Native Build Fails, when Reusing Existing Executable with Compression Enabled

[gcw-it]

Describe the bug

Building a native executable with <quarkus.native.reuse-existing>true</quarkus.native.reuse-existing> set, and compression enabled, leads to a failing build, when the artifact already exists.

Expected behavior

The plugin skips the compression step for the already compressed executable, and the build is finished without error.

Actual behavior

The build fails with: Build step io.quarkus.deployment.pkg.steps.UpxCompressionBuildStep#compress threw an exception: java.lang.IllegalStateException: Unable to compress the native executable

How to Reproduce?

1. quarkus create
2. add the following properties to the native profile in pom.xml.
   <quarkus.native.container-build>true</quarkus.native.container-build>
<quarkus.native.reuse-existing>true</quarkus.native.reuse-existing>
<quarkus.native.compression.level>1</quarkus.native.compression.level>
3. mvn package -DskipTests -Dnative
4. When running the build a 2nd time (mvn package -DskipTests -Dnative) the build fails, because the plugin tries to recompress the already compressed executable.

Output of uname -a or ver

Darwin 23.5.0 Darwin Kernel Version 23.5.0: Wed May 1 20:09:52 PDT 2024; root:xnu-10063.121.3~5/RELEASE_X86_64 x86_64

Output of java -version

java 22.0.1 2024-04-16

Mandrel or GraalVM version (if different from Java)

quay.io/quarkus/ubi-quarkus-mandrel-builder-image:jdk-21

Quarkus version or git rev

3.11.3

Build tool (ie. output of mvnw --version or gradlew --version)

Apache Maven 3.9.8 (36645f6c9b5079805ea5009217e36f2cffd34256)

Additional information

No response

[quarkus-bot[bot]]

/cc @Karm (mandrel), @galderz (mandrel), @zakkak (mandrel,native-image)

[galderz]

@gcw-it Fancy trying to implement this yourself? UpxCompressionBuildStep could be enhanced to see if the binary is already there (and reuse existing is on) and check if the file is already compressed.

[gcw-it]

@galderz I could tackle this in the next few days, when I find some spare time (Probably on the weekend).

But first and foremost we should come to an understanding, what the expected behaviour should be.

Probably simplest and least surprising would be, to leave an existing artifact alone and skip the compression step.

If we go down the road to compress an existing uncompressed executable, if the compression settings have changed between builds, then we have to consider some other situations:

• What happens, when the file is already compressed and the compression level is changed?
• What happens to a compressed file, if compression is disabled?

In my opinion, the cleanest approach would be to take Reuse Existing Executable literally and leave the existing artifact alone. What are your thoughts?