The self-deployment of PR#61 failed on its first real attempt, at the
one check that exists for it:
version `GLIBC_2.38' not found (required by …/jre/lib/libjli.so)
ERROR: the staged runtime does not run on this host
jlink cuts the shipped runtime out of the toolchain's own kit, native
libraries included, so the kit's build platform decides where that
runtime can start. Both kits in play were Java 21, which is why the
version number gave no warning: Debian trixie's OpenJDK in the build
sandbox needs glibc 2.38, the production host has 2.36, and Temurin on
the developer machine needs 2.14 — which is why deployments from there
always worked and hid the problem.
The toolchain now names Temurin as its vendor, and the foojay resolver
lets Gradle fetch it where none is installed. Measured on the runtime
this branch produces: no native library asks for more than GLIBC_2.15.
A kit baked into the rootfs would fix today's failure too. It was
rejected because it moves the Java version onto the host: every upgrade
would become a gigabyte of rootfs upload, which is the manual step the
self-deployment exists to remove.
Nothing on the host was damaged by the failed attempt. The staged swap
stopped before moving anything, the service stayed up, and the transient
restart unit was never scheduled.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Two scenarios stopped being expectations. The CI build of b9203be
produced a runtime reporting Temurin-21.0.12.1, which the sandbox had to
fetch because its rootfs carries Debian's kit and no other. That runtime
was then started on the production host out of the build worktree,
without installing it, and printed its version.
So the toolchain download works in the sandbox and the result runs where
it has to. What remains unproven is the deployment as a whole, which
this branch cannot show: deploy has never completed on the host, and the
first merge after this one is what will run it.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
mi
merged commit 99647ea154 into main2026-09-15 09:49:53 +02:00
mi
deleted branch adoptium-toolchain2026-09-15 09:49:53 +02:00
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
The self-deployment of PR#61 failed on its first real attempt, at the one check that exists for it: version `GLIBC_2.38' not found (required by …/jre/lib/libjli.so) ERROR: the staged runtime does not run on this host jlink cuts the shipped runtime out of the toolchain's own kit, native libraries included, so the kit's build platform decides where that runtime can start. Both kits in play were Java 21, which is why the version number gave no warning: Debian trixie's OpenJDK in the build sandbox needs glibc 2.38, the production host has 2.36, and Temurin on the developer machine needs 2.14 — which is why deployments from there always worked and hid the problem. The toolchain now names Temurin as its vendor, and the foojay resolver lets Gradle fetch it where none is installed. Measured on the runtime this branch produces: no native library asks for more than GLIBC_2.15. A kit baked into the rootfs would fix today's failure too. It was rejected because it moves the Java version onto the host: every upgrade would become a gigabyte of rootfs upload, which is the manual step the self-deployment exists to remove. Nothing on the host was damaged by the failed attempt. The staged swap stopped before moving anything, the service stayed up, and the transient restart unit was never scheduled. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>