Make netpollopen return what Windows GetLastError API returns.
It is probably copy / paste error from long time ago.

Change-Id: I28f78718c15fef3e8b5f5d11a259533d7e9c6185
Reviewed-on: https://go-review.googlesource.com/72592Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>

If n.Type==nil after typechecking, then we should have already
reported a more useful error somewhere else. Just return 0 in
evalunsafe without trying to do anything else that's likely to cause
problems.

Also, further split out issue7525.go into more test files, because
cmd/compile reports at most one typechecking loop per compilation
unit.

Fixes #22351.

Change-Id: I3ebf505f72c48fcbfef5ec915606224406026597
Reviewed-on: https://go-review.googlesource.com/72251
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Daniel Martí <mvdan@mvdan.cc>
Reviewed-by: Robert Griesemer <gri@golang.org>

decoder.decode() was defering close of lzw.decoders created for each
frame in a loop, thus increasing heap usage (referenced object + defered
function) until decode() returns. Memory increased proportionally to the
number of frames. Fix this by moving the sImageDescriptor case block
into its own method.

Fixes #22237

Change-Id: I819617ea7e539e13c04bc11112f339645391ddb9
Reviewed-on: https://go-review.googlesource.com/70370
Run-TryBot: Emmanuel Odeke <emm.odeke@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Emmanuel Odeke <emm.odeke@gmail.com>

Fixes #22282.

Change-Id: I4097e9cbd590ab6c8b1511a3b752c6ac98ac819b
Reviewed-on: https://go-review.googlesource.com/72792Reviewed-by: Rob Pike <r@golang.org>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>

When enhanced DWARF location list generation is enabled (via internal
option -dwarflocationlists), variable entries were missing for "large"
(non-decomposable) locals and formals. From the debugging perspective,
this makes it appear that the variable doesn't exist, which is
probably not what we want. This change insures that a formal/local DIE
is created for these vars (with correct type, line, etc) but with a
conservative ("no info") location.

Change-Id: I10b2e9a51a60c7b4c748e987cdec5f2d8b2837d5
Reviewed-on: https://go-review.googlesource.com/72630
Run-TryBot: Than McIntosh <thanm@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Heschi Kreinick <heschi@google.com>

Fixes #22377.

Change-Id: I0a0e1bde558df964f0961dc4cfc305e72d590e1a
Reviewed-on: https://go-review.googlesource.com/72690Reviewed-by: Alan Donovan <adonovan@google.com>

The addressing mode of global variable was missing, whereas the
compiler may make use of it, causing "illegal combination" error.
This CL adds support of that addressing mode.

Fixes #22390.

Change-Id: Ic8eade31aba73e6fb895f758ee7f277f8f1832ef
Reviewed-on: https://go-review.googlesource.com/72610
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: David Chase <drchase@google.com>

* replace a copy of IsMethod with a call of it.
* a few more switches where they simplify the code.
* prefer composite literals over "n := new(...); n.x = y; ...".
* use defers to get rid of three goto labels.
* rewrite updateHasCall into two funcs to remove gotos.

Passes toolstash-check on std cmd.

Change-Id: Icb5442a89a87319ef4b640bbc5faebf41b193ef1
Reviewed-on: https://go-review.googlesource.com/72070
Run-TryBot: Daniel Martí <mvdan@mvdan.cc>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>

Currently, benchmarking compile performance under -l=4 is confounded
by -l=2 enabling eager typechecking of unused inline function bodies
for debugging. This isn't logically an "inlining aggressiveness"
level, so instead move this logic under the -d umbrella flag.

Change-Id: I713f68952efbe25b6941d3ebc2f3707ccbbd6240
Reviewed-on: https://go-review.googlesource.com/72253
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Robert Griesemer <gri@golang.org>

Increasing the map size with the benchmark iteration count
introduced non-linearities and made benchmark runs slow when
increasing benchtime.

Rework the benchmark to use a map size independent of the
iteration count and instead re-fill it when it becomes empty.

Fixes #21546

Change-Id: Iafb6eb225e81830263f30b3aba0d449c361aec32
Reviewed-on: https://go-review.googlesource.com/57650
Run-TryBot: Emmanuel Odeke <emm.odeke@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>

Previously, the code would only check (w*h), not (w*h*bpp).

