February 2024 - jextract-dev

RFR: 7903682: Misformatted --help message
by Jorn Vernee 23 Feb '24

23 Feb '24

Fix some formatting issues in the `--help` message. The new message looks correct to me now: > .\build\jextract\bin\jextract.bat --help Usage: jextract <options> <header file> Option Description ------ ----------- -?, -h, --help print help -D --define-macro <macro>=<value> define <macro> to <value> (or 1 if <value> omitted) -I, --include-dir <dir> add directory to the end of the list of include search paths --dump-includes <file> dump included symbols into specified file --header-class-name <name> name of the generated header class. If this option is not specified, then header class name is derived from the header file name. For example, class "foo_h" for header "foo.h". --include-function <name> name of function to include --include-constant <name> name of macro or enum constant to include --include-struct <name> name of struct definition to include --include-typedef <name> name of type definition to include --include-union <name> name of union definition to include --include-var <name> name of global variable to include -l, --library <libspec> specify a shared library that should be loaded by the generated header class. If <libspec> starts with :, then what follows is interpreted as a library path. Otherwise, <libspec> denotes a library name. Examples: -l GL -l :libGL.so.1 -l :/usr/lib/libGL.so.1 --use-system-load-library libraries specified using -l are loaded in the loader symbol lookup (using either System::loadLibrary, or System::load). Useful if the libraries must be loaded from one of the paths in java.library.path. --output <path> specify the directory to place generated files. If this option is not specified, then current directory is used. --source generate java sources -t, --target-package <package> target package name for the generated classes. If this option is not specified, then unnamed package is used. --version print version information and exit ------------- Commit messages: - fix help message formatting Changes: https://git.openjdk.org/jextract/pull/225/files Webrev: https://webrevs.openjdk.org/?repo=jextract&pr=225&range=00 Issue: https://bugs.openjdk.org/browse/CODETOOLS-7903682 Stats: 4 lines in 1 file changed: 0 ins; 0 del; 4 mod Patch: https://git.openjdk.org/jextract/pull/225.diff Fetch: git fetch https://git.openjdk.org/jextract.git pull/225/head:pull/225 PR: https://git.openjdk.org/jextract/pull/225

2 2

Incomplete usage documentation due to misformatted Messages.properties
by Joel Knighton 23 Feb '24

23 Feb '24

Hello, On line 78 of Messages.properties, the final backslash is trailed by a space character. This leads to the termination of the element value associated with the key jextract.usage. As a result, usage is not displayed for --output through --version options when the jextract command line utility is run. Best, Joel Knighton

2 1

Re: Growing jextract
by Manuel Bleichenbacher 23 Feb '24

23 Feb '24

Hi Panama team Thanks again for quickly addressing the issues. I was able to upgrade my library to JDK 22 using the latest jextract version on all three platforms. All issues that I have reported previously are fixed. I've discovered one more issue. For a particular type (on macOS), jextract creates code because a dependency was not specified. I guess it would be supposed to abort with an error regarding the missing dependency. The problem is quickly fixed if the dependency is included. The minimal commands to reproduce it is: JEXTRACT=../../../../jextract/build/jextract/bin/jextract SDK_DIR=/Applications/Xcode.app/Contents/Developer/Platforms/MacOSX.platform/Developer/SDKs/MacOSX.sdk $JEXTRACT --output ../../src/main/java \ -I $SDK_DIR/usr/include \ -lIOKit.framework \ --header-class-name IOKit \ --target-package net.codecrete.usb.macos.gen.iokit \ --include-struct IOCFPlugInInterfaceStruct \ iokit_helper.h The file iokit_helper.h: #include <IOKit/usb/IOUSBLib.h> #include <IOKit/IOCFPlugIn.h> The missing dependency is CFUUIDBytes and the path to it is that a typedef (REFIID) referring to it is used in a parameter of a function pointer (QueryInterface) of IOCFPlugInInterfaceStruct. I notice that I don't use any "--include-typedef" anymore. Since jextract resolves typedefs to the underlying type, only the referred struct needs to be included. Is there any use left for "--include-typedef"? The biggest annoyance is that the generated code still needs to be manually modified (unless I've overlooked a jextract option). The details can be found here: https://github.com/manuelbl/JavaDoesUSB/blob/jdk22/java-does-usb/jextract/R… It's related to two cases of locating the library at run-time: referring to major version 1 of libudev in Linux, and referring to a system framework on macOS. Regards Manuel

