]> Cypherpunks repositories - gostls13.git/log
gostls13.git
6 years agocmd/go: add go list -m -f {{.GoMod}} to show path to go.mod file
Russ Cox [Tue, 24 Jul 2018 19:01:22 +0000 (15:01 -0400)]
cmd/go: add go list -m -f {{.GoMod}} to show path to go.mod file

"go env GOMOD" gives this for the main module already
but it's useful to be able to query other modules.
Using {{.Dir}} does not work if the go.mod was auto-synthesized.

Change-Id: If4844571e9e429b541de0d40c36ff4c5743b2031
Reviewed-on: https://go-review.googlesource.com/125656
Reviewed-by: Bryan C. Mills <bcmills@google.com>
6 years agocmd/go: document go env GOMOD
Russ Cox [Tue, 24 Jul 2018 18:41:44 +0000 (14:41 -0400)]
cmd/go: document go env GOMOD

Fixes #26500.

Change-Id: I0a00009ce10f7aab3e6e79f7218307c3008422d1
Reviewed-on: https://go-review.googlesource.com/125655
Reviewed-by: Bryan C. Mills <bcmills@google.com>
6 years agocmd/cgo: document that #including source files in subdirectories is a bad idea
Russ Cox [Fri, 20 Jul 2018 15:22:21 +0000 (11:22 -0400)]
cmd/cgo: document that #including source files in subdirectories is a bad idea

Suggested in discussion on #26366.

Change-Id: Id9ad2e429a915f88b4c4b30fc415c722eebe0ea4
Reviewed-on: https://go-review.googlesource.com/125297
Reviewed-by: Ian Lance Taylor <iant@golang.org>
6 years agogo/build: invoke go command to find modules during Import, Context.Import
Russ Cox [Fri, 20 Jul 2018 14:59:57 +0000 (10:59 -0400)]
go/build: invoke go command to find modules during Import, Context.Import

The introduction of modules has broken (intentionally) the rule
that the source code for a package x/y/z is in GOPATH/src/x/y/z
(or GOROOT/src/x/y/z). This breaks the code in go/build.Import,
which uses that rule to find the directory for a package.

In the long term, the fix is to move programs that load packages
off of go/build and onto golang.org/x/tools/go/packages, which
we hope will eventually become go/packages. That code invokes
the go command to learn what it needs to know about where
packages are.

In the short term, though, there are lots of programs that use go/build
and will not be able to find code in module dependencies.
To help those programs, go/build now runs the go command to
ask where a package's source code can be found, if it sees that
modules are in use. (If modules are not in use, it falls back to the
usual lookup code and does not invoke the go command, so that
existing uses are unaffected and not slowed down.)

Helps #24661.
Fixes #26504.

Change-Id: I0dac68854cf5011005c3b2272810245d81b7cc5a
Reviewed-on: https://go-review.googlesource.com/125296
Reviewed-by: Michael Matloob <matloob@golang.org>
Reviewed-by: Bryan C. Mills <bcmills@google.com>
6 years agonet/http: document Transport.Proxy's https support
Craig Citro [Fri, 27 Jul 2018 20:39:49 +0000 (13:39 -0700)]
net/http: document Transport.Proxy's https support

The net/http module added support for HTTPS proxies in CL 68550, but the
Transport.Proxy docstring was never updated to reflect this. This (doc-only)
update adds "https" to the list of supported schemes.

Change-Id: I0570fcdae8232bb42d52c4dd739dd09ee8dfd612
Reviewed-on: https://go-review.googlesource.com/126495
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
6 years agosyscall: support Faccessat flags argument
Ian Lance Taylor [Fri, 27 Jul 2018 17:48:18 +0000 (10:48 -0700)]
syscall: support Faccessat flags argument

The Linux kernel faccessat system call does not take a flags parameter.
The flag parameter to the C library faccessat function is implemented in C.
The syscall.Faccessat function takes a flags parameter. In older releases
we have passed the flags parameter to the kernel, which ignored it.
In CL 120015 we started returning an error if any flags were set.
That seems clearly better than ignoring them, but it turns out that some
code was using the flags. The code was previously subtly broken.
Now it is obviously broken. That is better, but we can do better still:
we can implement the flags as the C library does. That is what this CL does.

