang.String fileURLString, FileMetaData filemd, ClassLoaderResolver clr)
Then, I tried the API enhancement, so I added the following lines:
JDOEnhancer enhancer = JDOHelper.getEnhancer();
enhancer.setVerbose(true);
enhancer.addClasses(ClassToPersist.class.getName()).enhance();
getClass().getClassLoader().loadClass(ClassToPersist.class.getName());
The following jars are in the classpath: datanucleus-api-jdo.jar, datanucleus-connectionpool.jar, datanucleus-core.jar datanucleus-rdbms.jar, jdo-api.jar, asm.jar.
But when I launch my app, I get this exception:
Caused by: mypackage.MyException:
org.datanucleus.api.jdo.exceptions.ClassNotPersistenceCapableException: The class "mypackage.ClassToPersist" is not persistable. This means that it either hasnt been enhanced, or that the enhanced version of the file is not in the CLASSPATH (or is hidden by an unenhanced version), or the Meta-Data/annotations for the class are not found.
Do you please have any idea how to fix this ?
PS: I also noticed that the method enhance returns 0 which indicates that the class has not been enhanced (we could exclude the other options)
Thanks
I think I found an easy way to fix this.
At runtime, I created a jar that contains the updated metadata file as META-INF/package.jdo. Then I added this jar to the classpath as described here
Using this hacks, I didn't need to re-enhace my class since it is enhanced after compilation.
(But for information, I had to call JDOEnhancer.addFiles() to enhance my class.)
So your call to loadClass has loaded the unenhanced class (since it would have been loaded in order to perform the enhancement presumably), and you didn't follow the tutorial that is provided at http://www.datanucleus.org/documentation/development/dynamic_class_metadata_enhance_runtime.html