Chiebot-Mirror/grpc - grpc - Gitea: Git with a cup of tea

Commit Graph

Author	SHA1	Message	Date
Yash Tibrewal	5eeb651a5f	Add extra space for opt_stats as part of the formula as opposed to 1024	6 years ago
Yash Tibrewal	6a372ff442	Use constexpr	6 years ago
Yash Tibrewal	5a44f700bb	Increase allocated space for cmsgs	6 years ago
Vijay Pai	a47c979ba0	Enable TCP callback tests if the event engine allows	6 years ago
Arjun Roy	195a30bb8b	Grpc: Change grpc_handshake and grpc_handshake_mgr to use CPP implementations. grpc_handshake is renamed to GrpcHandshake, using C++ class definitions instead of C-style vtable classes. Update callers to use new interfaces. We use RefCountedPtr to simplify reference tracking.	6 years ago
Alexander Polcyn	93b7acd9bf	Disable service config resolution with c-ares by default	6 years ago
Vijay Pai	f96d630c33	Document ApplicationCallbackExecCtx, update ExecCtx comments	6 years ago
Eric Gribkoff	b90dd36270	add comment	6 years ago
Soheil Hassas Yeganeh	611857accb	Address guantaol@'s comments on Pull #17964	6 years ago
Jan Tattermusch	a832d66b09	upgrade System.Interactive.Async to 3.2.0	6 years ago
Alexander Polcyn	5c85f5a1a0	Ruby: refactor init/shutdown logic to avoid using atexit; fix windows	6 years ago
Eric Gribkoff	34965961cf	check grpc is init before creating execctx	6 years ago
Muxi Yan	40544bb112	Min deployment target change	6 years ago
Yash Tibrewal	30d8f7a626	Memset before setting length	6 years ago
Vijay Pai	85d76b2888	Add ApplicationCallbackExecCtx in other spots that may trigger application work	6 years ago
Yash Tibrewal	cddb5519f2	Add a test only method to reset global interceptor	6 years ago
Vijay Pai	67c010b44f	Add default initialization value	6 years ago
Mark D. Roth	0794507043	Don't pass service config from parent channel to grpclb balancer channel.	6 years ago
Muxi Yan	ae88ee803d	clang-format	6 years ago
Muxi Yan	cd22177e04	Add manual test suite	6 years ago
Vijay Pai	f815656256	Instantiate an application callback exec ctx in security filter	6 years ago
Yash Tibrewal	c71b2f4fb7	Global Interceptor Registration allowed only once	6 years ago
Yash Tibrewal	ff72f3eeff	Skip the test instead	6 years ago
Yash Tibrewal	56a93d4c18	Add no enum sanitizer annotations around functions that need to fill/load in grpc_status_code	6 years ago
Juanli Shen	db1c09ad49	Fix subchannel call destruction	6 years ago
Eric Gribkoff	a61785f184	Don't count internal ApplicationCallbackExcCtx against ExcCtx count This was originally done for ExcCtx in https://github.com/grpc/grpc/pull/15825. This avoids a hang in our pre-fork handler, where grpc_core::Executor::RunClosures attempts to increment the thread count from an invocation of grpc_prefork():	6 years ago
Jan Tattermusch	8a9e074237	build -dev nugets with timestamp suffix	6 years ago
Jan Tattermusch	70a05a7c53	unify usage of Version vs VersionPrefix	6 years ago
Jan Tattermusch	9197a6ea25	simplify c# build_packages scripts, get rid of a template	6 years ago
Jan Tattermusch	aaaa32ef7d	change Grpc.Core.NativeDebug metapackage into csproj	6 years ago
Jan Tattermusch	236d657afc	change Grpc.nuspec metapackage into csproj	6 years ago
Keith Moyer	619e6c8ef6	Compare mask with zero Masking a value and comparing to one will only work if the mask itself is equal to one (which is not the case here). Comparing to zero works for any mask.	6 years ago
Juanli Shen	1fab48edfc	Add null input handling in grpc_json_destroy()	6 years ago
