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

Add module-info.java to checker-qual #6326

Merged
merged 13 commits into from
Mar 6, 2024
Merged

Conversation

mtf90
Copy link
Contributor

@mtf90 mtf90 commented Nov 30, 2023

This PR adds a module-info.java to the checker-qual project so that the checker-framework annotations can be directly used in projects that want to use the jlink or jpackage tools for distributing software. The current automatic modules (via MANIFEST.MF's Automatic-Module-Names) are not supported by jlink.

The corresponding issue is #4559.

@mtf90 mtf90 mentioned this pull request Nov 30, 2023
@mernst
Copy link
Member

mernst commented Jan 16, 2024

@mtf90
./gradlew javadoc is failing with this error (and others):

framework/src/main/java/org/checkerframework/common/basetype/BaseTypeChecker.java:3: error: package com.google.common.collect is not visible
import com.google.common.collect.ImmutableSet;
                        ^
  (package com.google.common.collect is declared in module com.google.common, which is not in the module graph)

Could you please fix that? I would love to get this pull request merged. Thanks!

this is in accordance with the declaration order of JDK module definitions
make sure to compile all files with Java 9 first so that the compile may detect missing "requires" statements
@mtf90
Copy link
Contributor Author

mtf90 commented Jan 17, 2024

Dear @mernst, I disabled the module path for the other javadoc processes as well which seemed to do the trick.

However, I am still not that satisfied with the PR.

  • The code for disabling module semantics is copied across various task configurations. Is there a global plugin/task/way to set this at a central place? Unfortunately, I'm not very well-versed with Gradle.
  • The module dependencies are mainly specified for resolving links in the javadocs (hence the static modifiers). Since the javadoc generation does not use module semantics, you may as well delete them. Especially if the other modules may at one point support named modules as well, you wouldn't even be able to specify the static dependency to org.checkerframework.checker because modules don't allow for cyclic dependencies. Currently, I cannot see how this can be implemented cleanly without a major refactoring. You probably have to live with the fact that most IDEs do not properly resolve the {@link}s now.
  • I don't know anything about Android specifics, so I have just disabled the module definition for this artifact completely.
  • The one good thing is that you probably do not need extra integration-tests. Since the compile tasks (hopefully) pick up the module definition, they should respect the specified exports declarations as well. So if you add a new annotation package and forget to export it as well, you should directly run into compilation errors when referencing it elsewhere.

@mtf90
Copy link
Contributor Author

mtf90 commented Jan 17, 2024

I managed to clean up the configs some more, given that slowly I get the hang of sourceSets.

Also, the two pipelines seem to have failed due to a timeout, not an actual error (as all other pipelines have passed).

@mtf90 mtf90 marked this pull request as ready for review January 17, 2024 20:57
@sgammon
Copy link

sgammon commented Mar 6, 2024

@mtf90 / @mernst friendly bump? we ran into this in a downstream project. i would be happy to test if helpful

@mtf90
Copy link
Contributor Author

mtf90 commented Mar 6, 2024

@sgammon, as mentioned in #6326 (comment), I don't see a clean way of implementing this without some larger refactoring (which is not my decision to make). So I can understand the reluctance of the maintainers.

As mentioned in #4559 (comment), there is a way to workaround this in downstream projects. For example, I managed to successfully implement it in LearnLib/learnlib#122 including a passing jlink integration test.

@mernst
Copy link
Member

mernst commented Mar 6, 2024

My apologies; I lost track of this pull request. I'm going to assign it for review and hopefully we can get it merged.

@smillst smillst merged commit 2776c6d into typetools:master Mar 6, 2024
29 checks passed
@sgammon
Copy link

sgammon commented Mar 6, 2024

@mernst No problem, and thank you guys for merging this!

(Tagging google/guava#2970 as well)

copybara-service bot pushed a commit to google/error-prone that referenced this pull request Mar 8, 2024
## Summary

Minimal set of changes to ship a `module-info.java`, and support JVM 1.8 for the `annotations` module only.

### Reasoning

It costs almost nothing to support _only_ the annotations on JVM 1.8, for projects which transitively include `error-prone-annotations`, which can happen via a simple dependency on something like Guava. But, it would be ideal to [ship a `module-info.java`](#2649), so `annotations` is now a `Multi-Release` JAR, with a `module-info.class` in `META-INF/versions/9`.

I am working downstream to [support JPMS in Guava](google/guava#2970), and this PR will be necessary for that to eventually land (without hackery). [Checker Framework recently merged their support](typetools/checker-framework#6326) which means Error Prone is one of the last things on the classpath for Guava without a `module-info.java` definition.

Automatic Modules also aren't the answer because they cannot be used with `jlink`. Such dependencies must be processed by tools like Moditect before they can be used in fully modular Java builds. Because Error Prone Annotations is a dependency in Guava, and is itself very popular, many projects need Error Prone to adopt JPMS so they can ship their own code with modular Java support.

There may be libraries out there that depend on the annotations _and not the compiler_, who wish to ship a MRJAR and retain JVM 1.8 support (Guava).

