Kokoro build runners have an old version of the Go command. Update the
run_local.sh script that they use to build in Docker.
For golang/go#40451
Change-Id: I89fc422e6824045ffdb610fe0f9a106c6cdb875e
Reviewed-on: https://go-review.googlesource.com/c/tools/+/247681
Reviewed-by: Rebecca Stambler <rstambler@golang.org>
Progress reporting has gotten complicated, and has had a couple bugs.
Factor out progress-related behavior to a new progressTracker type, and
use this to implement some unit tests.
Also rename some methods to remove stuttering, and reorganize the code
to be more logical.
Fixesgolang/go#40527
Change-Id: I93d53a67982460e7171f892021e99f4523fe3e5d
Reviewed-on: https://go-review.googlesource.com/c/tools/+/247407
Run-TryBot: Robert Findley <rfindley@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Rebecca Stambler <rstambler@golang.org>
Our WorkDone reporting was generating a random token for each unit of
work, even if a token was supplied by the client. Change this to use
the client token if it is non-empty, and skip the
workDoneProgress/create request.
After this change we can no longer rely on tokens being a string.
Update our progress tracking accordingly.
For golang/go#40527
Change-Id: I702f739c466efb613b69303aaf07005addd3b5e2
Reviewed-on: https://go-review.googlesource.com/c/tools/+/247321
Run-TryBot: Robert Findley <rfindley@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Rebecca Stambler <rstambler@golang.org>
Having a private field with a json tag does not necessary means the
field needs to be exposed publicly. This change ignores the warning of
such fields that contain a json tag with the field tag "-".
The lines added in the unit tests will fail without these changes with
the error:
```
a/a.go:36:2: unexpected diagnostic: struct field b has json tag but is not exported
```
Change-Id: Ife987b99c264ae3b60a702e43a1f9c778f8c2d62
Reviewed-on: https://go-review.googlesource.com/c/tools/+/245857
Run-TryBot: Rebecca Stambler <rstambler@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Rebecca Stambler <rstambler@golang.org>
The gc annotation details code lens did build binaries during
diagnostics to the module root. When trying to enable details for a
main package in a sub directory, it failed since the output binary name
conflicted with the sub directory name.
Instead, specify output to the package directory during build to avoid
conflicts.
Change-Id: Idc11e0c6a9ba15a66645aeef89bffb5abde76928
Reviewed-on: https://go-review.googlesource.com/c/tools/+/246419
Run-TryBot: Pontus Leitzler <leitzler@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Peter Weinberger <pjw@google.com>
Now we downrank candidates that have already been used in other switch
cases. For example:
switch time.Now().Weekday() {
case time.Monday:
case time.<> // downrank time.Monday
}
It wasn't quite as simple as tracking the seen types.Objects.
Consider this example:
type foo struct {
i int
}
var a, b foo
switch 123 {
case a.i:
case <>
}
At <> we don't want to downrank "b.i" even though "i" is represented
as the same types.Object for "a.i" and "b.i". To accommodate this, we
track having seen ["a", "i"] together. We will downrank "a.i", but not
"b.i".
I applied only a minor 0.9 downranking when the candidate has already
been used in another case clause. It is hard to know what the user is
planning. For instance, in the preceding example the user could intend
to write "a.i + 1", so we mustn't downrank "a.i" too much.
Change-Id: I62debc5be3d5d310deb69d11770cf5f8bd9add1d
Reviewed-on: https://go-review.googlesource.com/c/tools/+/247337
Run-TryBot: Muir Manders <muir@mnd.rs>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Rebecca Stambler <rstambler@golang.org>
We keep track of the N highest seen scores so we can quickly skip deep
completions not in the top N. Our logic for maintaining the top N list
wasn't quite right, resulting in certain cases where we would let
non-high scoring candidates through. I don't think the bug impacted
correctness.
Change-Id: Ic105617523c82f0e71e4f95ef0ee216182a84252
Reviewed-on: https://go-review.googlesource.com/c/tools/+/247418
Run-TryBot: Muir Manders <muir@mnd.rs>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Rebecca Stambler <rstambler@golang.org>
Now we will filter out the types already used in other case
statements:
switch ast.Node(nil).(type) {
case *ast.Ident:
case *ast.I<> // don't offer "Ident" since it has been used
}
Note that the implementation was not able to use a map to track the
seen types.Types because we build up types.Type entries dynamically
when searching for completions (e.g. types.NewPointer() to make a
pointer type). We must use types.Identical() instead of direct pointer
equality.
Change-Id: I316638bb48bfd6802e2caea671f297d640291010
Reviewed-on: https://go-review.googlesource.com/c/tools/+/247098
Run-TryBot: Muir Manders <muir@mnd.rs>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Rebecca Stambler <rstambler@golang.org>
This change adds the notion of a "workspace directory", which is
basically the set of directories that contains workspace packages. These
are mainly used for replace targets right now. It's a little trickier
than expected because the set of workspace directories can technically
change on any go.mod change.
At first, I wanted DidModifyFiles to report whether there was a change,
but I don't think it's actually that expensive to check on each call
and it complicates the code a bit. I can change it back if you think
it's worth doing.
The parse mod handle changes are because I needed an unlocked way of
parsing the mod file, but I imagine they'll conflict with CL 244769
anyway.
The next CL will be to "promote" replace targets to the level of
workspace packages, meaning we will be able to find references in them.
Change-Id: I5dd58fe29415473496ca6634a94a3134923228dc
Reviewed-on: https://go-review.googlesource.com/c/tools/+/245327
Run-TryBot: Rebecca Stambler <rstambler@golang.org>
Reviewed-by: Heschi Kreinick <heschi@google.com>
This CL is mostly a refactoring of the logic to compute errors for
`go mod tidy` diagnostics. It had been getting a little confusing, so
hopefully this makes things easier to read.
I made a few other small changes, such as slightly changing a few error
messages and showing diagnostics in the go.mod file for all missing
modules.
Change-Id: Ia8cf580731b997248591a2d64dff133accd9c5aa
Reviewed-on: https://go-review.googlesource.com/c/tools/+/244610
Run-TryBot: Rebecca Stambler <rstambler@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Heschi Kreinick <heschi@google.com>
* adds gopls command line tool for call hierarchy
* adds lsp setup for call hierarchy
* adds handler for textDocument/prepareCallHierarchy to display selected
identifier and get incoming/outgoing calls for it
* setup testing
Change-Id: I0a0904abdbe11273a56162b6e5be93b97ceb9c26
Reviewed-on: https://go-review.googlesource.com/c/tools/+/246521
Run-TryBot: Danish Dua <danishdua@google.com>
Reviewed-by: Heschi Kreinick <heschi@google.com>
We missed a possible case in checking for gopackagesdriver - a binary
named gopackagesdriver works the same way as setting GOPACKAGESDRIVER.
Change-Id: I676800d253950cb35d74211558bafab340310653
Reviewed-on: https://go-review.googlesource.com/c/tools/+/247179
Run-TryBot: Rebecca Stambler <rstambler@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Michael Matloob <matloob@golang.org>
We did not adjust the range in extractFunction(). We only adjusted
the range in canExtractFunction().
Change-Id: Idc1eab775988ab61be6df8b4afd4b1a36a8bb0ff
Reviewed-on: https://go-review.googlesource.com/c/tools/+/247405
Run-TryBot: Josh Baum <joshbaum@google.com>
Reviewed-by: Rebecca Stambler <rstambler@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
In the previous implementation, we kept the first variable in the
return statement that matched the each given return type. Now, we
keep searching for a non-"zero" value, even if we have already found
a "zero" value.
Change-Id: Icf0987bab90239781452319979e7a30502807e36
Reviewed-on: https://go-review.googlesource.com/c/tools/+/246917
Run-TryBot: Josh Baum <joshbaum@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Rebecca Stambler <rstambler@golang.org>
Now we prefer functions when completing "go" and "defer" statements.
Previously we had no preference for the type of object. Further, we
will now also properly invoke functions.
var f1 int
var f2 func()
go f<> // prefers "f2" and expands to "f2()"
Change-Id: I213551b74ba453c337ac89e825b5d495659e9d65
Reviewed-on: https://go-review.googlesource.com/c/tools/+/246359
Run-TryBot: Muir Manders <muir@mnd.rs>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Rebecca Stambler <rstambler@golang.org>
Consider this example:
var foo, bar int
if foo == 123 && b<> {
}
Completing at "<>" previously preferred the unimported
"bytes.Contains()" function because it returns a bool. You often need
to compose a boolean expression from non-boolean candidates, so
preferring only bool candidates gives unhelpful results. Now we don't
infer any expected type for "&&" and "||", which allows the example to
prefer "bar" as the top candidate.
Fixesgolang/go#37163.
Change-Id: Ic341da11dd626439cfb265d129288c5ca6008270
Reviewed-on: https://go-review.googlesource.com/c/tools/+/246362
Run-TryBot: Muir Manders <muir@mnd.rs>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Rebecca Stambler <rstambler@golang.org>
I noticed an annoying completion ranking issue:
// ranks "HandlerFunc" over "HandleFunc"
http.HandleFunc<>
This was due to us downranking function calls to prefer fields/vars. I
tweaked the logic to only downrank methods (with a receiver).
Change-Id: Ia4040dc8a35f641be2d7d934bf555090831219ac
Reviewed-on: https://go-review.googlesource.com/c/tools/+/246357
Run-TryBot: Muir Manders <muir@mnd.rs>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Rebecca Stambler <rstambler@golang.org>
In the current implementation, the return value of wasFirstChanges() is
reversed and did not record the URI of the changed file. In addition,
the snapshot was not stored in the snapshotByURI.
So I fixed when didChange() is executed, the URI of the edited file is
registered and wasFirstChanges() returns true if the URI is not
registered. Also the snapshot is now stored in snapshotByURI.
Fixesgolang/go#40531
Change-Id: I0adbf7459593d70660beb3b37900ffc88f707917
Reviewed-on: https://go-review.googlesource.com/c/tools/+/247118
Run-TryBot: Rebecca Stambler <rstambler@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Rebecca Stambler <rstambler@golang.org>
In the previous implementation, the extracted function would
sometimes include superfluous parameters and return values. It
might also unnecessarily initialize variables. This CL introduces
3 rules to limit this behavior. (1) a variable is not passed as a
parameter to the extracted function if its first use within the
function is its own redefinition. (2) a variable is not returned
from the extracted function if its first use after the function is its
own redefinition. (3) when possible, we redefine variables in the call
expression to the extracted function.
Change-Id: Ideb5a7eff8a1bf462c83271a2f043116ff5d8b76
Reviewed-on: https://go-review.googlesource.com/c/tools/+/244770
Run-TryBot: Josh Baum <joshbaum@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Rebecca Stambler <rstambler@golang.org>
CL 246757 resulted in an infinite loop because the value of "o" is
never updated.
Change-Id: I79cf265349838de19089c4468128c565a9a3cda3
Reviewed-on: https://go-review.googlesource.com/c/tools/+/247182
Run-TryBot: Rebecca Stambler <rstambler@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Josh Baum <joshbaum@google.com>
In the example:
append([]T{}, <>)
We used to track the "objType" as "[]T", and "variadicType" separately
as "T". However, most things are more interested in "T" vs "[]T", so
they had to fiddle around with swapping types. Now instead we track
objType as T, and add a "variadic=true" flag indicating that "[]T" is
also an acceptable type.
Change-Id: I8ee3ef840917378c8406368cb5c660a377498dfd
Reviewed-on: https://go-review.googlesource.com/c/tools/+/246698
Run-TryBot: Muir Manders <muir@mnd.rs>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Rebecca Stambler <rstambler@golang.org>
Fix a minor completion ranking issue:
foo := func(int, int) {}
foo(123, <>)
Previously we were preferring "foo()" at "<>" even though it can't be
used. We mistakenly thought we were completing the first param because
the *ast.CallExpr appears to only have a single param.
Change-Id: Iedbbb1870a4b9eb5d5be4ed266b8bb3e313b496b
Reviewed-on: https://go-review.googlesource.com/c/tools/+/246697
Run-TryBot: Muir Manders <muir@mnd.rs>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Rebecca Stambler <rstambler@golang.org>
The gc_details command, which shows the gc compiler's decisions, can
produce thousands of diagnostics for a package. New gopls options
'noBounds', 'noEscape', 'noInline', 'noNilcheck' will suppress diagnostics
of less interest to the user. These are in a new 'annotations' section
parallel to 'codelens' or 'analyses'.
Change-Id: Ica75de25b14f38b67ddfa9f997f581674f45221d
Reviewed-on: https://go-review.googlesource.com/c/tools/+/246477
Run-TryBot: Peter Weinberger <pjw@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Rebecca Stambler <rstambler@golang.org>
snapshot.View().Session().Cache().FileSet() has been driving me crazy
for a while. Add it to snapshot. Along the way, discover that the Cache
interface is now totally unused and delete it.
I also changed a bunch of View arguments to Snapshot while I was in the
area.
Change-Id: I1064d0020b1567c2ed28d2d55e0f4649eb94c060
Reviewed-on: https://go-review.googlesource.com/c/tools/+/245324
Run-TryBot: Heschi Kreinick <heschi@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Rebecca Stambler <rstambler@golang.org>
Only add files in errors if either the error's import stack is empty,
or the import stack's top package is the same as the package itself,
so we have more confidence that the error applies to the package.
This is bound to be flaky, but shouldn't be worse than the current
state. (And it unblocks a cl from going into the RC...).
Updates golang/go#40544
Change-Id: Ie21a8abec7150800d3d34b94a7ec90fd40d93fe9
Reviewed-on: https://go-review.googlesource.com/c/tools/+/246758
Run-TryBot: Michael Matloob <matloob@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Rebecca Stambler <rstambler@golang.org>
Ran into this while debugging another issue.
Change-Id: I154493418c7676a24457a4e11431ad4f0311c07a
Reviewed-on: https://go-review.googlesource.com/c/tools/+/246757
Run-TryBot: Rebecca Stambler <rstambler@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Josh Baum <joshbaum@google.com>
In the previous implementation, fillreturns only altered return
statements that contained too few values. Now, fillreturns also examines
return statements with too many return values. In these situations,
we remove any value that is a "zero value" and does not match a type
in the return signature.
Change-Id: I7548307234ff4b16534b72a8aead95a322eb535a
Reviewed-on: https://go-review.googlesource.com/c/tools/+/246520
Run-TryBot: Josh Baum <joshbaum@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Rebecca Stambler <rstambler@golang.org>
This CL addresses completion prefix not being overwritten for completion
in comments for exported variables/functions/types etc. Instead of
setting the surrounding range as cursor position, we expand out from
cursor instead to replace the word we're currently on.
Fixesgolang/go#39262
Change-Id: I90c28562e3ef285ce6848598f8d7bd7545d5c957
Reviewed-on: https://go-review.googlesource.com/c/tools/+/246237
Run-TryBot: Danish Dua <danishdua@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Heschi Kreinick <heschi@google.com>
Previously, the suggested fix tests did not properly handle the case
in which one fix contained at least two edits. We also prevent
the server from panicing when we cannot extract the selection.
Change-Id: I38f7b6d871b2f2741349a3fd94fd95b396f5fd33
Reviewed-on: https://go-review.googlesource.com/c/tools/+/246458
Reviewed-by: Rebecca Stambler <rstambler@golang.org>
Run-TryBot: Rebecca Stambler <rstambler@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
In the previous implementation, a test would pass if the given
command could not be applied to the given range.
Change-Id: I2e63972472cbd146cb5f27a3e27c878387222cb6
Reviewed-on: https://go-review.googlesource.com/c/tools/+/246517
Reviewed-by: Rebecca Stambler <rstambler@golang.org>
We were returning the AST node for the identifier that the rename was
called from, not the actual declaration, so the doc comments weren't
getting updated.
Fixesgolang/go#40463
Change-Id: Id8ba0b0aeb8f42d2aaa561e7a964edcca5202916
Reviewed-on: https://go-review.googlesource.com/c/tools/+/245817
Run-TryBot: Rebecca Stambler <rstambler@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Robert Findley <rfindley@google.com>
People reading this doc may not know, how versions in "go get" work
or how Git tags map to Go module versions. So, add an example
of getting a specific version of gopls.
Change-Id: Id75cc8829bbbf710c3c021eab7d956433ba4110e
Reviewed-on: https://go-review.googlesource.com/c/tools/+/246418
Reviewed-by: Rebecca Stambler <rstambler@golang.org>
parseGoHandles have lifetimes separate from the packages they belong to.
For example, a package may be invalidated by a change to one of its
files, but we still want to retain the parse results for all the rest.
Track them explicitly.
Change-Id: I03a4ffe283bf2b252d2d838bdb2cf332cd981075
Reviewed-on: https://go-review.googlesource.com/c/tools/+/245059
Run-TryBot: Heschi Kreinick <heschi@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Robert Findley <rfindley@google.com>
Reviewed-by: Rebecca Stambler <rstambler@golang.org>
This CL removes duplicate code in lineFoldingRange function under
lsp/source/folding_range.go and generally improves code quality.
Fixes bug with composite literal folding where gopls was folding literals
with braces on the same line as end token (paranthesis/braces).
Change-Id: I742f285d866d72a243129c0aef0935fe2a1ad0dd
Reviewed-on: https://go-review.googlesource.com/c/tools/+/245757
Reviewed-by: Heschi Kreinick <heschi@google.com>
FileHandle currently includes LSP-level information about Version and
Session. That's dangerous, because the cache operates in terms of
URIs and content only -- we explicitly want to share results across
sessions and versions if they happen to be the same.
Split the LSP information into separate types, VersionedFileHandle and
VersionedFileIdentity.
Change-Id: I158646b783375b58245468599301e2a29c657e71
Reviewed-on: https://go-review.googlesource.com/c/tools/+/245058
Run-TryBot: Heschi Kreinick <heschi@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Rebecca Stambler <rstambler@golang.org>
Reviewed-by: Robert Findley <rfindley@google.com>
The builtin package was the one special case where we parsed Go outside
the context of a Snapshot. Move it up.
Change-Id: I1f4bb536adb40019e0ea9c5c89f38b15737abb8c
Reviewed-on: https://go-review.googlesource.com/c/tools/+/245057
Run-TryBot: Heschi Kreinick <heschi@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Rebecca Stambler <rstambler@golang.org>
To manually collect cache entries, we need to know when a snapshot is
idle. Add a reference count in the form of a WaitGroup and keep track of
its uses. The pattern is that any time a snapshot is returned, it comes
with a release function that decrements the ref count.
Almost all uses of a snapshot originate in a user-facing request,
handled in beginFileRequest. There it's mostly an exercise in passing
Snapshots around instead of Views.
In the other places I took the path of least resistance. For file
modifications I tried to minimize the amount of code that needed to deal
with snapshots. For diagnostics I just acquired the snapshot at the
diagnostics call.
Change-Id: Id48a2df3acdd97f27d905e2c2be23072f28f196b
Reviewed-on: https://go-review.googlesource.com/c/tools/+/241837
Run-TryBot: Heschi Kreinick <heschi@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Rebecca Stambler <rstambler@golang.org>
Previously, we were only invalidating workspace packages when the go.mod
changed, but we really need to be invalidating all known packages.
Fixesgolang/go#40456
Change-Id: I9ad353a26ab40c74c7760ed7a1c5de517640cfab
Reviewed-on: https://go-review.googlesource.com/c/tools/+/245779
Run-TryBot: Rebecca Stambler <rstambler@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Heschi Kreinick <heschi@google.com>
This change is generated whenever you run go test . --golden to generate
updated golden files. Commiting it so it doesn't show up in other CLs.
The module names are now sorted and hence this change shouldn't happen again.
Change-Id: I5d6c44d9a4f3ca24e336afad301136af7123eef7
Reviewed-on: https://go-review.googlesource.com/c/tools/+/245997
Reviewed-by: Heschi Kreinick <heschi@google.com>
Run-TryBot: Danish Dua <danishdua@google.com>
Packages that have gc_details enabled is stored internally in a
map[span.URI]struct{}, but when toggling on/off via ExecuteCommand, the
scheme is lost. The mismatch lead to no diagnostics from gc_details.
This fix uses span.URIFromPath() to parse the file path, so that the
internal keys are found when diagnostics are calculated.
Change-Id: I53b4e869625251c366b12b570eb03212319d2770
Reviewed-on: https://go-review.googlesource.com/c/tools/+/245858
Run-TryBot: Pontus Leitzler <leitzler@gmail.com>
Reviewed-by: Heschi Kreinick <heschi@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Continuing the massacre, remove ParseModHandle, and Mod*Handle, from the
source API.
Notably, having the snapshot available means we can simplify the go
command invocation paths a lot.
Change-Id: Ief4ef41e42f93d653f719a230004861e5e1ef70b
Reviewed-on: https://go-review.googlesource.com/c/tools/+/244769
Run-TryBot: Heschi Kreinick <heschi@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Rebecca Stambler <rstambler@golang.org>
We're going to switch to running govim tests at main as post-submit CI
rather than presubmit, and will also switch to running them via Kokoro
using the run_local script rather than cloud build.
Enable this by changing the semantics of run_local.sh to default to
main.
For golang/go#40451
Change-Id: I9c311dea8326a36a3f8335eddbfae0ce7f02f6bf
Reviewed-on: https://go-review.googlesource.com/c/tools/+/245539
Run-TryBot: Robert Findley <rfindley@google.com>
Reviewed-by: Rebecca Stambler <rstambler@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
There were a few merge conflict-related issues in the GC optimization
details CL. Also fixed a few things I noticed after the fact, like
separating out a new mutex.
Staticcheck caught a few things, and I also fixed a bug I noticed
in the cache package.
Change-Id: I3fc519373253418586dca08fdec3114b30a247ea
Reviewed-on: https://go-review.googlesource.com/c/tools/+/245399
Run-TryBot: Rebecca Stambler <rstambler@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Peter Weinberger <pjw@google.com>
