Gopls behavior on disconnection is currently somewhat undefined, because
it hasn't mattered when there was a single gopls session per binary
invocation. With golang/go#34111, this changes.
Checks are added to ensure clients and sessions are cleaned up when an LSP
connection closes. Also, normal client disconnection is differentiated
with the jsonrpc2.ErrDisconnected value.
Updates golang/go#34111
Change-Id: I74d48ad6dcfc30d11f7f751dcffb20c18a4cbaa3
Reviewed-on: https://go-review.googlesource.com/c/tools/+/220519
Run-TryBot: Robert Findley <rfindley@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Heschi Kreinick <heschi@google.com>
When running gopls against an automatically started remote instance, we
want the lifecycle of the remote to be detached from that of its
clients, so that it doesn't shut down while clients are still connected.
On the other hand, a gopls process can consume significant resources, so
we don't want it to remain when there are no more connected clients.
The jsonrpc2 package is updated to support the concept of idle timeout:
a duration after which the server is shut down when there are no
connected clients. This is exposed in the gopls serve command via the
-listen.timeout flag.
Update golang/go#34111
Change-Id: Id62b3d4a2fa66de2c9306d130ca431717f01d1e5
Reviewed-on: https://go-review.googlesource.com/c/tools/+/220281
Run-TryBot: Robert Findley <rfindley@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Heschi Kreinick <heschi@google.com>
Not all regtests resulted in LSP shutdown, which caused temp modfiles to
be leaked. After this fix I have confirmed that /tmp is clean after a
successful run of the regtests.
Also proactively clean up the unix socket file when serving jsonrpc2
over UDS.
Change-Id: I745fbd3d2adeeb165cadf7c54fd815d8df81d4e4
Reviewed-on: https://go-review.googlesource.com/c/tools/+/220061
Run-TryBot: Robert Findley <rfindley@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Rohan Challa <rohan@golang.org>
Reviewed-by: Heschi Kreinick <heschi@google.com>
For tests (and perhaps later, for daemon discovery), unix domain sockets
offer advantages over TCP: we can know the exact socket address that will be
used when starting a server subprocess. They also offer performance and
security advantages over TCP, and were specifically requested on
golang.org/issues/34111.
This CL adds support for listening on UDS, and uses this to implement an
additional regtest environment mode that starts up an external process.
This mode is disabled by default, but may be enabled by the
-enable_gopls_subprocess_tests.
The regtest TestMain may be hijacked to instead run as gopls, if a
special environment variable is set. This allows the the test runner to
start a separate process by using os.Argv[0]. The -gopls_test_binary
flag may be used to point tests at a separate gopls binary.
Updates golang/go#36879
Updates golang/go#34111
Change-Id: I1cfdf55040e81ffa69a6726878a96529e5522e82
Reviewed-on: https://go-review.googlesource.com/c/tools/+/218839
Run-TryBot: Robert Findley <rfindley@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Heschi Kreinick <heschi@google.com>
Update the servertest package to support connecting to a jsonrpc2 server
using either TCP or io.Pipes. The latter is provided so that regtests
can more accurately mimic the current gopls execution mode, where gopls
is run as a sidecar and communicated with via a pipe.
Updates golang/go#36879
Change-Id: I0e14ed0e628333ba2cc7b088009f1887fcaa82a5
Reviewed-on: https://go-review.googlesource.com/c/tools/+/218777
Run-TryBot: Robert Findley <rfindley@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Heschi Kreinick <heschi@google.com>
We had a deadlock in cases where a request was cancelled (1) after being
written to the stream, but (2) before a response was received. This
resulted in the request ID being removed from the pending map while the
server has the request, after which point the server response would hang
in Conn.Run trying to send to a nil channel.
After fixing this nil send there was still a race: it was possible that
Conn.Run could get the pending request, and Conn.Call would select
ctx.Done before Conn.Run could send to the response channel, again
resulting in a blocking send. Fix this by adding a buffer to the
response channel.
The response channel management is also made less forgiving, because we
should be able to reason precisely about how many sends and receives
will occur:
+ Don't close the response channel after sending a response: there
should only be one recipient.
+ Don't delete the ID from pending map twice: it should only be cleaned
up by Conn.Call.
Cancellation tests in the lsprpc package are updated to exercise the
race conditions.
Fixesgolang/go#37159
Change-Id: Ie3207442ea910f79247b18d8647fd52f39fb15db
Reviewed-on: https://go-review.googlesource.com/c/tools/+/219126
Run-TryBot: Robert Findley <rfindley@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Heschi Kreinick <heschi@google.com>
Previously, the process of instantiating and running the LSP server was
sharded across the lsp, protocol, and cmd packages, and this resulted in
some APIs that are hard to work with. For example, it's hard to guess
the difference between lsp.NewClientServer, lsp.NewServer,
protocol.NewServer (which returns a client), and protocol.NewClient
(which returns a server).
