Bazel 8 no longer supports legacy toolchains, so all of our release toolchains have been updated to use platforms.
#test-continuous
PiperOrigin-RevId: 870017275
This docker image includes ruby gems v4.0.5, which includes https://github.com/ruby/rubygems/pull/9245 that fixes building multiple native extensions.
PiperOrigin-RevId: 863399053
The pyi generator now includes Kythe annotations for:
* Extension field constants (e.g., `EXTENSION_FIELD`).
* Field number constants (e.g., `STRING_FIELD_FIELD_NUMBER`).
* `Create` methods in generated Stubby client classes.
These annotations allow the Python indexer to link these generated symbols back to their definitions in the `.proto` files.
PiperOrigin-RevId: 852971617
These are by far our slowest tests due to their QEMU use. They rarely catch unique issues, so we can handle this in post-submits instead of paying ~4 minutes on every presubmit
PiperOrigin-RevId: 818873557
These tests are currently flaky but should be fixed soon by a new image
rollout.
I plan on rolling back this change in a couple of days once the issue is fixed.
PiperOrigin-RevId: 720598019
Migrate all tests to run on bazel 7 and fix errors that came up in the process. 30.x will no longer guarantee support for bazel 6.
#test-continuous
PiperOrigin-RevId: 703590770
This ensures the names of these jobs will properly be prefixed with "[SKIPPED] (Continuous)" and will show as passed instead of skipped.
PiperOrigin-RevId: 666918220
Before this PR, we stored a list internally of tests that must pass on presubmit and tried to keep it up to date.
This PR moves that information keeping into GitHub by adding a 'continuous-only' variable to most testing matrices to allow authors to specify which of their tests should be skipped on presubmit. During presubmit, tests that were specified to not run on presubmit will not be run and their names will be prefixed with "[SKIPPED]". All continuous only tests will be suffixed with "(Continuous)".
At the end of running all the tests, we have a single "All Blocking Tests" signal that will tell us whether all of the necessary tests have passed (either for presubmit or continuous based on how the test was triggered).
I've tested this from a different branch [here](https://github.com/protocolbuffers/protobuf/actions/runs/9602443750?pr=17151) and from a different fork [here](https://github.com/protocolbuffers/protobuf/actions/runs/9602554500?pr=17192). These should be the same and are as far as I can tell.
I also have a continuous test run [here](https://github.com/protocolbuffers/protobuf/actions/runs/9603824200) which runs the entire test suite.
Closes#17198
PiperOrigin-RevId: 662940724
This will allow them to reuse our bazelrc and remote caching setup. This also silences the non-bzlmod windows test that's hitting the windows path length.
PiperOrigin-RevId: 626390416
Latest bundler 2.5.0 release results in the following error: `The last version of bundler (>= 0) to support your Ruby & RubyGems was 2.4.22. Try installing it with gem install bundler -v 2.4.22 bundler requires Ruby version >= 3.0.0. The current ruby version is 2.7.3.183.`
PiperOrigin-RevId: 591288515
This will retry up to 3 times if we hit networks flakes updating our submodules. It will also allow us to easily inject other stability fixes to this step in the future.
PiperOrigin-RevId: 568306356
This removes the need for us to manually rebuild these amalgamation files. I
added the "manual" tag to the corresponding staleness tests so that we won't
get presubmit errors if the amalgamation files are out of date. I also had to
tweak the Ruby BUILD.bazel file to make sure it uses Bazel-generated
amalgamations instead of relying on the checked-in files.
I added deliberate errors to the Ruby and PHP amalgamation files to test that
everything is working. We should not see any CI errors, and the files should
quickly be auto-updated after the change lands on main.
PiperOrigin-RevId: 561187352
https://github.com/protocolbuffers/protobuf/pull/13204 refactored the Ruby object cache to use a key of `LL2NUM(key_val)` instead of `LL2NUM(key_val >> 2)`. On 32-bit systems, `LL2NUM(key_val)` returns inconsistent results because a large value has to be stored as a Bignum on the heap. This causes cache lookups to fail.
This commit restores the previous behavior of using `ObjectCache_GetKey`, which discards the lower 2 bits, which are zero. This enables a key to be stored as a Fixnum on both 32 and 64-bit platforms.
As https://patshaughnessy.net/2014/1/9/how-big-is-a-bignum describes, a Fixnum uses:
* 1 bit for the `FIXNUM_FLAG`.
* 1 bit for the sign flag.
Therefore the largest possible Fixnum value on a 64-bit value is 4611686018427387903 (2^62 - 1). On a 32-bit system, the largest value is 1073741823 (2^30 - 1).
For example, a possible VALUE pointer address on a 32-bit system:
0xff5b4af8 => 4284173048
Dropping the lower 2 bits makes up for the loss of range to these flags. In the example above, we see that shifting by 2 turns the value into a 30-bit number, which can be represented as a Fixnum:
(0xff5b4af8 >> 2) => 1071043262
This bug can also manifest on a 64-bit system if the upper bits are 0xff.
Closes#13481Closes#13494
COPYBARA_INTEGRATE_REVIEW=https://github.com/protocolbuffers/protobuf/pull/13494 from stanhu:sh-fix-ruby-protobuf-32bit d63122a6fc
PiperOrigin-RevId: 557211768
