Avoiding Exceptions during automatic testing

RainerHeintzmann commented 3 years ago

JavaCall throws a java.awt.HeadlessException in Linux systems due to a missing "X11" display specification, when Git-Hub automatic testing is performed. Does anyone know how to avoid this? Also on Mac the tests fail due to a missing java installation. Are there easy fixed for this?

mkitti commented 3 years ago

For Mac, set the JAVA_HOME environmental variable to tell JavaCall where to find Java. Where are you getting Java for Mac from?

mkitti commented 3 years ago

I will get around to setting GitHub Actions CI here soon.

I would look into something like https://github.com/marketplace/actions/gabrielbb-xvfb-action

RainerHeintzmann commented 3 years ago

For Mac, set the JAVA_HOME environmental variable to tell JavaCall where to find Java. Where are you getting Java for Mac from? Thanks for the hints. Would be great, if the GitHub Actions CI can be convinced to deal with this. These tests work well on the Windows virtual machine (which seems to include a java installation) but they fail for the named reasons on Linux and Mac.

aviks commented 3 years ago

For headless Java, set java.awt.headless=true

mkitti commented 3 years ago

If you need to install Java, then use something like https://github.com/actions/setup-java

RainerHeintzmann commented 3 years ago

Thanks. This is in my ci.yml file in .github\workflows, but it seems to have no positive effect on the tests.

    steps:
      - uses: actions/checkout@v2
      - uses: actions/setup-java@v2
        with:
          distribution: 'adopt' # See 'Supported distributions' for available options
          java-version: '11'

mkitti commented 3 years ago

The issue is that JavaCall.jl does not know where adopt put Java, so you need to set the JAVA_HOME environmental variable where it got installed.

RainerHeintzmann commented 3 years ago

The problem seems to be the X-Windows system under Linux. GitHub actions works pretty well. Even the installation of the x-windows system seems to run through but it always stops when trying to connect to DISPLAY=:0.0 or any other number I tried. I disabled the tests on Linux and Mac for now, since I couldn't get it to work. This was the code, which did not work when linux and mac were enabled: https://github.com/RainerHeintzmann/View5D.jl/blob/master/.github/workflows/ci.yml

mkitti commented 3 years ago

For headless Java, set java.awt.headless=true

Did you try aviks's comment above?

Concretely,

using JavaCall
JavaCall.addOpts("-Djava.awt.headless=true")
# Configure classpath here
JavaCall.init()

mkitti commented 3 years ago

More details on that here: https://www.oracle.com/technical-resources/articles/javase/headless.html

RainerHeintzmann commented 3 years ago

Thanks for the hint! This causes other problems, but they are possibly fixable in my Java code. The headless mode is understandably unhappy to report about its screensize:

Exception in thread "main" java.awt.HeadlessException
        at sun.awt.HeadlessToolkit.getScreenSize(Unknown Source)
        at view5d.View5D.Start5DViewer(View5D.java:532)
start viewers: Error During Test at C:\Users\pi96doc\Documents\Programming\Julia\View5D.jl\test\runtests.jl:9
  Got exception outside of a @test
  JavaCall.JavaCallError("Error calling Java: java.awt.HeadlessException")

I will fix this...

RainerHeintzmann commented 3 years ago

Bad news: Even a simple new of a java.applet.Applet object causes the same exception. Rewinding ....

mkitti commented 3 years ago

Aren't applets deprecated as of Java 9? I'm not really sure if there is any way to still use them in a modern browser.

https://www.oracle.com/technetwork/java/javase/migratingfromapplets-2872444.pdf

mkitti commented 3 years ago

I see View5D extends java.applet.Applet so I could see how that might be a complicated to extricate.

RainerHeintzmann commented 3 years ago

Thanks for the hint. I wasn't really aware of this deprecation. So I changed the main class from extending Applet to Container and fixed a few related issues and got it working again, but then the next problem pops up. Looks a bit like a never-ending story with headless graphics:

Exception in thread "main" java.awt.HeadlessException
        at java.awt.GraphicsEnvironment.checkHeadless(Unknown Source)
        at java.awt.Label.<init>(Unknown Source)
        at java.awt.Label.<init>(Unknown Source)
        at view5d.ImgPanel.<init>(ImgPanel.java:199)
        at view5d.View5D.initLayout(View5D.java:867)
        at view5d.View5D.Prepare5DViewer(View5D.java:750)
        at view5d.View5D.Start5DViewer(View5D.java:518)
        at view5d.View5D.Start5DViewer(View5D.java:544)

mkitti commented 3 years ago

See https://www.oracle.com/technical-resources/articles/javase/headless.html

For Linux, this can create a virtual X server: https://github.com/marketplace/actions/gabrielbb-xvfb-action

JuliaInterop / JavaCall.jl

Avoiding Exceptions during automatic testing #143