This change reorganizes Server instantiation as follows:
+ The lsp.Server is now purely an implementation of the protocol.Server
interface. It is no longer responsible for installing itself into the
jsonrpc2 Stream, nor for running itself.
+ A new package 'lsprpc' is added, to implement the logic of binding an
incoming connection to an LSP server session. This is put in a
separate package for lack of a clear home: it didn't really
philosophically belong in any of the lsp, cmd, or protocol packages.
We can perhaps move it to cmd in the future, but I'd like to keep it
as a separate package while I develop request forwarding.
simplified import graph:
jsonrpc2 ⭠ lsprpc ⭠ cmd
⭩ ⭦
lsp (t.b.d. client tests)
⭩ ⭨
protocol source
+ The jsonrpc2 package is extended to have a minimal API for running a
'StreamServer': something analogous to an HTTP server that listens
for new connections and delegates to a handler (but we couldn't use
the word 'Handler' for this delegate as it was already taken).
After these changes, I hope that the concerns of "serving the LSP",
"serving jsonrpc2", and "installing the LSP on jsonrpc2" are more
logically organized, though one legitimate criticism is that the word
'Server' is still heavily overloaded.
This change prepares a subsequent change which hijacks the jsonrpc2
connection when forwarding messages to a shared gopls instance.
To test this change, the following improvements are made:
+ A servertest package is added to make it easier to run a test against
an in-process jsonrpc2 server. For now, this uses TCP but it could
easily be modified to use io.Pipe.
+ cmd tests are updated to use the servertest package. Unfortunately it
wasn't yet possible to eliminate the concept of `remote=internal` in
favor of just using multiple sessions, because view initialization
involves calling both `go env` and `packages.Load`, which slow down
session startup significantly. See also golang.org/issue/35968.
Instead, the syntax for `-remote=internal` is modified to be
`-remote=internal@127.0.0.1:12345`.
+ An additional test for request cancellation is added for the
sessionserver package. This test uncovered a bug: when calling
Canceller.Cancel, we were using id rather than &id, which resulted in
incorrect json serialization (as only the pointer receiver implements
the json.Marshaller interface).
Updates golang/go#34111
Change-Id: I75c219df634348cdf53a9e57839b98588311a9ef
Reviewed-on: https://go-review.googlesource.com/c/tools/+/215742
Run-TryBot: Robert Findley <rfindley@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Rebecca Stambler <rstambler@golang.org>
Reviewed-by: Heschi Kreinick <heschi@google.com>
Found by running the go vet pass 'testinggoroutine'.
Change-Id: I38f17877e2a97ffb823bb97850d21107743271d7
Reviewed-on: https://go-review.googlesource.com/c/tools/+/217179
Run-TryBot: Heschi Kreinick <heschi@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Heschi Kreinick <heschi@google.com>
CL 215738 didn't work because the canceller was embedded in the
serverHandler, which already had a Deliver method. Add it as an actual
handler instead.
Change-Id: I0c79f1bee67aa3b4da53d92547804de859f1938c
Reviewed-on: https://go-review.googlesource.com/c/tools/+/216303
Run-TryBot: Heschi Kreinick <heschi@google.com>
Reviewed-by: Rebecca Stambler <rstambler@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
This change allows us to hanel cancel messages as they go into the queue, and
cancel messages that are ahead of them in the queue but not being processed yet.
This should reduce the amount of redundant work that we do when we are handling
a cancel storm.
Change-Id: Id1a58991407d75b68d65bacf96350a4dd69d4d2b
Reviewed-on: https://go-review.googlesource.com/c/tools/+/200766
Run-TryBot: Ian Cottrell <iancottrell@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Rebecca Stambler <rstambler@golang.org>
A detatched context ends up attributing all background work to the initialize
function.
Change-Id: I81206462752228d5ac81408fb1e3fb86ab36796e
Reviewed-on: https://go-review.googlesource.com/c/tools/+/186457
Run-TryBot: Ian Cottrell <iancottrell@google.com>
Reviewed-by: Rebecca Stambler <rstambler@golang.org>
We merge them into a single interface and support multiple of them rather than
just one.
This will allow us to stack handlers with different responsabilities and extract
some core logic (like tracing) out to a handler where it belongs.