Fixes #22304

Change-Id: Iaca26d916fe4b894d460448c416b1e0b9fd68e44
Reviewed-on: https://go-review.googlesource.com/72350Reviewed-by: Rob Pike <r@golang.org>

Running test.bash goes from 30s to 10s on a linux workstation.

(The coming pkg cache work in cmd/go would presumably do the same thing,
but this makes all.bash faster today.)

Change-Id: I8c9b0400071a412fce55b386e939906bb1c1d84d
Reviewed-on: https://go-review.googlesource.com/72330
Run-TryBot: David Crawshaw <crawshaw@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>

CL 72150 fixes #22352 by reverting the problematic parts of that CL
where the line number and column number were inconsistent with each other.
This CL adds back functionality to address the issue that CL 72150
was trying to solve in the first place. That is, it reports the starting
line of the record, so that users have a frame of reference to start with
when debugging what went wrong.

In the event of gnarly CSV files with multiline quoted strings, a parse
failure likely occurs somewhere between the start of the record and
the point where the parser finally detected an error.
Since ParserError.{Line,Column} reports where the *error* occurs, we
add a RecordLine field to report where the record starts.

Also take this time to cleanup and modernize TestRead.

Fixes #19019
Fixes #22352

Change-Id: I16cebf0b81922c35f75804c7073e9cddbfd11a04
Reviewed-on: https://go-review.googlesource.com/72310Reviewed-by: Ian Lance Taylor <iant@golang.org>

For #22095

Change-Id: Idcfdfe8a94db8626392658bb93429454238f648a
Reviewed-on: https://go-review.googlesource.com/70835
Run-TryBot: David Crawshaw <crawshaw@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>

Code essentially mirrors AMD64 implementation.

Change-Id: I39f7f099ce11fdc3772df039998cc11947bb22a2
Reviewed-on: https://go-review.googlesource.com/72270
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>

windows version of Pipe function is implemented by calling
syscall.Pipe which returns handles inheritable by client process,
and then adjusting returned handles with syscall.CloseOnExec.

Just create non-inheritable handles in the first place.
Now that we don't have a race window in the code, drop use
of syscall.ForkLock.

Change-Id: Ie325da7c2397b5995db4a5ddb0117e2ce1745187
Reviewed-on: https://go-review.googlesource.com/72010Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

NewEncoder returns an io.Writer that writes all incoming bytes as
hexadecimal characters to the underlying io.Writer. NewDecoder returns an
io.Reader that does the inverse.

Fixes #21590

Change-Id: Iebe0813faf365b42598f19a9aa41768f571dc0a8
Reviewed-on: https://go-review.googlesource.com/70210Reviewed-by: Joe Tsai <thebrokentoaster@gmail.com>
Reviewed-by: Ian Lance Taylor <iant@golang.org>

For #22095

Change-Id: Iba3dffc782cecc15ea0e90a971a2734729984945
Reviewed-on: https://go-review.googlesource.com/70834
Run-TryBot: David Crawshaw <crawshaw@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>

For #22095

Change-Id: Ica6b3391541fe5a0355620d7c4a5107cf53eee82
Reviewed-on: https://go-review.googlesource.com/70833
Run-TryBot: David Crawshaw <crawshaw@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Daniel Martí <mvdan@mvdan.cc>
Reviewed-by: Ian Lance Taylor <iant@golang.org>

The Reader implementation is slow because it operates on a rune-by-rune
basis via bufio.Reader.ReadRune. We speed this up by operating on entire
lines that we read from bufio.Reader.ReadSlice.

In order to ensure that we read the full line, we augment ReadSlice
in our Reader.readLine method to automatically expand the slice if
bufio.ErrBufferFull is every hit.

This change happens to fix #19410 because it no longer relies on
rune-by-rune parsing and only searches for the relevant delimiter rune.

In order to keep column accounting simple and consistent, this change
reverts parts of CL 52830.

