Spark is a tool for testing game performance, identifying performance bottlenecks and stuttering. When you encounter issues like game lag or high latency, you can use Spark.
After version 0.5.3 (presumably?), spark has been removed from the gto package. (/gtocore spark is a legacy command; it is invalid if spark is not installed.)
spark does not work properly on some kernels, e.g. Danger! MC mod causes crash: the culprit is actually this line of code..._bilibili
used by Sparkasync-profiler is incompatible with some Linux kernels
Closest spark issues are these two:
#512 Crash using spark-1.10.53 (Forge) on Alpine 3.18 with openjdk21-jre-headless This is Forge + 1.20.1 + spark 1.10.53 + JDK 21 a direct native crash. Although it's Alpine/aarch64, not your Debian/amd64, the nature is the same: both are spark's native profiler directly crashes in the host environment.
#535 Spark 1.20.1 Forge-47.4.10 (spark version 1.10.53) incompatible with Java 25 This is Forge 1.20.1 + spark 1.10.53 + Java 25 incompatibility, consistent with your previous phase.
Additionally:
#546 Spark 1.20.1 Profiler doesn't work
#557 Disable spark profiler on linux
#560 Could not find VMThread bridge. Unsupported JVM?
So at least two points can be confirmed:
This is not a unique strange phenomenon for you alone. spark 1.10.53 on 1.20.1, Forge, Linux, and different JDK combinations, there have indeed been reports of profiler-related crashes/unavailability.
Your problem here is not just Java 25 You have already tested:
JDK 25 crashes
JDK 21 also crashes. So now it's more like spark 1.10.53's native profiler is incompatible with the combination of your Linux container environment, not just a Java main version issue.
I also checked async-profiler upstream releases:
Here you can see some very relevant fix directions:
v4.2: Re-implement SafeAccess crash protection
v4.3: Harden crash protection in StackWalker
v4.3: Support compilation on modern JDKs
v4.3: Workaround for the kernel PERF_EVENT_IOC_REFRESH bug
v4.3: Fix races related to VM termination