Change-Id: I6aab92138550c5062fcb1bed86171e0850d1eb38
Reviewed-on: https://go-review.googlesource.com/c/tools/+/185879
Reviewed-by: Rebecca Stambler <rstambler@golang.org>
also change the return type to be and end function and not an incomplete span
Change-Id: Icd99d93ac98a0f8088f33e905cf1ee3fe410c024
Reviewed-on: https://go-review.googlesource.com/c/tools/+/185349
Run-TryBot: Ian Cottrell <iancottrell@google.com>
Reviewed-by: Rebecca Stambler <rstambler@golang.org>
This is the basic library that allows for recording of stats about the program
operation.
Change-Id: I09f7e3de5fc37aaf29bc0db46f15b15056fc0eb2
Reviewed-on: https://go-review.googlesource.com/c/tools/+/185338
Run-TryBot: Ian Cottrell <iancottrell@google.com>
Reviewed-by: Rebecca Stambler <rstambler@golang.org>
This maps more directly to the basic telementery tagging requirements and uses
the context package in a way that is more idomatic.
Change-Id: If08c429b897bddfe014224ac2d92d7796a521ab9
Reviewed-on: https://go-review.googlesource.com/c/tools/+/184941
Run-TryBot: Ian Cottrell <iancottrell@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Rebecca Stambler <rstambler@golang.org>
This abandons the limited size queue
Instead we kick a go-routine per request, but have each request wait for the
previous request to say it is okay to continue. This allows each request to
control when it is finished with tasks that required strict ordering without
every blocking the routine that handles replies.
It also protects against repeated or missing replies.
Fixesgolang/go#32631Fixesgolang/go#32589Fixesgolang/go#32467Fixesgolang/go#32360Fixesgolang/go#31977
Change-Id: Icd071620052351ec7f8fac136f1b8e3f97d4bb2d
Reviewed-on: https://go-review.googlesource.com/c/tools/+/183718
Run-TryBot: Ian Cottrell <iancottrell@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Rebecca Stambler <rstambler@golang.org>
This separates hides the wire structures, and then exposes a new Request
type to allow for it to carry advanced features.
It also embeds the connection into the request and changes the signature of the
handler to no longer require a separate Conn argument.
Change-Id: I20b54f146285f7a9cb5f279c6ebdf0f286f4b829
Reviewed-on: https://go-review.googlesource.com/c/tools/+/183717
Run-TryBot: Ian Cottrell <iancottrell@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Rebecca Stambler <rstambler@golang.org>
This uses the opencensus compatability later to track all the json rpc calls in
and out.
Change-Id: Ib719879a8d6855b6e6479a4f1b01fe823b548110
Reviewed-on: https://go-review.googlesource.com/c/tools/+/183248
Run-TryBot: Ian Cottrell <iancottrell@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Rebecca Stambler <rstambler@golang.org>
The means that the stream reader can move forward while a message is being processed.
This will significantly improve responsivness and cancellation handling, and also
allow message handlers to send messages themselves, reducing the need to spin up
new go routines inside handlers.
The flow control changes from blocking to failing when a server is busy, which removes
the main current cause of deadlock, but may break non deadlock cases that currently wait
if the queue is not sufficiently large.
Change-Id: Ia73eb049b38d0651344abdbf16c477a8ce1a6fd1
Reviewed-on: https://go-review.googlesource.com/c/tools/+/170007
Reviewed-by: Rebecca Stambler <rstambler@golang.org>
This changes the basic API of a jsonrpc2 connection to run the
read loop as a method rather than in a go routine launched in
the NewConn. This allows the handler to be created and bound
between construction and the read loop starting, which fixes
the race.
Fixesgolang/go#30091
Change-Id: I8201175affe431819cf473e5194d70c019f58425
Reviewed-on: https://go-review.googlesource.com/c/tools/+/170003
Run-TryBot: Ian Cottrell <iancottrell@google.com>
Reviewed-by: Rebecca Stambler <rstambler@golang.org>
Delivering each message in a go routine turned out to be problematic, there are some messages
that must be fully processed before later messages are started, and there was no way to guarantee that.
We now push concurrence handling up to the higher level, this has the disadvantage of not being able to guarantee
we respond to call messages correctly, but its a small price to pay.
The LSP currently processes each message fully in order blocking the handler, while we still work on basic
functionality.
Change-Id: If0648c77713ddbe4fed69da97a57696f433b8002
Reviewed-on: https://go-review.googlesource.com/c/149497
Run-TryBot: Ian Cottrell <iancottrell@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Rebecca Stambler <rstambler@golang.org>
This improves the logging capabilities of the jsonrpc 2 library to always
include the method and also an optional elapsed time.
This is used to implement an lsp inspector compatible logging mode in the golsp.
Change-Id: I2f7ac8b9298c4364b1b89cf6f696b534557ed139
Reviewed-on: https://go-review.googlesource.com/c/146157
Reviewed-by: Rebecca Stambler <rstambler@golang.org>
