Repository navigation
Fix tf-text OSS nightly and presubmit builds against current tf-nightly. - #1571
Open
copybara-service[bot] wants to merge 1 commit into
Open
copybara-service[bot] wants to merge 1 commit into
copybara-service[bot] wants to merge 1 commit into
Conversation
copybara-service
Bot
force-pushed
the
test_1000522266
branch
from
October 16, 2026 01:57
da28014 to
767c8a0
Compare
copybara-service
Bot
changed the title
1. When `IS_NIGHTLY=nightly`, `oss_scripts/configure.sh` runs `bazel run //oss_scripts/pip_package:requirements.update -- --upgrade` (`pip-compile`) on `@release_or_nightly//:requirements.in`. Because `tf-keras-nightly` on PyPI currently declares `Requires-Dist: tf-nightly~=2.21.0.dev` (with no other dependencies), `pip-compile` downgrades `tf-nightly` from `2.23.0.dev` back to `2.21.0.dev20260203` (Feb 2026). Subsequent `prepare_tf_dep.sh` then pins `@org_tensorflow` in `WORKSPACE` to that Feb 2026 commit, which predates `@rules_cc` in `tf_workspace3()` and breaks the build with `Failed to load Starlark extension '@@rules_cc//cc:extensions.bzl'`. - Exclude `tf-keras` from `@release_or_nightly//:requirements.in` in `py_deps_profile.bzl` (while keeping the `@release_or_nightly//:tf_keras_pkg` target alias and adding `IS_NIGHTLY` to `environ`). - Persist `HERMETIC_PYTHON_VERSION` and `IS_NIGHTLY` via `build --repo_env` in `oss_scripts/configure.sh` and append the latest `tf-keras-nightly` pin and sha256 hash from PyPI directly to `requirements_lock_<py>.txt` after `requirements.update` so `pip_parse` still populates `@pypi_tf_keras_nightly`. 2. Add `common --features=-external_include_paths` and `common --host_features=-external_include_paths` in `oss_scripts/configure.sh` (matching TensorFlow `master`'s `.bazelrc`) so TFLite targets (`@@org_tensorflow//tensorflow/compiler/mlir/lite/schema:schema_utils`) build cleanly with the updated `rules_ml_toolchain` when `prepare_tf_dep.sh` pins `WORKSPACE` to a newer TensorFlow commit than `v2.22.0-rc0`. 3. Mark `tokenizers_benchmarks` and `ops_benchmarks` with `tags = ["manual"]` so `bazel test tensorflow_text:all` does not run them without `--benchmark_filter` and fail with `NO TESTS RAN` (exit code 5) under Python 3.12+. PiperOrigin-RevId: 1000522266
copybara-service
Bot
force-pushed
the
test_1000522266
branch
from
October 16, 2026 02:24
767c8a0 to
7d27935
Compare
This branch has not been deployed
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Fix tf-text OSS nightly and presubmit builds against current tf-nightly.
IS_NIGHTLY=nightly,oss_scripts/configure.shrunsbazel run //oss_scripts/pip_package:requirements.update -- --upgrade(pip-compile) on@release_or_nightly//:requirements.in. Becausetf-keras-nightlyon PyPI currently declaresRequires-Dist: tf-nightly~=2.21.0.dev(with no other dependencies),pip-compiledowngradestf-nightlyfrom2.23.0.devback to2.21.0.dev20260203(Feb 2026). Subsequentprepare_tf_dep.shthen pins@org_tensorflowinWORKSPACEto that Feb 2026 commit, which predates@rules_ccintf_workspace3()and breaks the build withFailed to load Starlark extension '@@rules_cc//cc:extensions.bzl'.tf-kerasfrom@release_or_nightly//:requirements.ininpy_deps_profile.bzl(while keeping the@release_or_nightly//:tf_keras_pkgtarget alias and addingIS_NIGHTLYtoenviron).HERMETIC_PYTHON_VERSIONandIS_NIGHTLYviabuild --repo_envinoss_scripts/configure.shand append the latesttf-keras-nightlypin and sha256 hash from PyPI directly torequirements_lock_<py>.txtafterrequirements.updatesopip_parsestill populates@pypi_tf_keras_nightly.common --features=-external_include_pathsandcommon --host_features=-external_include_pathsinoss_scripts/configure.sh(matching TensorFlowmaster's.bazelrc) so TFLite targets (@@org_tensorflow//tensorflow/compiler/mlir/lite/schema:schema_utils) build cleanly with the updatedrules_ml_toolchainwhenprepare_tf_dep.shpinsWORKSPACEto a newer TensorFlow commit thanv2.22.0-rc0.tokenizers_benchmarksandops_benchmarkswithtags = ["manual"]sobazel test tensorflow_text:alldoes not run them without--benchmark_filterand fail withNO TESTS RAN(exit code 5) under Python 3.12+.