This only affects pull requests for the `Bazel RBE Build Tests` job. The
equivalent master CI job will still build all end2end tests, including
experiments.
Built upon @Vignesh2208 's work in #33156
This adds ref counting to the http_proxy fixture object, fixing test
flakes identified by the introduction of EventEngine listeners. Proxy
objects were either being deleted twice, or sometimes not at all,
resulting in two different sorts of flakes.
The `bazel_build_with_strict_warnings_linux` job was taking nearly an
hour to complete. This change brings it down to ~20m by splitting up the
build targets into 7 separate RBE actions.
---------
Co-authored-by: drfloob <drfloob@users.noreply.github.com>
This reduces clang-format duration to under 10s on my 128 core machine.
Previously I stopped the process at 10 minutes. Major changes:
Eliminates 18,000 files
```
find ... \
-and -not -path '*/cmake/build/*' \
-and -not -path '*/huffman_geometries/*' \
```
Also, this now runs `$CPU_COUNT` number of parallel clang-format
processes with a roughly equal number of files. Previously, this script
was formatting one file at a time in separate processes (`xargs -n 1`).
Not all POSIX platforms have rt as a separate library. QNX has rt as
part of libc (same as pthreads).
Add condition to check if the library can be found and link with it only
in positive scenario.
A solution for: https://github.com/grpc/grpc/issues/34254
Add the flag `enable_csm_observability` to the c++ PSM interop testing
image, such that when enabled from the PSM interop testing framework,
the C++ client/server app will enable the CSM Observability plugin.
Add a `PodMonitoring` resource type to the PSM interop testing
framework. This is needed so that GMP (Google Managed Prometheus) can
scrape the matching GKE pods Prometheus endpoint for Prometheus metrics.
`succeed-on-retry-attempt-<int>` is what is being currently used in the
Java server implementation and also by the test client. The spec was
probably written after the stuff was implemented and this typo is better
fixed before other languages implement their server side logic. Go is in
the process of doing so.
Relands #34785, which was reverted in #34818.
The first commit is the revert. The second commit removes the gtest
dependency from the xds_server library, which should address the
testonly problem internally.
Returning a reference to a GUARDED_BY variable while only holding the
lock in the getter lets callers access the variable without holding the
lock.
See https://github.com/llvm/llvm-project/pull/67776.
This is only used for testing, so just return a copy for simplicity.
<!--
If you know who should review your pull request, please assign it to
that
person, otherwise the pull request would get assigned randomly.
If your pull request is for a specific language, please add the
appropriate
lang label.
-->
As such, `alts_zero_copy_grpc_protector_create` will take a
`GsecKeyFactoryInterface` to create `GsecKeyInterface` objects for the
underlying crypter to use.
This enables the caller to control how all the key related buffers are
prepared and protected.
`gsec_aes_gcm_aead_crypter` holds the raw pointer to `GsecKeyInterface`
instead of a `unique_ptr` possibly because somewhere in the test (and
maybe production code as well), the structure is getting copied. A SEGV
error would be caused with `unique_ptr` which doesn't support copy
operations.
The frame fuzzer constructs random bytes and frame header to test
serialization and deserialization. There are error cases where the frame
header isn't constructed correctly and isn't align with the serialized
bytes. Added checks in frame header parser and initialize frame header's
flags and header/trailer length in serialization.
<!--
If you know who should review your pull request, please assign it to
that
person, otherwise the pull request would get assigned randomly.
If your pull request is for a specific language, please add the
appropriate
lang label.
-->
Fixes two issue with the `event_engine_listener` experiment:
* a race in `on_connect` with the read/write path
* a long delay due to pollset_work being effectively a `sleep` in the
EventEngine shim implementation
<!--
If you know who should review your pull request, please assign it to
that
person, otherwise the pull request would get assigned randomly.
If your pull request is for a specific language, please add the
appropriate
lang label.
-->
Currently it is very easy to use the `TlsCredentialsOptions` in such a
way that it produces a memory leak. For example, the code block
```
{
TlsCredentialsOptions options;
}
```
produces a memory leak. This PR fixes up the ownership bugs in this
class and its `grpc_tls_credentials_options`, the C-core analogue.
Just the right combination of timeout values could lead to a large
amount of growth in the timeout cache, leading to high memory usage and
high cpu utilization on the client attempting that encoding.
Fix: cap the number of cached timeout values we'll consider to a very
small number (I'm choosing five).
Add a fuzzer that ensures that we're actually respecting the limits of
inaccuracy we're imposing upon ourselves.
---------
Co-authored-by: ctiller <ctiller@users.noreply.github.com>
In case of test fails, the clean up script will try delete some resource
we didn't create and resulting lots of 404 errors, we should exclude
those status code since we have specific handling for 404.
<!--
If you know who should review your pull request, please assign it to
that
person, otherwise the pull request would get assigned randomly.
If your pull request is for a specific language, please add the
appropriate
lang label.
-->
The original logic from #34615 was incorrect in cases where one address
family has a different number of addresses than the other(s).
Fixes b/307937051.
Also break the filter stack and promise based versions apart so that I
can re-understand this code.
---------
Co-authored-by: ctiller <ctiller@users.noreply.github.com>
- add entry for aggregate and logical DNS cluster support
- add entry for least_request LB policy
- add a bunch of missing versions in various languages
This fixes a bug which could happen if multiple WorkStealingThreadPools
existed (wstpA and wstpB), and a thread in wstpA called
`wstpB->Run(closure)`. Previously, this would have scheduled the closure
on the current wstpA thread worker's local queue.
`bm_thread_pool` results look unchanged on RBE.
Ben Beasley
Stanley Cheung
AJ Heller
Mark D. Roth
Vignesh Babu
Roman-Byshliaha-Bose
Brad Hess
paskozdilar
Easwar Swaminathan
Arthur Eubanks
Arvind Bright
Luwei Ge
Yash Tibrewal
Craig Tiller
Alisha Nanda
nanahpang
Xuan Wang
Matthew Stevenson
Kevin Cooney