Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Partial handling for restrictive annotations on varargs in unannotated code #1029

Merged
merged 13 commits into from
Sep 7, 2024

Conversation

msridhar
Copy link
Collaborator

@msridhar msridhar commented Sep 2, 2024

Partially addresses #1027. Most importantly, handles the case with unannotated methods and no restrictive annotations; we were getting warnings in some cases with the main branch, which impacts backward compatibility.

I started implementing full support for restrictive annotations on varargs parameters and then realized it would take some refactoring (since there can be a restrictive annotation on the varargs array itself or on its contents, and we have no way to express this in the current code structure). This PR contains some incomplete code towards that change, which I figured does no harm so I left it in for now. If desired I can remove it and put it in a separate PR.

@msridhar msridhar marked this pull request as ready for review September 2, 2024 23:04
Copy link

codecov bot commented Sep 2, 2024

Codecov Report

Attention: Patch coverage is 80.00000% with 6 lines in your changes missing coverage. Please review.

Project coverage is 87.23%. Comparing base (fe9476a) to head (118dc17).
Report is 1 commits behind head on master.

Files with missing lines Patch % Lines
...c/main/java/com/uber/nullaway/NullabilityUtil.java 61.53% 2 Missing and 3 partials ⚠️
...away/src/main/java/com/uber/nullaway/NullAway.java 87.50% 0 Missing and 1 partial ⚠️
Additional details and impacted files
@@             Coverage Diff              @@
##             master    #1029      +/-   ##
============================================
- Coverage     87.28%   87.23%   -0.05%     
- Complexity     2117     2129      +12     
============================================
  Files            83       83              
  Lines          6943     6965      +22     
  Branches       1346     1356      +10     
============================================
+ Hits           6060     6076      +16     
- Misses          456      458       +2     
- Partials        427      431       +4     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

@@ -200,7 +200,8 @@ public void testSkipJetbrainsNotNullOnVarArgsFromThirdPartyJars() {
" ThirdParty.takesNullableVarargs(o1);", // Empty var args passed
" ThirdParty.takesNullableVarargs(o1, o4);",
" ThirdParty.takesNullableVarargs(o1, (Object) null);",
" // BUG: Diagnostic contains: passing @Nullable parameter '(Object[]) null' where @NonNull",
" // No error: we require a type-use annotation for nullability of the varargs array itself",
// TODO should we preserve the old behavior in legacy mode?
Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@lazaroclapp with this change, we no longer detect an error when passing a null array in the varargs position to a third-party method with a JetBrains @NotNull annotation on the varargs parameter. On the one hand, getting an error here is the "desired" behavior, in that the annotation is supposed to apply to the array itself (see #720). But, everywhere else, a declaration annotation on the varargs argument is treated as applying to the contents of the array, not the array itself. To me, adding the extra code complexity to detect these errors in legacy mode for this corner case is not worth it; hopefully Kotlinc will be convinced to put annotations in the "right" place soon. But I'd be interested in your opinion.

Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I think without a coordinated change on kotlinc, I would prefer to keep the hack. But I don't feel super strongly about it. It's just an additional test of (a) is the annotation in bytecode, (b) is it the jetbrains annotation, (c) is it in var args... if all that is true, hacky override with a comment linking to the issue. No? Would this be hard to maintain?

Unfortunately, Java code interfacing with Kotlin code is a big portion of what we will see in Android for the foreseeable future, so kotlinc quirks might be the kind of thing we can't ignore for the sake of uniformity. Though if you can convince them to generate the "right" bytecode, I'd be happy to see this hack disappear (hopefully we don't need then to analyze magic numbers to see which kotlinc version produced the jar)

Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

You're right; it looks like for now kotlinc will be emitting similar bytecodes. Fixed in 7330a20

@msridhar msridhar marked this pull request as draft September 2, 2024 23:45
@msridhar msridhar marked this pull request as ready for review September 3, 2024 03:03
Copy link
Collaborator

@lazaroclapp lazaroclapp left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Definitely an incomplete fix, as the description clearly states, but also a step in the right direction. See comments below. All addressable in follow ups, of course.

if ((arraySymbol.flags() & Flags.VARARGS) != 0) {
return Nullness.hasNonNullDeclarationAnnotation(arraySymbol, config);
}
return false;
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Is codecov correct here that we never test the case where this method returns false?

Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

It is correct. We should get coverage of this line when we add logic to process restrictive @NonNull annotations on array elements in positions other than the varargs position, which we will do in a follow up.

@@ -200,7 +200,8 @@ public void testSkipJetbrainsNotNullOnVarArgsFromThirdPartyJars() {
" ThirdParty.takesNullableVarargs(o1);", // Empty var args passed
" ThirdParty.takesNullableVarargs(o1, o4);",
" ThirdParty.takesNullableVarargs(o1, (Object) null);",
" // BUG: Diagnostic contains: passing @Nullable parameter '(Object[]) null' where @NonNull",
" // No error: we require a type-use annotation for nullability of the varargs array itself",
// TODO should we preserve the old behavior in legacy mode?
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I think without a coordinated change on kotlinc, I would prefer to keep the hack. But I don't feel super strongly about it. It's just an additional test of (a) is the annotation in bytecode, (b) is it the jetbrains annotation, (c) is it in var args... if all that is true, hacky override with a comment linking to the issue. No? Would this be hard to maintain?

Unfortunately, Java code interfacing with Kotlin code is a big portion of what we will see in Android for the foreseeable future, so kotlinc quirks might be the kind of thing we can't ignore for the sake of uniformity. Though if you can convince them to generate the "right" bytecode, I'd be happy to see this hack disappear (hopefully we don't need then to analyze magic numbers to see which kotlinc version produced the jar)

Copy link
Collaborator Author

@msridhar msridhar left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thanks for the review!

if ((arraySymbol.flags() & Flags.VARARGS) != 0) {
return Nullness.hasNonNullDeclarationAnnotation(arraySymbol, config);
}
return false;
Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

It is correct. We should get coverage of this line when we add logic to process restrictive @NonNull annotations on array elements in positions other than the varargs position, which we will do in a follow up.

@@ -200,7 +200,8 @@ public void testSkipJetbrainsNotNullOnVarArgsFromThirdPartyJars() {
" ThirdParty.takesNullableVarargs(o1);", // Empty var args passed
" ThirdParty.takesNullableVarargs(o1, o4);",
" ThirdParty.takesNullableVarargs(o1, (Object) null);",
" // BUG: Diagnostic contains: passing @Nullable parameter '(Object[]) null' where @NonNull",
" // No error: we require a type-use annotation for nullability of the varargs array itself",
// TODO should we preserve the old behavior in legacy mode?
Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

You're right; it looks like for now kotlinc will be emitting similar bytecodes. Fixed in 7330a20

@msridhar msridhar enabled auto-merge (squash) September 7, 2024 18:45
@msridhar msridhar merged commit e2e5394 into uber:master Sep 7, 2024
9 of 11 checks passed
@msridhar msridhar deleted the varargs-unannotated branch September 7, 2024 18:54
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

2 participants