Change-Id: I259bd6f240c3951e939b81c3032dead3d9c567b4
Reviewed-on: https://go-review.googlesource.com/126415
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
6 years agonet/http: try to document ServeFile security more
Brad Fitzpatrick [Wed, 25 Jul 2018 18:19:51 +0000 (18:19 +0000)]
net/http: try to document ServeFile security more

We've expanded this several times. Try more.

Fixes #18837

Change-Id: I03b699391351a30ee60a15d7aa712c6c66444cf9
Reviewed-on: https://go-review.googlesource.com/125875
Reviewed-by: Ian Lance Taylor <iant@golang.org>
6 years agocmd/link: added test of availability of gcc
Constantin Konstantinidis [Thu, 26 Jul 2018 14:01:34 +0000 (16:01 +0200)]
cmd/link: added test of availability of gcc

The test RuntimeTypeAttr always failed when gcc
was unavailable. The test is duplicated for internal
and external linking. The usual verification
of host linker is added at the beginning of the
external link test.

Fixes #26621

Change-Id: I076d661f854c8a6de8fa5e7b069942a471445047
Reviewed-on: https://go-review.googlesource.com/126075
Run-TryBot: Heschi Kreinick <heschi@google.com>
Reviewed-by: Heschi Kreinick <heschi@google.com>
6 years agosyscall: improve NewCallback documentation and panic message
Jeet Parekh [Fri, 27 Jul 2018 15:52:36 +0000 (15:52 +0000)]
syscall: improve NewCallback documentation and panic message

Fixes #26138

Change-Id: If77b2839bccc600223735df42438a19131cd051c
GitHub-Last-Rev: 64ceaea9f1cb7afb3d552dce0223b818bac1faf9
GitHub-Pull-Request: golang/go#26617
Reviewed-on: https://go-review.googlesource.com/126035
Reviewed-by: Austin Clements <austin@google.com>
Run-TryBot: Austin Clements <austin@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

6 years agocmd/cgo: don't resolve typedefs in -godefs mode
Ian Lance Taylor [Wed, 25 Jul 2018 22:24:27 +0000 (15:24 -0700)]
cmd/cgo: don't resolve typedefs in -godefs mode

In -godefs mode any typedefs that appear in struct fields and the like
will presumably be defined in the input file. If we resolve to the
base type, those cross-references will not work. So for -godefs mode,
keep the Go 1.10 behavior and don't resolve the typedefs in a loop.

Fixes #26644

Change-Id: I48cf72d9eb5016353c43074e6aff6495af326f35
Reviewed-on: https://go-review.googlesource.com/125995
Reviewed-by: Keith Randall <khr@golang.org>
6 years agocrypto/x509: skip TestSystemRoots
Martin Möhrmann [Fri, 20 Jul 2018 14:41:26 +0000 (16:41 +0200)]
crypto/x509: skip TestSystemRoots

cgo and non-cgo code paths can disagree
on the number of root certificates:
=== RUN   TestSystemRoots
--- FAIL: TestSystemRoots (0.31s)
    root_darwin_test.go:31:     cgo sys roots: 93.605184ms
    root_darwin_test.go:32: non-cgo sys roots: 213.998586ms
    root_darwin_test.go:44: got 168 roots
    root_darwin_test.go:44: got 427 roots
    root_darwin_test.go:73: insufficient overlap between cgo and non-cgo roots; want at least 213, have 168
FAIL
exit status 1

Updates #21416
Updates #24652

Change-Id: Idb6d35b17c142dfff79a10cf6b40a42d12f9d17e
Reviewed-on: https://go-review.googlesource.com/125259
Run-TryBot: Martin Möhrmann <moehrmann@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
6 years agocmd/go/internal/par: fix TestWorkParallel retries
Daniel Martí [Fri, 27 Jul 2018 10:24:00 +0000 (11:24 +0100)]
cmd/go/internal/par: fix TestWorkParallel retries

When the test retried multiple times, it reused the same Work variable,
causing in the builders being flaky due to panics. I was able to
immediately reproduce the failure with stress and -race:

$ go test -race -c && stress -p 32 ./par.test -test.run=TestWorkParallel$

/tmp/go-stress909062277
--- FAIL: TestWorkParallel (0.07s)
panic: par.Work.Do: already called Do [recovered]
panic: par.Work.Do: already called Do

Instead, use a new Work variable at each retry. Now, the line above
seems to never fail. Of course, much higher 'stress -p' values will
still result in "does not seem to be parallel" test failures since the
machine lacks resources. But those are test failures, not panics.