3 5

RFR: 7903680: Dependency checker doesn't handle function pointers and nested types
by Maurizio Cimadamore 22 Feb '24

22 Feb '24

This patch fixes few issues in the new dependency checker. As usual, the new visitor was skipping a bunch of cases, which resulted in errors not being reported: * if a variable, function parameter/return, typedef had function pointer type, we did not propagate the visit to the function type * if there's nested declarations in a typedef, function, variable, we also did not propagate. I added the required propagation, and a new test which checks that errors are generated in these cornery situations. ------------- Commit messages: - Initial push Changes: https://git.openjdk.org/jextract/pull/224/files Webrev: https://webrevs.openjdk.org/?repo=jextract&pr=224&range=00 Issue: https://bugs.openjdk.org/browse/CODETOOLS-7903680 Stats: 154 lines in 3 files changed: 145 ins; 0 del; 9 mod Patch: https://git.openjdk.org/jextract/pull/224.diff Fetch: git fetch https://git.openjdk.org/jextract.git pull/224/head:pull/224 PR: https://git.openjdk.org/jextract/pull/224

2 2

RFR: 7903678: regression: jextract tool is broken in the latest early access build
by Athijegannathan Sundararajan 21 Feb '24

21 Feb '24

double-quote should be used instead of single-quote around the value of --add-options option. ------------- Commit messages: - 7903678: regression: jextract tool is broken in the latest early access build Changes: https://git.openjdk.org/jextract/pull/223/files Webrev: https://webrevs.openjdk.org/?repo=jextract&pr=223&range=00 Issue: https://bugs.openjdk.org/browse/CODETOOLS-7903678 Stats: 1 line in 1 file changed: 0 ins; 0 del; 1 mod Patch: https://git.openjdk.org/jextract/pull/223.diff Fetch: git fetch https://git.openjdk.org/jextract.git pull/223/head:pull/223 PR: https://git.openjdk.org/jextract/pull/223

2 2

RFR: Update README based on feedback
by Jorn Vernee 20 Feb '24

20 Feb '24

- Change references from jdk 21 to jdk 22 - Update JEP number and recommend LLVM version - Amend instructions for jdk version needed to run gradle - Add note that example indicating that a native library is required ------------- Commit messages: - Update README.md - Update README or comments Changes: https://git.openjdk.org/jextract/pull/222/files Webrev: https://webrevs.openjdk.org/?repo=jextract&pr=222&range=00 Stats: 9 lines in 1 file changed: 2 ins; 1 del; 6 mod Patch: https://git.openjdk.org/jextract/pull/222.diff Fetch: git fetch https://git.openjdk.org/jextract.git pull/222/head:pull/222 PR: https://git.openjdk.org/jextract/pull/222

2 2

RFR: 7903674: jextract should ignore non-enum constants inside enum
by Athijegannathan Sundararajan 20 Feb '24

20 Feb '24

* checking cursor kind of child cursor to be CursorKind.EnumConstantDecl. * Piggybacking to fix similar check of record child element kinds to be CursorKind.FieldDecl. * Added null check for createTree return value in collectNestedTypes method as well. * In addition to jextract tests, I ran all samples on Mac OS. All fine with this change. Also, ran jextract script for Mac OS from https://github.com/manuelbl/JavaDoesUSB.git. ------------- Commit messages: - 7903674: jextract should ignore non-enum constants inside enum Changes: https://git.openjdk.org/jextract/pull/221/files Webrev: https://webrevs.openjdk.org/?repo=jextract&pr=221&range=00 Issue: https://bugs.openjdk.org/browse/CODETOOLS-7903674 Stats: 18 lines in 1 file changed: 7 ins; 0 del; 11 mod Patch: https://git.openjdk.org/jextract/pull/221.diff Fetch: git fetch https://git.openjdk.org/jextract.git pull/221/head:pull/221 PR: https://git.openjdk.org/jextract/pull/221