This CL is an alternative to CL 36270 and builds on some of the ideas
from that change by Diogo Pinela.

name                                     old time/op    new time/op    delta
Read-8                                   3.12µs ± 1%    2.54µs ± 2%  -18.76%   (p=0.000 n=10+9)
ReadWithFieldsPerRecord-8                3.12µs ± 1%    2.53µs ± 1%  -18.91%    (p=0.000 n=9+9)
ReadWithoutFieldsPerRecord-8             3.13µs ± 0%    2.57µs ± 3%  -18.07%  (p=0.000 n=10+10)
ReadLargeFields-8                        52.3µs ± 1%     5.3µs ± 2%  -89.93%   (p=0.000 n=10+9)
ReadReuseRecord-8                        2.05µs ± 1%    1.40µs ± 1%  -31.48%   (p=0.000 n=10+9)
ReadReuseRecordWithFieldsPerRecord-8     2.05µs ± 1%    1.41µs ± 0%  -31.03%   (p=0.000 n=10+9)
ReadReuseRecordWithoutFieldsPerRecord-8  2.06µs ± 1%    1.40µs ± 1%  -31.70%   (p=0.000 n=9+10)
ReadReuseRecordLargeFields-8             50.9µs ± 0%     4.1µs ± 3%  -92.01%  (p=0.000 n=10+10)

name                                     old alloc/op   new alloc/op
Read-8                                       664B ± 0%      664B ± 0%
ReadWithFieldsPerRecord-8                    664B ± 0%      664B ± 0%
ReadWithoutFieldsPerRecord-8                 664B ± 0%      664B ± 0%
ReadLargeFields-8                          3.94kB ± 0%    3.94kB ± 0%
ReadReuseRecord-8                           24.0B ± 0%     24.0B ± 0%
ReadReuseRecordWithFieldsPerRecord-8        24.0B ± 0%     24.0B ± 0%
ReadReuseRecordWithoutFieldsPerRecord-8     24.0B ± 0%     24.0B ± 0%
ReadReuseRecordLargeFields-8               2.98kB ± 0%    2.98kB ± 0%

name                                     old allocs/op  new allocs/op
Read-8                                       18.0 ± 0%      18.0 ± 0%
ReadWithFieldsPerRecord-8                    18.0 ± 0%      18.0 ± 0%
ReadWithoutFieldsPerRecord-8                 18.0 ± 0%      18.0 ± 0%
ReadLargeFields-8                            24.0 ± 0%      24.0 ± 0%
ReadReuseRecord-8                            8.00 ± 0%      8.00 ± 0%
ReadReuseRecordWithFieldsPerRecord-8         8.00 ± 0%      8.00 ± 0%
ReadReuseRecordWithoutFieldsPerRecord-8      8.00 ± 0%      8.00 ± 0%
ReadReuseRecordLargeFields-8                 12.0 ± 0%      12.0 ± 0%

Updates #22352
Updates #19019
Fixes #16791
Fixes #19410

Change-Id: I31c27cfcc56880e6abac262f36c947179b550bbf
Reviewed-on: https://go-review.googlesource.com/72150Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Joe Tsai <thebrokentoaster@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

On systems that use kqueue, we always register descriptors for both
EVFILT_READ and EVFILT_WRITE. On at least FreeBSD and OpenBSD, when
the write end of a pipe is registered for EVFILT_READ and EVFILT_WRITE
events, and the read end of the pipe is closed, kqueue reports an
EVFILT_READ event with EV_EOF set, but does not report an EVFILT_WRITE
event. Since the write to the pipe is waiting for an EVFILT_WRITE
event, closing the read end of a pipe can cause the write end to hang
rather than attempt another write which will fail with EPIPE.

Fix this by treating EVFILT_READ with EV_EOF set as making both reads
and writes ready to proceed.

The real test for this is in CL 71770, which tests using various
timeouts with pipes.

Updates #22114

Change-Id: Ib23fbaaddbccd8eee77bdf18f27a7f0aa50e2742
Reviewed-on: https://go-review.googlesource.com/71973Reviewed-by: Matthew Dempsky <mdempsky@google.com>

Fixes #22342.

Change-Id: Ic942162e98dce5749e381a31d58b0bf16c7d6f98
Reviewed-on: https://go-review.googlesource.com/72132
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Austin Clements <austin@google.com>

Since inlining budget calculation is fixed in CL 70151
runtime.nextFreeFast is no longer inlineable on MIPS64x because
it does not support Ctz64 as intrinsic. Skip the test.

Updates #22239.

Change-Id: Id00d55628ddb4b48d27aebfa10377a896765d569
Reviewed-on: https://go-review.googlesource.com/72271
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>

This hack existed because cmd/go used to install (write) and then run
cmd/cgo in the same invocation, and writing and then running a program
is a no-no in modern multithreaded Unix programs (see #22315).

As of CL 68338, cmd/go no longer installs any programs that it then
tries to use. It never did this for any program other than cgo, and
CL 68338 removed that special case for cgo.