Fixes #26642.

Change-Id: I5e962eca7602cf413d911ff5669f56d4f52da5a7
Reviewed-on: https://go-review.googlesource.com/126355
Run-TryBot: Daniel Martí <mvdan@mvdan.cc>
Run-TryBot: Bryan C. Mills <bcmills@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Bryan C. Mills <bcmills@google.com>
6 years agocmd/go: remove unused global var isGoRelease
Yury Smolsky [Fri, 27 Jul 2018 14:01:34 +0000 (17:01 +0300)]
cmd/go: remove unused global var isGoRelease

This variable is not used anymore. It was used in the
TestPackageNotStaleWithTrailingSlash test.

Change-Id: I5e52d4d1d91592dd21e2d9cff96974a49d07f5f8
Reviewed-on: https://go-review.googlesource.com/126376
Run-TryBot: Yury Smolsky <yury@smolsky.by>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
6 years agosyscall: implement pipe() on linux/mips
David Wimmer [Wed, 25 Jul 2018 19:16:07 +0000 (19:16 +0000)]
syscall: implement pipe() on linux/mips

Change the Pipe() function to use the pipe() syscall (which has a unique
calling convention on linux/mips) instead of using pipe2(). This allows
it work on kernels <2.6.27 when pipe2() was introduced.

Change-Id: I65dfbd2a02b64e777a8eb13013d718e356521be6
GitHub-Last-Rev: c483a06168387c2cf151b6419e4e16576fab640a
GitHub-Pull-Request: golang/go#26608
Reviewed-on: https://go-review.googlesource.com/125915
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Vladimir Stefanovic <vladimir.stefanovic@mips.com>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
6 years agocrypto/tls: fix deadlock when Read and Close called concurrently
Minaev Mike [Fri, 26 Jan 2018 09:17:46 +0000 (09:17 +0000)]
crypto/tls: fix deadlock when Read and Close called concurrently

The existing implementation of TLS connection has a deadlock. It occurs
when client connects to TLS server and doesn't send data for
handshake, so server calls Close on this connection. This is because
server reads data under locked mutex, while Close method tries to
lock the same mutex.

Fixes #23518

Change-Id: I4fb0a2a770f3d911036bfd9a7da7cc41c1b27e19
Reviewed-on: https://go-review.googlesource.com/90155
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Filippo Valsorda <filippo@golang.org>
6 years agocmd/go: for missing binary-only package, say where it should be
Ian Lance Taylor [Wed, 25 Jul 2018 17:03:40 +0000 (10:03 -0700)]
cmd/go: for missing binary-only package, say where it should be

Before this CL the user effectively has to guess at the expected
location of a binary-only package. While the location is normally
obvious ($GOPATH/pkg/GOOS_GOARCH/PATH/PKG.a) it is much less so when
building with options that implicitly add an -installsufix option.

Fixes #26590

Change-Id: I753ef54d6dcf733bb456dba65a4a92e4db57a1b0
Reviewed-on: https://go-review.googlesource.com/125818
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Reviewed-by: Bryan C. Mills <bcmills@google.com>
6 years agoruntime: ignore GNU/Linux sigaction errors for signals 32 and 33
Ian Lance Taylor [Wed, 25 Jul 2018 21:09:02 +0000 (14:09 -0700)]
runtime: ignore GNU/Linux sigaction errors for signals 32 and 33

This avoids problems when running under QEMU. It seems that at least
some QEMU versions turn the sigaction implementation into a call to
the C library sigaction function. The C library function will reject
attempts to set the signal handler for signals 32 and 33. Ignore
errors in that case.

Change-Id: Id443a9a32f6fb0ceef5c59a398e7ede30bf71646
Reviewed-on: https://go-review.googlesource.com/125955
Run-TryBot: Ian Lance Taylor <iant@golang.org>
Reviewed-by: Heschi Kreinick <heschi@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

6 years agonet: fix handling of Conns created by Resolver.Dial
Ian Gudger [Tue, 24 Jul 2018 23:19:19 +0000 (16:19 -0700)]
net: fix handling of Conns created by Resolver.Dial

The DNS client in net is documented to treat Conns returned by
Resolver.Dial which implement PacketConn as UDP and those which don't as
TCP regardless of what was requested. golang.org/cl/37879 changed the
DNS client to assume that the Conn returned by Resolver.Dial was the
requested type which broke compatibility.

