I have a function in Kotlin that I want to inline despite not being high-order.
I find that through out the project I have numerous occurrences of such scenario.
Therefore, to ignore compiler warnings about using inline functions, I have to use lots of Suppress("NOTHING_TO_INLINE") annotations through out my project and it's starting to bother me.
Is there any way to block this warning for the whole project by, for instance, a compiler option?
I'd like to know how to do this in IntelliJ IDEA.
It doesn't look like it's possible to disable the inspection globally. I can't find a setting in IntelliJ at least. You can, however, disable the inspection for an entire file:
#file:Suppress("NOTHING_TO_INLINE")
If you press ALT+Enter on the warning, you'll get an option to suppress it which just contains suppressing for the function, class (if it's in one), or for the entire file. Any inspections you can disable usually has a "Disable inspection" option on this list (which NOTHING_TO_INLINE does not have).
You can, however, disable the warnings when compiling (but it does not affect the inspection; you'll still see warnings while editing) by adding suppressWarnings to the compileKotlin task. Note that this disables ALL warnings, not just specific ones.
compileKotlin {
kotlinOptions.suppressWarnings = true
}
Related
I have tried annotating a field with
org.checkerframework.checker.nullness.qual.NonNull
org.jetbrains.annotations.NotNull
javax.annotation.Nonnull
And in all cases, assigning a null to it generates no complaints from IntelliJ 2016.2.
public class GreetingController {
#NotNull Integer x = 3;
public void foo() { x = null; }
}
That all compiles fine according to IntelliJ.
This page from IntelliJ specifically states that "IntelliJ IDEA highlights the problems “on-the-fly”, so you can see the inspection results right in the editor." I have even copied the example code (public class TestNullable) into my editor and it produces no errors.
This other page from IntelliJ states you can change the annotations it responds to. So I chose javax.annotation.Nonnull and made sure that was the one I was using in my code, still no luck.
To be clear, what I'm hoping for, and what I understand should be provided, is that the editor window / compiler alerts me to the problem (I am not looking for a runtime check, NullPointerException already works fine at runtime.)
In case it didn't work in real time, I tried "Rebuild Project".
I'm sure this must work, what am I doing wrong?
I have uploaded an example of this not working here: ZIP download.
As I can see from your screenshots and the sample project, IntelliJ IDEA does show you the warnings. Note that these warnings are shown by the code inspections which are running on the fly and will be displayed in the editor or in the Analyze | Inspect Code results. These warnings will not be displayed by the compiler.
Note that you can configure the warnings highlighting if needed (for example add the underwave effect):
You can also change the severity of the inspection (like to Error):
You may also want to vote for this feature request:
IDEA-78625 Provide inspection severity level that will work like validation and abort compilation
As a bonus, pay attention to the javax.annotation.Nullable annotation, it may be not what you think it's for, see this comment and the documentation. For some years IntelliJ IDEA has incorrectly suggested to use this annotation, while the correct one for such cases would be javax.annotation.CheckForNull:
This annotation is useful mostly for overriding a Nonnull annotation.
Static analysis tools should generally treat the annotated items as
though they had no annotation, unless they are configured to minimize
false negatives. Use CheckForNull to indicate that the element value
should always be checked for a null value.
"Settings" > "Inspections" > "Probable Bugs" > "Constant conditions & exceptions"
Tick the first option: "Suggest #NotNull annotation for methods that possibly return null and report nullable values passed to non-annotated parameters.
Click "Configure Annotations". By default, Intellij will use their own annotations from org.jetbrains.annotation. I was using the more general (my own opinion) annotations from javax.annotation.
I set Nullable to: javax.annotation.Nullable
I set NotNUll to : javax.annotation.NotNull
In order to set these new options, you must click them, then click the tiny checkmark button to the right to set it. Selecting the javax.annotation annotations then hitting "OK" will NOT lock in the new settings, you must use the checkbox button.
After successfully specifying javax.annotation.Nullable and javax.annotation.NotNull, the code correctly highlighted null problems.
The best that this can do is offer up warnings. It will not stop compilation from happening, since the annotations do not prohibit or preclude code compilation from taking place.
Be sure that you have the appropriate inspections enabled in your IDE, and be sure that you remain aware of what parameters you're passing into your method. The IDE can at best warn you, but it can't really stop you.
