[windows/arm64] Add missing JDK toolchain for java build #14700
Conversation
|
@meteorcloudy and @philwo - Can you please review this change? This would be needed for the java builds. |
|
Huh, not sure why the mirror URL is giving HTTP 404... let me check. Edit: Fixed! I'll rerun the failed CI jobs. |
Thanks! Looks like the failing job also has trouble downloading JDK package for darwin from Bazel mirror. If this is already fixed or just an intermittent issue, could you retrigger the build please ? |
|
@nsait-linaro I'm really sorry, I don't know why this test is suddenly failing with this error. The URL seems fine and I can download it on my own machine. I'll retry it later today, maybe it's some really weird kind of network flakyness between RBE and the remote server? I don't see anything in your PR that would cause it. @meteorcloudy Any idea? |
|
Hmm, it does seem to be flaky in another presubmit: https://buildkite.com/bazel/bazel-bazel-github-presubmit/builds/10200#f7c14b0e-60b7-4f4f-9440-e78c19da31ef But not sure what caused it.. |
Thanks, @philwo. Could you do one more retrigger please ? |
I noticed in this in one of my PRs before. At that PR it looked like we're hitting some kind of traffic limiter - one url in the list consistently failed remotely (i'm only assuming it was the last one), and the same url didn't fail locally. The issue was resolved, because somebody else did a cleanup, removing a couple of remote repos. |
|
I'll submit a change to make this test easier to debug, as we currently have no idea or log output why it's failing. :/ @comius That's a good idea. If it's indeed a traffic limit, maybe RBE team can help us out, as this is only failing on RBE, so it must be something specific to the environment. |
|
@nsait-linaro Could you please rebase this PR against master to pick-up 938e209? I'm curious if this will tell us why the test is failing. |
niyas-sait
force-pushed
the
fix_jdk_win_arm64
branch
from
725d08f
to
401ae7d
Compare
Looks like still timing out |
|
Nice, at least now we get an error message. I'll ask the RBE folks for help. |
|
I filed a bug asking for help, will follow-up as soon as I hear back! |
@philwo Is CI issue specific to this patch ? I just retriggered and still seems to have the same issue |
|
@nsait-linaro Yes, because this PR adds more URLs, which I think happen to triggered some traffic limit. |
|
While the RBE team is looking into this (which looks like going to take some time), one mitigation I can think of is to deduplicate the urls we parsed from files: |
|
@meteorcloudy Yes, I also looked into this and one easy win would be to update all our other OpenJDKs to the latest version - we have at least one case where we have two separate versions for the same platform in the file. This should save some URLs 😅 Deduping the list would be a good improvement for the test anyway, great idea! |
@nsait-linaro Do you mind trying to fix the verify_workspace test? I'm quite overloaded currently. |
Yes of course! |
Hopefully, #14763 should fix! |
niyas-sait
force-pushed
the
fix_jdk_win_arm64
branch
from
ef44a2e
to
95c5360
Compare
|
Thanks, @meteorcloudy, and @philwo. Looking for the per-commit build for validation, Is it still at the same location? I've tried with the latest commit but cannot find anything there. |
|
@nsait-linaro Thanks for the contribution! |
|
Thanks, @meteorcloudy, I can confirm that java and c++ builds are working fine now. It will be nice to get this delivered with the next Bazel release with the |
|
@bazel-io fork 5.1 |
|
@nsait-linaro That's nice! Can you send a PR to cherry-pick Windows ARM64 related changes to Bazel 5.1? I have opened issues to track the progress.
Yes, the official release will of course come with ".exe" suffix. |
Extend configurations to add JDK 11 and 17 for windows/arm64 platforms. This should fix the Java builds on windows/arm64 Closes bazelbuild#14700. PiperOrigin-RevId: 427737536
bazelbuild#14700 added couple more URLs to fetch JDK package and seems to be causing some infrastructure as discussed in bazelbuild#14700. This patch workaround the issue by removing the duplicated URLs and reduce the crawl request. Closes bazelbuild#14763. PiperOrigin-RevId: 427464876
* Enable native support for Windows on arm64 (Part 1) Contains following changes to third_party: - Extended def_parser to handle ARM64 binaries - Add grpc patch to workaround build issues Closes: #14689 Partial commit for third_party/*, see #14689. Signed-off-by: Yun Peng <[email protected]> * Enable native support for Windows on arm64 This PR will enable cross-compilation of Bazel binaries for win/arm64 from win/x64 `bazel build -c opt --cpu=x64_arm64_windows //src:bazel.exe` Generated bazel executable can be used for native compilation in win/arm64 `bazel.exe build //main:hello-world` Following changes are included - Add win/arm64 JDK 17 - Fix AutoCpuConverter.java to identify win/arm64 CPU - Extend build_bazel_binaries.yml to cross-compile for win/arm64 - Fix msvc toolchain to look for tools in HostX86 directory as well - add clang-cl support for windows/arm64 host - Extend host_windows config to handle windows x64 and arm64 hosts. Closes #14340. PiperOrigin-RevId: 425919351 * [windows/arm64] Add missing JDK toolchain for java build Extend configurations to add JDK 11 and 17 for windows/arm64 platforms. This should fix the Java builds on windows/arm64 Closes #14700. PiperOrigin-RevId: 427737536 * add missing openjdk11_windows_arm64_archive * Deduplicate urls parsed to reduce crawl requests #14700 added couple more URLs to fetch JDK package and seems to be causing some infrastructure as discussed in #14700. This patch workaround the issue by removing the duplicated URLs and reduce the crawl request. Closes #14763. PiperOrigin-RevId: 427464876 * fix jdk_http_archives.tmpl
Extend configurations to add JDK 11 and 17 for windows/arm64 platforms.
This should fix the Java builds on windows/arm64