jdk10

• Path: arch/private/jdk10
• Oracle/Sun Java 10 (10.0.2), proprietary/EOL — AUR jdk10 PKGBUILD (maintainer Chris Severance), added here 2026-08-27 from the user’s own local jdk10.tgz (a full prior AUR git checkout: PKGBUILD, .install, desktop files, and the actual proprietary source tarball — the .git/ dir itself was dropped, not committed here, this project never nests a real git repo inside the arch/ submodule tree).
• Source rehosted to our own src/ mirror (jdk-10.0.2_linux-x64_bin.tar.gz, 355MB) — was https://download.oracle.com/otn-pub/java/jdk/10.0.2+13/<hash>/jdk-10.0.2_linux-x64_bin.tar.gz (kept commented above the new source= entry), which needs an Oracle account + license click-through now, not a direct download. Verified byte-identical (sha256 matches the existing PKGBUILD entry) before uploading.
• Simplified away the whole manual:// download dance the original PKGBUILD used to work around that click-through: a custom DLAGENTS+=("manual::...") entry running readme.sh (opens the Oracle page in a browser, expects the user to have already saved the file), plus an XDG_DOWNLOAD_DIR “scoop the file from ~/Downloads if it’s there” fallback block. Removed all of it along with the readme.sh source entry now that source= points straight at a plain https URL — normal makepkg -s (or -g/updpkgsums) works again. Also dropped the paired DLAGENTS=("${DLAGENTS[@]//curl -/curl -b 'oraclelicense=a' -}") / -gqb "" hacks (only existed to inject the Oracle license cookie, irrelevant once nothing hits Oracle’s servers) and the now-unused _build/_hash variables that only fed the old Oracle URL.
  • Caught a real bug while doing this: removing readme.sh from source=() initially left its old trailing md5sums/sha256sums entries dangling (5 checksums for 4 sources) — arrays are positional, this would have silently mis-paired every checksum after that point if the extra pair hadn’t been dropped too. Verified the final 4-source/4-checksum alignment by re-hashing each local file directly before trusting .SRCINFO.
• First publish 2026-08-27: x86_64 built+signed+published cleanly at the AUR’s existing pkgrel=1 (first time in archlinuxaba, no pkgrel bump needed, same reasoning as [[jdk6]]).
• arch=('x86_64') only (no i686/arm variants ever existed for JDK10) — repo_build.sh’s remote-build fallback never attempted, correctly.