Fixes #26573
Updates #16218

Change-Id: Idf4f073a4cc3b1db36a3804898df206907f9c43c
Reviewed-on: https://go-review.googlesource.com/125735
Run-TryBot: Ian Gudger <igudger@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
6 years agodoc: fix a couple of drive-by review comments in FAQ
Rob Pike [Tue, 24 Jul 2018 22:26:01 +0000 (08:26 +1000)]
doc: fix a couple of drive-by review comments in FAQ

Change-Id: I10cc2073e28cefb1b9a10c0ae89d819ad6417d66
Reviewed-on: https://go-review.googlesource.com/125695
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
6 years agonet/http: fix and normalize the [Server.][ListenAnd]Serve[TLS] docs
Filippo Valsorda [Fri, 30 Mar 2018 23:09:59 +0000 (19:09 -0400)]
net/http: fix and normalize the [Server.][ListenAnd]Serve[TLS] docs

The only inaccurate part was the HTTP/2 caveat in Server.ServeTLS, which
only applies to the plain Serve variant.

The restriction implemented in shouldConfigureHTTP2ForServe is not on
the setupHTTP2_ServeTLS codepath because ServeTLS owns the tls.Listener,
so we fix it for the user instead of disabling HTTP/2.

Fixes #24607

Change-Id: Ie5f207d0201f09db27bf81b75535e5f6fdaf91e2
Reviewed-on: https://go-review.googlesource.com/103815
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
6 years agoruntime: traceback from outermost libc call
Keith Randall [Mon, 16 Jul 2018 21:39:40 +0000 (14:39 -0700)]
runtime: traceback from outermost libc call

If we're in a libc call and get a trap, don't try to traceback the libc call.
Start from the state we had at entry to libc.

If there are multiple libc calls outstanding, remember the outermost one.

Fixes #26393

Change-Id: Icfe8794b95bf3bfd1a0679b456dcde2481dcabf3
Reviewed-on: https://go-review.googlesource.com/124195
Reviewed-by: Austin Clements <austin@google.com>
Run-TryBot: Austin Clements <austin@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

6 years agocmd/compile: set stricter inlining threshold in large functions
Keith Randall [Mon, 23 Jul 2018 20:09:48 +0000 (13:09 -0700)]
cmd/compile: set stricter inlining threshold in large functions

If we're compiling a large function, be more picky about how big
the function we're inlining is.  If the function is >5000 nodes,
we lower the inlining threshold from a cost of 80 to 20.

Turns out reflect.Value's cost is exactly 80.  That's the function
at issue in #26546.

20 was chosen as a proxy for "inlined body is smaller than the call would be".
Simple functions still get inlined, like this one at cost 7:

func ifaceIndir(t *rtype) bool {
return t.kind&kindDirectIface == 0
}

5000 nodes was chosen as the big function size.  Here are all the
5000+ node (~~1000+ lines) functions in the stdlib:

5187 cmd/internal/obj/arm (*ctxt5).asmout
6879 cmd/internal/obj/s390x (*ctxtz).asmout
6567 cmd/internal/obj/ppc64 (*ctxt9).asmout
9643 cmd/internal/obj/arm64 (*ctxt7).asmout
5042 cmd/internal/obj/x86 (*AsmBuf).doasm
8768 cmd/compile/internal/ssa rewriteBlockAMD64
8878 cmd/compile/internal/ssa rewriteBlockARM
8344 cmd/compile/internal/ssa rewriteValueARM64_OpARM64OR_20
7916 cmd/compile/internal/ssa rewriteValueARM64_OpARM64OR_30
5427 cmd/compile/internal/ssa rewriteBlockARM64
5126 cmd/compile/internal/ssa rewriteValuePPC64_OpPPC64OR_50
6152 cmd/compile/internal/ssa rewriteValuePPC64_OpPPC64OR_60
6412 cmd/compile/internal/ssa rewriteValuePPC64_OpPPC64OR_70
6486 cmd/compile/internal/ssa rewriteValuePPC64_OpPPC64OR_80
6534 cmd/compile/internal/ssa rewriteValuePPC64_OpPPC64OR_90
6534 cmd/compile/internal/ssa rewriteValuePPC64_OpPPC64OR_100
6534 cmd/compile/internal/ssa rewriteValuePPC64_OpPPC64OR_110
6675 cmd/compile/internal/gc typecheck1
5433 cmd/compile/internal/gc walkexpr
14070 cmd/vendor/golang.org/x/arch/arm64/arm64asm decodeArg

There are a lot more smaller (~1000 node) functions in the stdlib.
The function in #26546 has 12477 nodes.

