7

I have some really noob question. I tried to create installation for my test app with jpackage in OpenJDK 14. Here is what I did:

first, created custom JRE with

jlink --module-path "C:\Java\javafx-sdk-14\lib" --add-modules javafx.controls,javafx.fxml --output hello\myjre

and that was successful. I copied arguments from my Eclipse from Run Configurations. After that made installation with jpackage

jpackage --name HelloFX --input hello --main-jar HelloFX.jar --runtime-image hello\myjre

That created .msi file, I run it and it created entry in my Win10 applications. Of course, I have no idea how to find that in windows menu, but it is placed in my C:\Program Files\HelloFX. I located Icon and Application file with Duke image, when I tried to run application meesage "Failed to launch JVM" pop up.

Can someone help me, what am I doing wrong? I really want to make this work and dive deeply in JavaFX.

Mark Rotteveel
  • 100,966
  • 191
  • 140
  • 197
Dalibor Krivokapic
  • 167
  • 3
  • 10

1 Answers1

16

The JavaFX JARs included with the JavaFX SDK do not include the native code. Instead, said code is in the bin directory. That means your custom runtime image created by jlink does not have the necessary native code to run JavaFX. You have two options:

  1. Download the JMOD files from Gluon and use those when creating the custom runtime image. You would put the JMOD files on the --module-path instead of the regular JAR files.

    • Note another way to get the JMOD files is to use a JDK distribution that includes JavaFX. Such distributions may no longer be provided by Oracle but other distributors still offer them (e.g. Azul Zulu, BellSoft Liberica, etc.). If you have a JDK that includes JavaFX then you don't need to place JavaFX on the --module-path, just make sure the JavaFX modules are resolved while packaging and during execution (via requires directives or --add-modules arguments).

  2. Use the JavaFX JARs that are published to Maven Central instead of the SDK. The Maven Central JARs embed the native code.

In both cases, make sure to use the JMOD/JAR files for your operating system—JavaFX is platform-specific.

I believe the first option is the best. When using JMOD files with jlink the native code is included with the custom runtime image in the same way as the native code specific to the JRE. If you use the second approach the native code will still be included with the custom runtime image but it will have to be extracted to some location on your computer (e.g. <user-home>/.openjfx) before it can be used1. In other words, the first option is cleaner.

1. Note this extraction is done automatically by JavaFX.

Slaw
  • 37,820
  • 8
  • 53
  • 80
  • When adding JavaFX JMOD directory to `--module-path`, is it necessary to still add JavaFX SDK to `--module-path`? – Hendra Anggrian Jan 02 '22 at 21:44
  • Not when using `jlink` and/or `jpackage`. – Slaw Jan 03 '22 at 07:46
  • 1
    An important note about jmod files from the [jmod documentation](https://docs.oracle.com/en/java/javase/11/tools/jmod.html): *”This format is transportable but not executable, which means that you can use it during compile time or link time but not at run time.”* Which, to me, is confusing. It means you can use `javac` to compile against jmods, *but* you **must** link the jmod files into an image for the `java` command from that image to be able to use the functions in them. – jewelsea Sep 01 '23 at 02:52
  • @jewelsea Yeah, it makes things a little more confusing. My guess as to why JMODs are not executable is due to the format explicitly allowing native code to be bundled. That native code does not matter at compile-time; only the Java code matters. And at link time that same native code is extracted and placed next to the built-in native code. But to use JMODs at run-time, I believe any native code would have to be extracted to some _writable_ (and executable?) and possibly platform-specific location. I suspect they didn't want to define a standard way to handle that. – Slaw Sep 01 '23 at 03:18