### Non-release version number change

One wrinkle here is that the Maven project version [is used unconditionally](https://github.com/apache/maven-compiler-plugin/blob/74cfc72acae4f55708bca189b2170167e83df6b3/src/main/java/org/apache/maven/plugin/compiler/CompilerMojo.java#L304-L305) [as the `--module-version`][0]; however, [`HEAD-SNAPSHOT` is not a valid module identifier](https://lists.apache.org/thread/qtghq13qgjoc4pn6ovsdxgnjnm4ozv4d). This leaves only a few choices to support JPMS through recent (`3.8.x+`) Maven Compiler plugin versions:

- `HEAD-SNAPSHOT` needs to become `1.0-HEAD-SNAPSHOT`, and then it is a valid `--module-version`
- Or, Maven Compiler Plugin needs to ship a bugfix and Error Prone will need to wait for a release (if a bugfix ships at all)

I understand this can be a breaking change somewhere within Google, but I don't see a way around this without resorting to severe build hacks. I've gotten it to build anyway by building the `module-info.java` only in a separate Maven project, and then copying it into the JAR at the last moment, but there are serious issues with that route so I suggest it be abandoned in favor of a version string change during dev, if possible.

## Changelog

- feat: add `module-info` to `annotations` module
- feat: ship `annotations` as a `Multi-Release` JAR
- feat: support `1.8` through latest JDK for `annotations` module
- fix: remove automatic module definition from `annotations` module
- fix: `HEAD-SNAPSHOT` → `1.0-HEAD-SNAPSHOT`, because of a Maven Compiler Plugin issue [precluding use as a module version][0]

Fixes and closes #2649

[0]: https://issues.apache.org/jira/browse/MCOMPILER-579

Fixes #4311

FUTURE_COPYBARA_INTEGRATE_REVIEW=#4311 from sgammon:feat/jpms d209b0c
PiperOrigin-RevId: 614005681
copybara-service bot pushed a commit to google/error-prone that referenced this pull request Mar 8, 2024
## Summary

Minimal set of changes to ship a `module-info.java`, and support JVM 1.8 for the `annotations` module only.

### Reasoning

It costs almost nothing to support _only_ the annotations on JVM 1.8, for projects which transitively include `error-prone-annotations`, which can happen via a simple dependency on something like Guava. But, it would be ideal to [ship a `module-info.java`](#2649), so `annotations` is now a `Multi-Release` JAR, with a `module-info.class` in `META-INF/versions/9`.

I am working downstream to [support JPMS in Guava](google/guava#2970), and this PR will be necessary for that to eventually land (without hackery). [Checker Framework recently merged their support](typetools/checker-framework#6326) which means Error Prone is one of the last things on the classpath for Guava without a `module-info.java` definition.

Automatic Modules also aren't the answer because they cannot be used with `jlink`. Such dependencies must be processed by tools like Moditect before they can be used in fully modular Java builds. Because Error Prone Annotations is a dependency in Guava, and is itself very popular, many projects need Error Prone to adopt JPMS so they can ship their own code with modular Java support.

There may be libraries out there that depend on the annotations _and not the compiler_, who wish to ship a MRJAR and retain JVM 1.8 support (Guava).

### Non-release version number change

One wrinkle here is that the Maven project version [is used unconditionally](https://github.com/apache/maven-compiler-plugin/blob/74cfc72acae4f55708bca189b2170167e83df6b3/src/main/java/org/apache/maven/plugin/compiler/CompilerMojo.java#L304-L305) [as the `--module-version`][0]; however, [`HEAD-SNAPSHOT` is not a valid module identifier](https://lists.apache.org/thread/qtghq13qgjoc4pn6ovsdxgnjnm4ozv4d). This leaves only a few choices to support JPMS through recent (`3.8.x+`) Maven Compiler plugin versions:

- `HEAD-SNAPSHOT` needs to become `1.0-HEAD-SNAPSHOT`, and then it is a valid `--module-version`
- Or, Maven Compiler Plugin needs to ship a bugfix and Error Prone will need to wait for a release (if a bugfix ships at all)

I understand this can be a breaking change somewhere within Google, but I don't see a way around this without resorting to severe build hacks. I've gotten it to build anyway by building the `module-info.java` only in a separate Maven project, and then copying it into the JAR at the last moment, but there are serious issues with that route so I suggest it be abandoned in favor of a version string change during dev, if possible.

## Changelog

- feat: add `module-info` to `annotations` module
- feat: ship `annotations` as a `Multi-Release` JAR
- feat: support `1.8` through latest JDK for `annotations` module
- fix: remove automatic module definition from `annotations` module
- fix: `HEAD-SNAPSHOT` → `1.0-HEAD-SNAPSHOT`, because of a Maven Compiler Plugin issue [precluding use as a module version][0]

Fixes and closes #2649

[0]: https://issues.apache.org/jira/browse/MCOMPILER-579

Fixes #4311

COPYBARA_INTEGRATE_REVIEW=#4311 from sgammon:feat/jpms d209b0c
PiperOrigin-RevId: 614026439
*/
module org.checkerframework.checker.qual {
// javadoc-only dependencies
requires static java.compiler;

Choose a reason for hiding this comment

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

If this is really for Javadoc only, wouldn't it be cleaner to remove those requires static statements and replace them by --add-reads options when invoking the javadoc tool?

Copy link
Contributor Author

Choose a reason for hiding this comment

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

Technically, you can even skip the --add-reads because the javadoc phase does not read the module descriptor (due to circular dependencies which are not allowed with JPMS).

I mainly added them to be as explicit as possible with the dependency declarations. For example, IntelliJ would flag even more references if these dependencies were not specified. I don't know how well IDEs are able to include phase-specific build-flags.

Do you have any issues with these declarations? As far as I know, requires static should not affect dependent modules.

Choose a reason for hiding this comment

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

No issue that I'm aware of for now. I thought that those declarations are a risk, because the code could accidentally use those dependencies without warning from the compiler. The argument about being explicit can also be turned the other way around, i.e. they are not describing the reality if the code has no dependency to those modules. Anyway, this is not a major issue.

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.

5 participants