At some point it might be nice to have a better heuristic for "inlined
body is smaller than the call", a non-cliff way to scale down the cost
as the function gets bigger, doing cheaper inlined calls first, etc.
All that can wait for another release. I'd like to do this CL for
1.11.

Fixes #26546
Update #17566

Change-Id: Idda13020e46ec2b28d79a17217f44b189f8139ac
Reviewed-on: https://go-review.googlesource.com/125516
Run-TryBot: Keith Randall <khr@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: David Chase <drchase@google.com>
6 years agocmd/go/internal/modfetch: run git log with "-c log.showsignature=false"
Baokun Lee [Fri, 20 Jul 2018 17:37:21 +0000 (01:37 +0800)]
cmd/go/internal/modfetch: run git log with "-c log.showsignature=false"

The old version git not support "--no-show-signature", git add this from
v2.10.0.

Fixes golang/go#26501.

Change-Id: Ia6b54488651e8687b08a4d40e092822bf960c4fe
Reviewed-on: https://go-review.googlesource.com/125315
Run-TryBot: Baokun Lee <nototon@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Bryan C. Mills <bcmills@google.com>
6 years agocmd/go: fix Go structs in -json documentation
Yann Hodique [Tue, 24 Jul 2018 06:23:50 +0000 (23:23 -0700)]
cmd/go: fix Go structs in -json documentation

"string" should really be "struct" in the structures describing the module.

Change-Id: I4e9cb12434bd33aa243622380c78e5e297d01d0b
Reviewed-on: https://go-review.googlesource.com/125638
Reviewed-by: Daniel Martí <mvdan@mvdan.cc>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Daniel Martí <mvdan@mvdan.cc>
TryBot-Result: Gobot Gobot <gobot@golang.org>

6 years agonet/rpc: clarify requirements for connections and codecs
Dmitry Vyukov [Tue, 26 Jun 2018 11:28:04 +0000 (13:28 +0200)]
net/rpc: clarify requirements for connections and codecs

1. Connections and codecs need to be partially safe for concurrent use.
   Namely, read side is serialized by one mutex,
   and writing side is serialized by another.
   Current comment says that they need to be fully thread-safe,
   which makes the default implementations (gobClientCodec/gobServerCodec)
   non-conforming.

2. Say that ServerCodec.Close can be called multiple times
   and must be idempotent. Server requires this and gobServerCodec
   accounts for this,  but the requirement is not documented.

Change-Id: Ie877e37891fed28056e3d9d1722edaed8e154067
Reviewed-on: https://go-review.googlesource.com/120818
Reviewed-by: Rob Pike <r@golang.org>
6 years agonet/http: document that Client methods always return *url.Error
Brad Fitzpatrick [Tue, 24 Jul 2018 00:24:49 +0000 (00:24 +0000)]
net/http: document that Client methods always return *url.Error

Updates #9424

Change-Id: If117ba3e7d031f84b30d3a721ef99fe622734de2
Reviewed-on: https://go-review.googlesource.com/125575
Reviewed-by: Ian Lance Taylor <iant@golang.org>
6 years agonet/http: deflake TestRetryRequestsOnError
Brad Fitzpatrick [Mon, 23 Jul 2018 23:42:10 +0000 (23:42 +0000)]
net/http: deflake TestRetryRequestsOnError

There's a 50ms threshold in net/http.Transport that this test
sometimes hitting on slower devices. That was unrelated to what this
test was trying to test. So instead just t.Skip on RoundTrip errors
unless the failure was quick (under 25ms), in which case the error
must've been about something else. Our fast machines should catch
regressions there.