Now this special case, added for #3001 long ago, can be removed too.

Change-Id: I338f1f8665e9aca823e33ef7dda9d19f665e4281
Reviewed-on: https://go-review.googlesource.com/71571
Run-TryBot: Russ Cox <rsc@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>

build.go - commands and misc helpers
action.go - action graph construction
exec.go - action graph execution
gc.go - gc toolchain
gccgo.go - gccgo toolchain

Change-Id: I39b6e2490ac05334c2321e9ad88df694a6efa82f
Reviewed-on: https://go-review.googlesource.com/70671
Run-TryBot: Russ Cox <rsc@golang.org>
Reviewed-by: David Crawshaw <crawshaw@golang.org>

It uses the build ID, which is soon to be internal to package work.
Luckily it is also only called from package work.

Change-Id: I5e6662cfe667bdc9190f086be733105ad65a3191
Reviewed-on: https://go-review.googlesource.com/70670
Run-TryBot: Russ Cox <rsc@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: David Crawshaw <crawshaw@golang.org>

Where GCC says "unrecognized command line option", clang says "unknown
argument". This distinction usually doesn't matter because the
compiler will also exit with a non-zero status, but clang 3.4
reportedly exits with a zero status after reporting an unknown argument.

Change-Id: Ieb69ea352a8de0cd4171a1c26708dfe523421cfa
Reviewed-on: https://go-review.googlesource.com/72151
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Damien Neil <dneil@google.com>

When traceviewer encounters a failure of json trace import
due to data error, onImportFail tried to access an error variable
which was not yet defined.

Change-Id: I431be03f179aafacaf1fd3c62a6337e8b5bd18fb
Reviewed-on: https://go-review.googlesource.com/71970Reviewed-by: Austin Clements <austin@google.com>

The test already contained logic to do this however it did not match
the error "cannot find 'ld'" which appears to be how gcc fails when
ld.gold is missing.

Fixes #22340.

Change-Id: I841248cc489b8fa72bc00a95000ad405f9ef8a4f
Reviewed-on: https://go-review.googlesource.com/72111
Run-TryBot: Michael Munday <mike.munday@ibm.com>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

A 1XX, 204, or 304 response may not include a response body according
to RFC 7230, section 3.3.3. If a buggy server returns a 204 or 304
response with a body that is chunked encoded, the invalid body is
currently made readable in the Response. This can lead to data races due
to the transport connection's read loop which does not wait for the body
EOF when the response status is 204 or 304.

The correct behavior is to ignore the body on a 204 or 304 response, and
treat the body data as the beginning of the next request on the
connection.

Updates #22330.

Change-Id: I89a457ceb783b6f66136d5bf9be0a9b0a04fa955
Reviewed-on: https://go-review.googlesource.com/71910Reviewed-by: Tom Bergan <tombergan@google.com>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Run-TryBot: Tom Bergan <tombergan@google.com>

For #22341.

Change-Id: I7e586c58075e4c835ef97701db2f38bdf7558521
Reviewed-on: https://go-review.googlesource.com/71974
Run-TryBot: Robert Griesemer <gri@golang.org>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

https://golang.org/cl/71750 specifies iota values as indices,
thus making them independent from nested constant declarations.
This CL removes some of the comments in the examples that were
still referring to the old notion of iotas being incremented
and reset.

As an aside, please note that the spec still permits the use
of iota in a nested function (like before). Specifically, the
following cases are permitted by the spec (as before):

1) const _ = len([iota]int{})
2) const _ = unsafe.Sizeof(func(){ _ = iota })

For #15550.

Change-Id: I9e5fec75daf7b628b1e08d970512397e9c348923
Reviewed-on: https://go-review.googlesource.com/71912Reviewed-by: Ian Lance Taylor <iant@golang.org>
Reviewed-by: Rob Pike <r@golang.org>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>

racewalk's "foreach" function applies a function to all of a Node's
immediate children, but with a non-idiomatic signature.

This CL reworks it to recursively iterate over the entire subtree
rooted at Node and provides a way to short-circuit iteration.

Passes toolstash -cmp for std cmd with -race.

