Update data-plane-api (core and filter packages)#2495
Update data-plane-api (core and filter packages)#2495htuch merged 17 commits intoenvoyproxy:masterfrom
Conversation
Signed-off-by: Kuat Yessenov <kuat@google.com>
Signed-off-by: Kuat Yessenov <kuat@google.com>
Signed-off-by: Kuat Yessenov <kuat@google.com>
Signed-off-by: Kuat Yessenov <kuat@google.com>
Signed-off-by: Kuat Yessenov <kuat@google.com>
Signed-off-by: Kuat Yessenov <kuat@google.com>
Signed-off-by: Kuat Yessenov <kuat@google.com>
Signed-off-by: Kuat Yessenov <kuat@google.com>
Signed-off-by: Kuat Yessenov <kuat@google.com>
| GrpcMuxImpl::~GrpcMuxImpl() { | ||
| // Hack to force linking of the service: https://github.com/google/protobuf/issues/4221 | ||
| envoy::service::discovery::v2::AdsDummy dummy; | ||
| envoy::service::ratelimit::v2::RateLimitRequest rls_dummy; |
There was a problem hiding this comment.
We shouldn't need rls_dummy here, right? I don't think it makes sense to include rls.pb.h here. And how come this wasn't an issue previously in general?
There was a problem hiding this comment.
I've discovered this by running envoy-static with ADS. It is missing ADS descriptor without this change.
There was a problem hiding this comment.
OK, can you remove rls_dummy from here and the rls.pb.h include? It should not be needed in this file.
There was a problem hiding this comment.
i am confused here. Why is AdsDummy needed here now? It was able to link properly earlier and why the new package need this? I ran in to the same issue with latest source, that it is not able to find the 'MethodDescriptor' for ADS and bazel clean seems to have resolved that. I may be wrong but Can you double check on this once please?
There was a problem hiding this comment.
it links correctly but if you try to call ADS methods, the reflection lookup fails since proto descriptor is missing. there's another PR that does a similar fix, so it is happening consistently. adding a dummy forces linking the proto descriptor.
| // services as well as types that are referenced in Any messages. IF THIS TEST BREAKS YOU | ||
| // HAVE DONE SOMETHING BAD. Consult with the larger dev team on how to handle. | ||
| TEST(ProtoDescriptorTest, BackCompat) { | ||
| // Hack to force linking of the service: https://github.com/google/protobuf/issues/4221 |
There was a problem hiding this comment.
I'm confused. Without linking to grpc_mux stuff, how does this work now?
There was a problem hiding this comment.
You are right, I need it in two places :(
Signed-off-by: Kuat Yessenov <kuat@google.com>
Signed-off-by: Kuat Yessenov <kuat@google.com>
| } | ||
|
|
||
| GrpcMuxImpl::~GrpcMuxImpl() { | ||
| // Hack to force linking of the service: https://github.com/google/protobuf/issues/4221 |
There was a problem hiding this comment.
Sorry, can you add a TODO here to fix this so we can search for it? Also please add a TODO to my new test which I didn't do. Otherwise LGTM.
|
I'd like to re-run mac build (flaked out). Is there some button in circle to do that? |
|
@kyessenov I just kicked it, but typically we will marge without OSX passing if it looks like a flake. |
|
@kyessenov sorry you lost the merge race. Can you merge master? |
|
Merged and fixed. |
| GrpcMuxImpl::GrpcMuxImpl(const envoy::api::v2::core::Node& node, Grpc::AsyncClientPtr async_client, | ||
| Event::Dispatcher& dispatcher, | ||
| const Protobuf::MethodDescriptor& service_method) | ||
| : node_(node), async_client_(std::move(async_client)), service_method_(service_method) { |
There was a problem hiding this comment.
should we add some kind of an ASSERT here on service_method_? with out that if the MethodByName is not found in generic pool, it just segfaults.
There was a problem hiding this comment.
Merged to keep things moving, tracking this concern at #2528.
htuch
left a comment
There was a problem hiding this comment.
Agree with @ramaraochavali and @mattklein123 that we should make the proto descriptor dynamic lookups more bullet proof, let's merge this one to avoid further drift and we can iterate.
Signed-off-by: Kuat Yessenov <kuat@google.com>
Description: Update the documentation of `find_class` method as it seems to accept `/` and `.` and using `/` should be less confusing as this is what `env->findClass` expects. Risk Level: None Testing: Manual confirmed that the current implementing of tagging works and it does uses `/` as opposed to `.`. Docs Changes: N/A Release Notes: N/A Signed-off-by: Rafał Augustyniak <Augustyniak@users.noreply.github.com> Signed-off-by: JP Simard <jp@jpsim.com>
Description: Update the documentation of `find_class` method as it seems to accept `/` and `.` and using `/` should be less confusing as this is what `env->findClass` expects. Risk Level: None Testing: Manual confirmed that the current implementing of tagging works and it does uses `/` as opposed to `.`. Docs Changes: N/A Release Notes: N/A Signed-off-by: Rafał Augustyniak <Augustyniak@users.noreply.github.com> Signed-off-by: JP Simard <jp@jpsim.com>
title: update data-plane-api
Description:
Update to changes in data-plane-api
Risk Level: Low
Testing:
bazel test.
Docs Changes:
envoyproxy/data-plane-api#452
Release Notes: