Foivos Zakkak

> But it might make debugging easier if the files were different so that we can distinguish between those Quarkus produces and those that come from the native image agent...

> Do we prefer minimal filtering that we now for sure Quarkus takes care of? Or do we prefer aggressive filtering? I think I would probably go for minimal because...

Great work, @galderz. I like the new approach even better :)

## Reproducer ``` git clone https://github.com/vincejv/qbittorrent-exporter cd qbittorrent-exporter git worktree add ../qbittorrent-exporter-3.6 ./gradlew build -Dquarkus.package.type=native cd ../qbittorrent-exporter-3.6 sed -i 's/3\.7\.1/3.6.7/g' gradle.properties ./gradlew build -Dquarkus.package.type=native -D ``` ## Build output ###...

> > The binary size increases from 52.95MB to 73.89MB (3.6.7 vs 3.7.1) > > This is pretty bad if you ask me :) @geoand this is not the default...

If I am not mistaken 2.16 is using GraalVM/Mandrel 22.3 3.2 is using GraalVM/Mandrel 23.0 3.7 is using GraalVM/Mandrel 23.1 In each of these updates we know that GraalVM/Mandrel increased...

> @gsmet maybe you can add the graalVM version you used for each entry. Assuming that @gsmet used the default builder image of each version the table with the Mandrel/GraalVM...

> to also add it as a step summary (which allows to visualize things right away instead of downloading the file) That would be a json file per integration-test, wouldn't...

This is blocked by https://github.com/Karm/collector/issues/27

Unable to reproduce locally with latest `main` (a9f0f54e3665dba51b26a8190323f45f65c3416e) so far... Also not able to reproduce locally with 4d68cc6d36e7a0f1d25cfb3b618a4ecccb7674ba which fails in https://github.com/graalvm/mandrel/actions/runs/8761583406/job/24048892524 :/