Change-Id: I738b73953d608709802c97945b7e0f4e4940d3f4
Reviewed-on: https://go-review.googlesource.com/71911
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Robert Griesemer <gri@golang.org>

Change-Id: I094af156e7b8c31af3162d943a8c61268803ab89
Reviewed-on: https://go-review.googlesource.com/71830Reviewed-by: Gabriel Aszalos <gabriel.aszalos@gmail.com>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Gabriel Aszalos <gabriel.aszalos@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

On modern Unix systems it is basically impossible for a multithreaded
program to open a binary for write, close it, and then fork+exec that
same binary. So don't write the binary if we're going to fork+exec it.

This fixes the ETXTBSY flakes.

Fixes #22220.
See also #22315.

Change-Id: I6be4802fa174726ef2a93d5b2f09f708da897cdb
Reviewed-on: https://go-review.googlesource.com/71570
Run-TryBot: Russ Cox <rsc@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>

Issue #15550 is clearly an esoteric case but the spec was silent
about it and we had diverging implementations. By making `iota`
and index that is relative to the respective constant declaration,
nested const declarations won't affect outer values of `iota`.

cmd/compile and go/types already follow this semantics.

Fixes #15550.

Change-Id: If138189e3ea4373f8ba50ac6fb1d219b481f8698
Reviewed-on: https://go-review.googlesource.com/71750Reviewed-by: Rob Pike <r@golang.org>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
Reviewed-by: Ian Lance Taylor <iant@golang.org>

golang.org/cl/37258 was committed to fix issue #16146.

This patch seemed intent to allow at most one dangling byte.  But, as
implemented, many more bytes may actually slip through.  This is because
the LZW layer creates a bufio.Reader which will itself consume data
beyond the end of the LZW stream, and this isn't accounted for anywhere.

This change means to avoid the allocation of the bufio.Reader by making
blockReader implement io.ByteReader.  Further, it adds a close() method
which detects extra data in the block sequence.  To avoid any
regressions with poorly encoded GIFs which may have worked accidentally,
there are no restrictions on how many extra bytes may exist in the final
full sub-block that contained LZW data.  If the end of the LZW stream
happened to align with the end of a sub-block, at most one more
sub-block with a length of 1 byte may exist before the block terminator.

This change aims to be at least as performant as the prior
implementation.  But the primary gain is avoiding the allocation of a
bufio.Reader per frame:

name      old time/op    new time/op    delta
Decode-8     276µs ± 0%     275µs ± 2%    ~     (p=0.690 n=5+5)

name      old speed      new speed      delta
Decode-8  55.9MB/s ± 0%  56.3MB/s ± 2%    ~     (p=0.690 n=5+5)

name      old alloc/op   new alloc/op   delta
Decode-8    49.2kB ± 0%    44.8kB ± 0%  -9.10%  (p=0.008 n=5+5)

name      old allocs/op  new allocs/op  delta
Decode-8       269 ± 0%       267 ± 0%  -0.74%  (p=0.008 n=5+5)

Change-Id: Iec4f9b895561ad52266313fbc73ec82c070c3349
Reviewed-on: https://go-review.googlesource.com/68350
Run-TryBot: Emmanuel Odeke <emm.odeke@gmail.com>
Reviewed-by: Nigel Tao <nigeltao@golang.org>

When building test binaries, we build one archive with all of the test
sources and a second archive with the generated test package main and
link them together. If the test sources are themselves in package main
and the test was compiled with non-default compiler flags, then both
archives will contain a go.cuinfo.producer.main symbol, leading to a
duplicate symbol failure.

This has been causing test build failures on darwin-arm-a1428ios,
darwin-arm64-a1549ios, linux-amd64-noopt, android-arm-wiko-fever, and
android-arm64-wiko-fever since CL 71430 added this symbol. This CL
should fix the build.

Change-Id: I69051c846e7c0d97395a865a361cae07f411f9ad
Reviewed-on: https://go-review.googlesource.com/71771
Run-TryBot: Austin Clements <austin@google.com>
Reviewed-by: Than McIntosh <thanm@google.com>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

CL 70310 dropped TODO while moving code.
Add TODO back, so we do not forget.

Change-Id: I3599ac02743bd35fb9556fdc238e9c72cf7f718f
Reviewed-on: https://go-review.googlesource.com/71590Reviewed-by: David Crawshaw <crawshaw@golang.org>

Fixes #22258.

Change-Id: I43e68f1cf3163e1a041ebff2734ff2cb7943f695
Reviewed-on: https://go-review.googlesource.com/71431Reviewed-by: Rob Pike <r@golang.org>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
Reviewed-by: Ian Lance Taylor <iant@golang.org>