Fixes #25366

Change-Id: Ibe8e2716a5b68558b57d0b8b5c46f38e46a2cba2
Reviewed-on: https://go-review.googlesource.com/125555
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
6 years agocmd/vet: if a function modifies its args, it's not a print wrapper
Ian Lance Taylor [Thu, 19 Jul 2018 21:35:02 +0000 (14:35 -0700)]
cmd/vet: if a function modifies its args, it's not a print wrapper

Fixes #26486
Updates #26555

Change-Id: I402137b796e574e9b085ab54290d1b4ef73d3fcc
Reviewed-on: https://go-review.googlesource.com/125039
Reviewed-by: Russ Cox <rsc@golang.org>
6 years agonet/http: update Serve docs on when HTTP/2 is enabled
Brad Fitzpatrick [Mon, 23 Jul 2018 18:31:23 +0000 (18:31 +0000)]
net/http: update Serve docs on when HTTP/2 is enabled

Contains portions and modified portions of CL 103815

Fixes #24607

Change-Id: Ic330850a0f098f183315f04ea4780eded46c5b77
Reviewed-on: https://go-review.googlesource.com/125515
Reviewed-by: Ian Lance Taylor <iant@golang.org>
6 years agocmd/go/internal/modfetch: do not panic when zip is incorrect
marwan-at-work [Sun, 22 Jul 2018 20:23:29 +0000 (20:23 +0000)]
cmd/go/internal/modfetch: do not panic when zip is incorrect

Fixes #26536

Change-Id: I75d284bd39af0a06e31d18c7f7745c5a8ecbe6ac
GitHub-Last-Rev: 88040a9edb0cb547a498cb8c8b662eb8729eb8bf
GitHub-Pull-Request: golang/go#26537
Reviewed-on: https://go-review.googlesource.com/125436
Reviewed-by: Daniel Martí <mvdan@mvdan.cc>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Run-TryBot: Daniel Martí <mvdan@mvdan.cc>
TryBot-Result: Gobot Gobot <gobot@golang.org>

6 years agomisc/cgo/test: disable some Darwin tests in cgo mode
Austin Clements [Fri, 20 Jul 2018 19:51:40 +0000 (15:51 -0400)]
misc/cgo/test: disable some Darwin tests in cgo mode

For unknown reasons, linking against CoreFoundation on macOS 10.10
sometimes causes mmap to ignore the hint address, which makes the Go
allocator incompatible with TSAN. Currently, the effect of this is to
run the allocator out of arena hints on the very first allocation,
causing a "too many address space collisions for -race mode" panic.

This CL skips the cgo tests that link against CoreFoundation in race
mode.

Updates #26475.
Updates #26513.

Change-Id: I52ec638c99acf5d4966e68ff0054f7679680dac6
Reviewed-on: https://go-review.googlesource.com/125304
Run-TryBot: Austin Clements <austin@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
6 years agotest: add test for gccgo bug #26495
Cherry Zhang [Fri, 20 Jul 2018 18:25:31 +0000 (14:25 -0400)]
test: add test for gccgo bug #26495

Gccgo produced incorrect order of evaluation for expressions
involving &&, || subexpressions. The fix is CL 125299.

Updates #26495.

Change-Id: I18d873281709f3160b3e09f0b2e46f5c120e1cab
Reviewed-on: https://go-review.googlesource.com/125301
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
6 years agodoc: explain minor change to gofmt in go1.11
Daniel Martí [Thu, 5 Jul 2018 13:17:46 +0000 (14:17 +0100)]
doc: explain minor change to gofmt in go1.11

Add the tools section with a Gofmt sub-section, just like in
go1.10.html. Instead of copying the two last paragraphs from 1.10, which
warn users about the hidden complexity of enforcing gofmt, move that to
go/format and link to it.

While at it, remove a duplicate "Tools" header that was likely added by
accident.

Fixes #26228.

Change-Id: Ic511c44b2b86f82a41f2b78dd7e7482d694b6c62
Reviewed-on: https://go-review.googlesource.com/122295
Reviewed-by: Ian Lance Taylor <iant@golang.org>
6 years agocm/go/internal/test: make vet run when using go test -c
LE Manh Cuong [Fri, 20 Jul 2018 06:21:45 +0000 (13:21 +0700)]
cm/go/internal/test: make vet run when using go test -c

Fixes #26451

Change-Id: Icd8d6d55dc42adb5c8953787eec7eb29ba46b2aa
Reviewed-on: https://go-review.googlesource.com/125215
Reviewed-by: Ian Lance Taylor <iant@golang.org>
6 years agonet/http/httptest: guarantee ResponseRecorder.Result returns a non-nil body
Jack [Thu, 19 Jul 2018 12:24:44 +0000 (12:24 +0000)]
net/http/httptest: guarantee ResponseRecorder.Result returns a non-nil body

The doc for ResponseRecorder.Result guarantees that the body of the returned
http.Response will be non-nil, but this only holds true if the caller's body is
non-nil. With this change, if the caller's body is nil then the returned
response's body will be an empty io.ReadCloser.

Fixes #26442

Change-Id: I3b2fe4a2541caf9997dbb8978bbaf1f58cd1f471
GitHub-Last-Rev: d802967d89e89c50077fb2d0d455163dcea0eb43
GitHub-Pull-Request: golang/go#26453
Reviewed-on: https://go-review.googlesource.com/124875
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>

6 years agocmd/link: skip a couple of DWARF tests in short mode
Than McIntosh [Fri, 20 Jul 2018 11:38:10 +0000 (07:38 -0400)]
cmd/link: skip a couple of DWARF tests in short mode

Rejigger the DWARF tests to ensure that they run in a reasonable
amount of time in short mode, particularly the "abstract origin
sanity" testpoints.

Updates #26470

Change-Id: Idae9763ac20ea999fa394595aacfcd1e271293ae
Reviewed-on: https://go-review.googlesource.com/125295
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Reviewed-by: Austin Clements <austin@google.com>
6 years agodoc/go1.11: fix bad link to Int.ModInverse
Alberto Donizetti [Fri, 20 Jul 2018 08:41:25 +0000 (10:41 +0200)]
doc/go1.11: fix bad link to Int.ModInverse

Change-Id: I659bc5b61dade9438b122e6994da022946f5e1d3
Reviewed-on: https://go-review.googlesource.com/125255
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
6 years agodoc/go1.11: fix unclosed html and typos in net section
Alberto Donizetti [Fri, 20 Jul 2018 09:00:28 +0000 (11:00 +0200)]
doc/go1.11: fix unclosed html and typos in net section

Change-Id: I3bcf8850ad3873f2627ba017cbfb8b7a8c9cf467
Reviewed-on: https://go-review.googlesource.com/125256
Reviewed-by: Daniel Martí <mvdan@mvdan.cc>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
6 years agocmd/go: fix module get -u to avoid spurious new deps
Russ Cox [Thu, 19 Jul 2018 00:08:04 +0000 (20:08 -0400)]
cmd/go: fix module get -u to avoid spurious new deps

If we have go get -u x1@v1 x2@v2 and x1 depends on x2,
use v2 as the "upgraded" x2 chosen by -u  instead of
letting -u pick something (say, v2.1) and then immediately
overriding it. This avoids chasing down the deps from v2.1
and also avoids them polluting the overall module graph.

This fix also lets us delete some code in the preparation step,
reducing the overall latency of get -u.

Suggested by Bryan Mills in
https://go-review.googlesource.com/c/vgo/+/122396/6#371.

Fixes #26342.

Change-Id: I50fa842304820d3f16f66a8e65dea695e2b0f88b
Reviewed-on: https://go-review.googlesource.com/124856
Reviewed-by: Bryan C. Mills <bcmills@google.com>
6 years agocmd/go: detect inconsistent 'go get' version requests
Russ Cox [Wed, 18 Jul 2018 21:15:32 +0000 (17:15 -0400)]
cmd/go: detect inconsistent 'go get' version requests

If x v1.0.0 requires y v1.2.0, then
go get x@v1.0.0 y@v1.0.0 needs to fail gracefully.

Fixes #25917.

Change-Id: I9b426af23a30310fcb0c3545a8d97feb58b8ddbe
Reviewed-on: https://go-review.googlesource.com/124800
Reviewed-by: Bryan C. Mills <bcmills@google.com>
6 years agocmd/go: case-encode versions as well as module paths in files, URLs
Russ Cox [Wed, 18 Jul 2018 20:09:36 +0000 (16:09 -0400)]
cmd/go: case-encode versions as well as module paths in files, URLs

While writing the GOPROXY docs it occurred to me that versions
can contain upper-case letters as well. The docs therefore say
that versions are case-encoded the same as paths in the proxy
protocol (and therefore in the cache as well). Make it so.

