Closed tetromino closed 2 weeks ago
Can you specify skip_in_bazel_downstream_pipeline: "<reason>"
for the task?
I see; and this is documented in https://github.com/bazelbuild/continuous-integration/blob/master/docs/downstream-testing.md#downstream-project-policies
Thanks for the explanation - I think that https://github.com/bazelbuild/stardoc/pull/244 will fix it
Originally reported at https://github.com/bazelbuild/stardoc/issues/242
For some input .bzl files, Stardoc is expected to produce different output depending on the version of Bazel it runs under (and sometimes - depending on the build flags passed to Bazel). Therefore, it comes with a number of golden tests which are tagged
"manual"
and which are expected to pass only under a specific version of Bazel with a specific set of flags. Examples:stardoc_self_gen_test
,proto_format_test
,multiple_files_noenable_bzlmod_test
,same_level_file_noenable_bzlmod_test
,macro_kwargs_legacy_test
, etc. in https://github.com/bazelbuild/stardoc/blob/da374a508b5a5f82e1e00f94889f04f9f86ca1fe/test/BUILDStardoc has tasks configured in its presubmit.yml to run these tests with the right versions of Bazel and the right flags. Such tasks have descriptive names such as "Stardoc golden tests requiring Bazel 7.2" and an explicitly set
bazel
version. Example: https://github.com/bazelbuild/stardoc/blob/da374a508b5a5f82e1e00f94889f04f9f86ca1fe/.bazelci/presubmit.yml#L116Unfortunately, bazel-at-head-plus-downstream pipeline attempts to run these tasks with Bazel at HEAD, and of course fails.
Possible solutions:
manual
CC @sgowroji