Soheil Hassas Yeganeh	75dec4d0f2	Track the pollsets of an FD in PO_MULTI mode for pollex. Each pollset in pollex has a lock, grabbed upon adding an FD to the pollset. Since this is called on a per-call basis, there is a flat array caching the FDs of the pollset, to avoid unnecessarily calling epoll_ctl multiple times for the same FD. This has two problems: 1) When multiple threads add FDs to the same pollset, we will have contention on the pollset lock. 2) When we have many FDs we simply run out of cache storage, and call epoll_ctl(). This commit changes the caching strategy by simply storing the epfd of pollsets of an FD inside that FD, when we are in PO_MULTI mode. This results in address in both (1) and (2). Moreover, this commit fixes another performance bug. When we have a release FD callback, we do not call close(). That FD will remain in our epollset, until the new owner of the FD actually call close(). This results in a lot of spurious wake ups when we simply hand off gRPC FDs to other FDs. Note that this is a revision on the reverted commit `e83e463b5a` (PR #17823). The main change is to track the epfd of the pollset instead of the pollset pointer, so that if the pollset is deleted we can still access the FD. It also halves the size of the cache vector for 64-bit machines.	6 years ago
Yash Tibrewal	3af64e8495	Reduce logging level to info from error for goaway	6 years ago
Yash Tibrewal	25797cf806	Add logging around GOAWAYs and keepalives	6 years ago
Nicolas Noble	26605fa309	Revert "Track the pollsets of an FD in PO_MULTI mode for pollex."	6 years ago
Jan Tattermusch	8a5c52e53f	move Logging types back to Grpc.Core	6 years ago
Soheil Hassas Yeganeh	e2b668e314	Fix an issue upon setting kEmptinessTheshold. The bug is that kEmptinessTheshold is always set to 0.	6 years ago
Jan Tattermusch	bc4c77bf95	ignore reserved bit in WINDOW_UPDATE frame	6 years ago
ncteisen	ade38d75c1	Fix tsan	6 years ago
Jan Tattermusch	0da52ab133	Fix typo in flow control trace	6 years ago
Juanli Shen	fdae4dc833	Second attemp to fix use-after-free in health check client	6 years ago
Yash Tibrewal	84a537b1d1	Default compression level quick fix	6 years ago
Vijay Pai	45c684f894	Allow an alarm to be set again after firing	6 years ago
Prashant Jaikumar	bca92b2b7e	Added test for wall-clock time change on the client	6 years ago
Eric Gribkoff	5a20b60cda	fix flake in test_abort_does_not_leak_local_vars	6 years ago
Yash Tibrewal	cc19a55338	Revert "Revert "Fix for 17338. Delay shutdown of buffer list till tcp_free to avoid races""	6 years ago
Soheil Hassas Yeganeh	e83e463b5a	Track the pollsets of an FD in PO_MULTI mode for pollex. Each pollset in pollex has a lock, grabbed upon adding an FD to the pollset. Since this is called on a per-call basis, there is a flat array caching the FDs of the pollset, to avoid unnecessarily calling epoll_ctl multiple times for the same FD. This has two problems: 1) When multiple threads add FDs to the same pollset, we will have contention on the pollset lock. 2) When we have many FDs we simply run out of cache storage, and call epoll_ctl(). This commit changes the caching strategy by simply storing the pollsets of an FD inside that FD, when we are in PO_MULTI mode. This results in address in both (1) and (2). Moreover, this commit fixes another performance bug. When we have a release FD callback, we do not call close(). That FD will remain in our epollset, until the new owner of the FD actually call close(). This results in a lot of spurious wake ups when we simply hand off gRPC FDs to other FDs.	6 years ago
Yash Tibrewal	f37e18b8fd	Dummy Shutdown should still unref the error	6 years ago

... 6 7 8 9 10 ...

21005 Commits (f01997b377a989ff2aa79fee70e2f6f0efc55cd4)