Alternatively, introduce a unit test to fail if that method receives a null parameter, and rely on that to ensure that you're not breaking code or expectations.
I want to have some "god mode" boolean toggles to quickly change functionality inside my app.
The previous developer created #define macros for these in a file that's included (indirectly) in every other file, but there are some disadvantages:
Every change has to be commited in git
Every change forces a build of the entire project
Is an xconfig file suited for this case? I accomplished what I want by:
Creating an .xconfig file
Adding it to the project
Adding HACKS_TEST = YES to it
Adding preprocessor macro HACKS_TEST=${HACKS_TEST} to the project target
Adding static BOOL const IOHacksTest = HACKS_TEST
and upon logging it, the value is printed correctly. But, if I change the HACK_TEST's value in the .xconfig file, it still builds the entire project. Also, adding one 'hack', needs modifications in the .xconfig file, project target's preprocessor macro section and project's global file.
Is it possible to achieve this without the need of an entire project build?
Is there another solution that doesn't require these many modifications on new 'hacks' adding?
Is there another solution that is more appropriated to my needs and/or easier/better?
What is the best way to have them set to NO for the Release configuration?
Edit:
Along the selected solution, I also added this in the Globals.m file:
if #debug
BOOL IOHacksTest = YES;
#else
BOOL IOHacksTest = NO;
#endif
so for Release builds, all the "hacks" are turned off automatically.
Every change has to be commited in git
You can't, or really shouldn't try to, avoid this. If there's some setting that can affect how the whole project works, it should be archived with the project.
As you've discovered, putting the values in an .xcconfig file doesn't have much advantage over having #defines for everything - if you change the xcconfig file, yo have to rebuild everything.
To avoid that, you'll want to split the declaration and definition of the variables between a header and a source file. So, you'd have a goobals.h file that contains "extern" declarations, like this:
extern BOOL IOHacksTest;
And a goobals.m file that has:
BOOL IOHacksTest = YES;
Then you include goobals.h wherever you need the declarations. If you need to change a value, you only need to recompile a single file.
1, Is it possible to achieve this without the need of an entire
project build?
If you want code to be compiled differently, based on those changing values, then there is no way to do it. In fact, you want the project to be recompiled, because the generated code needs to change.
The best way to prevent an entire project build is to put those values into a header file, and selectively include that header file only in files that need to know about the values that could change.
Is there another solution that doesn't require these many modifications on new 'hacks' adding?
There is no way getting around it if you want compile-time detection. If you change the values, then the compiler has to generate new code. How could the compiler generate the right code if it does not know about the custom changes you want to make with your "changes?"
Is there another solution that is more appropriated to my needs and/or easier/better?
That depends on your actual needs, which you didn't state in the original question. If you must have compile-time knowledge of the changes, you can use the header files, or you can add values to .xcconfig files, or just set them in the project. However, you will still have to recompile all the code that is impacted every time the values change.
If you want the settings to be changed at run time, then this is precisely what NSUserDefaults is designed for. Set an initial value in the source code, or in the Info.plist file, and set it to user-defaults on the app's first launch. Thereafter, you can manage the values from user-defaults.
What is the best way to have them set to NO for the Release configuration?
static BOOL builtInDebugMode = !!(DEBUG);
EDIT
Also, what does !!(DEBUG) do? – Iulian Onofrei
That's a logical-not operator, employed twice. It ensures that its operand will always be either 1 or 0. It was there to guide you to how it could be done (i.e., consider DEBUG or NDEBUG at compile time).
By default, your xcode configuration will have DEBUG=1 for debug builds. You could set it to DEBUG=0 for release builds, and use the above code.
As long as your code properly handles DEBUG, this should be fine. Unfortunately, some code incorrectly uses #ifdef DEBUG rather than #if DEBUG which could cause issues.
Thus, you may be better off with something like...
#if DEBUG
static BOOL builtInDebugMode = YES;
#else
static BOOL builtInDebugMode = NO;
#endif
When compiled in debug mode, DEBUG will be defined as 1. When not in debug mode, it will not be defined at all.
Or, you could add a definition to your project file DEBUG_VALUE
You can set some value in NSUserDefault, is very easy to use, I think.
You can add buttons to your app's UI to toggle these values, or you can let app query some website for this value at first launch, depends your needs.
NSUserDefault is fast.
And you can query at launch, and use it util app quit.
I'm trying to suppress an instance of a warning in my code. I use the intellij feature alt-enter to add the suppress warning annotation to the method, also tried with the class. But whenever I hit the rebuild button, the warning keeps coming up as if the suppress warning annotation isn't being recognised.
I've read a few answers that suggest disable completely, or you can suppress individual entry using this notation. I know I can disable it completely through settings, but I'd rather just suppress this instance. Does anyone know why it wouldn't be suppressing the warning?
Information:Using Eclipse compiler to compile java sources
Information:Compilation completed successfully with 1 warning in 9 sec
Information:0 errors
Information:1 warning
...TagValidator.java
Warning:(41, 16) java: Type safety: Unchecked cast from java.lang.Object to java.util.List<common.model.Tag>
Method in question:
#SuppressWarnings("unchecked")
#Override
public void validate(Object target, Errors errors) {
logger.entry();
List<Tag> list;
list = (List<Tag>) target;
for (Tag tag : list) {
if (tag.getTag().length() > 30) {
errors.rejectValue("tags", "tags.length");
break;
}
}
logger.exit();
}
This appears to be an issue with the eclipse (ejc) compiler. When the javac compiler is used, no warnings are present upon make/compilation. But if you switch to the eclipse compiler, then you get the warnings. In the (closed) IntelliJ IDEA bug report ecj doesn't use #SuppressWarnings the JetBrains development team indicated that
IDEA uses such suppressions for its own error highlighting, which is compiler-independent. Similarly, it runs external compiler as is, without any interference.
So IDEA isn't doing anything to tell ejc not to use/honor the SuppressWarnings annotations.
I've got the following code in a thread in my application:
while (true) {
if (ready) {
progressIndicatorController.value++;
return;
}
}
The ready variable is changed from a delegate method. This code works great when I open the application by clicking the "Run" button in Xcode's toolbar. However, if I open this application's .app (which I create by clicking Product > Archive and then following the steps) this code somehow doesn't work anymore.
progressIndicatorController.value is never incremented and this if-statement never evaluates to true. What could cause this problem?
This is probably caused by optimization from the compiler.
When you build with Archive, XCode enabled optimization in the compiler that could throw this kind of code away. I think setting the ready variable to volatile could fix your problem, altough if I were you I'd just try to rewrite it so it doesn't trigger this problem.
You can test with optimization turned on by choosing Edit Schemes in the scheme dropdown. Then set Build Configuration to Release in the Run MyApp.app. Don't forget to set it back to Debug when you're done though, as the debugger gets somewhat confused when optimization are on (i.e. you can't see the value of most variables, some breakpoints may behave erratically, etc...)
I'm using a static lib thats giving me a warning when uploading my binary for review by apple.
The method in the static lib that causes the warning(non-public selectors) is never called by me, its corresponding .h is deleted from my proj, but warning still persists.
Given that I know the method name causing the problem, is there a way for me to open/edit this .a and comment/delete the offending piece of code and then use the modified .a in my project.
I don't have access to the .a source to recompile it, and its very old and the creator of it has no contact details for me to track down.
Many Thanks,
-Cake
Quick and dirty solution: Open the .a file in a hex editor and change all instances of the name. Leave the function name the same length so that offsets in the file don't change, just change a letter or something like that. I did a quick test, adding a dummy function to a subproject we're building as a static library then tweaking the function name in the .a file (there were five instances, for what that's worth) and everything built okay. I don't see any reason it wouldn't pass the App Store check after that.
I'm really surprised the function was still there in the final build, though—I thought Dead Code Stripping was supposed to clean out any unused code. Huh.
http://opensource.apple.com/source/cctools/cctools-809/
I don't presume to get your bounty, because I haven't provided an easy solution. But yes, it in theory is possible. You have your work cut out for you.
There are several solutions, depending on your lib and project.
In your build settings :
Enable "dead code stripping" if possible : If the method is never used (even internally), the symbol will be deleted.
Use "Unexported symbol file" : Simply add the symbol into a file and it will be removed from the binary. This will work even if the symbol is used internally.
Enable "Deployment Postprocessing" and "Strip Linked Product" with "Strip Style" set to "All symbol"
(Not sure) Use "Symbols Hidden by Default". This is related to the code generation and should not affect linking, but just in case everything above failed...
No need to hack the binary files. Just turn off the compiler's "unused selectors" warning: -fno-unused-selectors.