3 17

RFR: 7903669: jextract-22 jextract tool shows enable native access warning
by Athijegannathan Sundararajan 19 Feb '24

19 Feb '24

--add-options was defined twice in jlink command line. The latter one overrides the previous one! To specify multiple launcher options, single --add-options with quoted, space separated options has to specified in the jlink command line. ------------- Commit messages: - 7903669: jextract-22 jextract tool shows enable native access warning Changes: https://git.openjdk.org/jextract/pull/220/files Webrev: https://webrevs.openjdk.org/?repo=jextract&pr=220&range=00 Issue: https://bugs.openjdk.org/browse/CODETOOLS-7903669 Stats: 3 lines in 1 file changed: 0 ins; 2 del; 1 mod Patch: https://git.openjdk.org/jextract/pull/220.diff Fetch: git fetch https://git.openjdk.org/jextract.git pull/220/head:pull/220 PR: https://git.openjdk.org/jextract/pull/220

2 3

RFR: 7903672: Some issues with missing dependency errors
by Jorn Vernee 19 Feb '24

19 Feb '24

- Only error on missing dependencies, if the dependent is not skipped - Scan element types of arrays for missing dependencies as well - Don't issue warnings about unsupported types/fields for skipped elements I've pulled the code that issues errors for missing dependencies into a separate phase (`MissingDepWarner`), since the control flow in some of the visitor methods was kinda tricky to follow if this stayed mixed in with the `IncludeFilter` code. The new phase just cleanly checks whether the visited element has a `Skip` on it and then returns at the start, before doing any work. I've also re-ordered the phases a bit so that the phases that add Skips run first, so that we don't issue warnings/errors for things that are skipped any way by a later phase. Since `UnsupportedFilter` both adds Skips and issues warnings, it sits right in the middle. Q: should we also return early in `NameMangler` if an element is skipped? (Right now it's just doing extra work). ------------- Commit messages: - MissingDepWarner + reorder passes - add dummy field - scan element types of array fields - Only error on skipped dependencies, if the dependent is not skipped Changes: https://git.openjdk.org/jextract/pull/217/files Webrev: https://webrevs.openjdk.org/?repo=jextract&pr=217&range=00 Issue: https://bugs.openjdk.org/browse/CODETOOLS-7903672 Stats: 274 lines in 9 files changed: 246 ins; 24 del; 4 mod Patch: https://git.openjdk.org/jextract/pull/217.diff Fetch: git fetch https://git.openjdk.org/jextract.git pull/217/head:pull/217 PR: https://git.openjdk.org/jextract/pull/217

2 4

RFR: 7903673: Jextract throws NPE when generating enum constant javadoc
by Maurizio Cimadamore 19 Feb '24

19 Feb '24

This PR fixes `TreeMaker` so that it doesn't try to generate javadoc for `null` enum constants. Unfortunately I was not able to reproduce on Linux. The problem seems to appear on MacOS when extracting this enum: enum __CFByteOrder { CFByteOrderUnknown, CFByteOrderLittleEndian, CFByteOrderBigEndian }; But there doesn't seem to be anything peculiar with this (and this does work when extracted in isolation). It might indicate a deeper issue in libclang. For now, I made the code more robust, so that it handles possible null enum constants (these are left out from the IR). ------------- Commit messages: - Initial push Changes: https://git.openjdk.org/jextract/pull/219/files Webrev: https://webrevs.openjdk.org/?repo=jextract&pr=219&range=00 Issue: https://bugs.openjdk.org/browse/CODETOOLS-7903673 Stats: 9 lines in 1 file changed: 3 ins; 2 del; 4 mod Patch: https://git.openjdk.org/jextract/pull/219.diff Fetch: git fetch https://git.openjdk.org/jextract.git pull/219/head:pull/219 PR: https://git.openjdk.org/jextract/pull/219

3 6