Change-Id: Ibc0c4af0192a4af251e5dd6f2d36cda7e529099a
Reviewed-on: https://go-review.googlesource.com/124795
Reviewed-by: Bryan C. Mills <bcmills@google.com>
6 years agocrypto/aes: implement AES-GCM AEAD for arm64
Vlad Krasnov [Sat, 14 Apr 2018 04:01:02 +0000 (04:01 +0000)]
crypto/aes: implement AES-GCM AEAD for arm64

Use the dedicated AES* and PMULL* instructions to accelerate AES-GCM

name              old time/op    new time/op      delta
AESGCMSeal1K-46     12.1µs ± 0%       0.9µs ± 0%    -92.66%  (p=0.000 n=9+10)
AESGCMOpen1K-46     12.1µs ± 0%       0.9µs ± 0%    -92.43%  (p=0.000 n=10+10)
AESGCMSign8K-46     58.6µs ± 0%       2.1µs ± 0%    -96.41%  (p=0.000 n=9+8)
AESGCMSeal8K-46     92.8µs ± 0%       5.7µs ± 0%    -93.86%  (p=0.000 n=9+9)
AESGCMOpen8K-46     92.9µs ± 0%       5.7µs ± 0%    -93.84%  (p=0.000 n=8+9)

name              old speed      new speed        delta
AESGCMSeal1K-46   84.7MB/s ± 0%  1153.4MB/s ± 0%  +1262.21%  (p=0.000 n=9+10)
AESGCMOpen1K-46   84.4MB/s ± 0%  1115.2MB/s ± 0%  +1220.53%  (p=0.000 n=10+10)
AESGCMSign8K-46    140MB/s ± 0%    3894MB/s ± 0%  +2687.50%  (p=0.000 n=9+10)
AESGCMSeal8K-46   88.2MB/s ± 0%  1437.5MB/s ± 0%  +1529.30%  (p=0.000 n=9+9)
AESGCMOpen8K-46   88.2MB/s ± 0%  1430.5MB/s ± 0%  +1522.01%  (p=0.000 n=8+9)

This change mirrors the current amd64 implementation, and provides optimal performance
on a range of arm64 processors including Centriq 2400 and Apple A12. By and large it is
implicitly tested by the robustness of the already existing amd64 implementation.

The implementation interleaves GHASH with CTR mode to achieve the highest possible
throughput, it also aggregates GHASH with a factor of 8, to decrease the cost of the
reduction step.

Even thought there is a significant amount of assembly, the code reuses the go
code for the amd64 implementation, so there is little additional go code.

Since AES-GCM is critical for performance of all web servers, this change is
required to level the playfield for arm64 CPUs, where amd64 currently enjoys an
unfair advantage.

Ideally both amd64 and arm64 codepaths could be replaced by hypothetical AES and
CLMUL intrinsics, with a few additional vector instructions.

Fixes #18498
Fixes #19840

Change-Id: Icc57b868cd1f67ac695c1ac163a8e215f74c7910
Reviewed-on: https://go-review.googlesource.com/107298
Run-TryBot: Vlad Krasnov <vlad@cloudflare.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
6 years agocmd/go: fix TestScript/list_test_e on Windows go1.11beta2
Bryan C. Mills [Thu, 19 Jul 2018 20:59:50 +0000 (16:59 -0400)]
cmd/go: fix TestScript/list_test_e on Windows

I forgot to run trybots on CL 123758, and the test failed on Windows because I
hard-coded a slash-delimited path.

Use the tent-in-a-box operator ([/\\]) to make the path platform-agnostic.

Change-Id: I9113ab60d21152c11e2ebdf822b58a44b1b38574
Reviewed-on: https://go-review.googlesource.com/125115
Run-TryBot: Bryan C. Mills <bcmills@google.com>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
6 years agoapi: add ErrorHandler function to httputil.ReverseProxy
Andrew Bonventre [Thu, 19 Jul 2018 20:31:12 +0000 (16:31 -0400)]
api: add ErrorHandler function to httputil.ReverseProxy

Change-Id: Ie706c43c9a11cf30ca0f2139a4e9eabac8c75c75
Reviewed-on: https://go-review.googlesource.com/125096
Run-TryBot: Andrew Bonventre <andybons@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
6 years agocmd/compile: run fmt test only in long mode
Keith Randall [Thu, 19 Jul 2018 20:19:54 +0000 (13:19 -0700)]
cmd/compile: run fmt test only in long mode

Update #26469

Change-Id: Id8b8d1c0db48374d5d3dc663a77187a73f60c9a5
Reviewed-on: https://go-review.googlesource.com/125037
Run-TryBot: Keith Randall <khr@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Reviewed-by: David Chase <drchase@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

6 years agocmd/go/internal/list: do not examine TestmainGo if the test fails to load
Bryan C. Mills [Fri, 13 Jul 2018 21:24:05 +0000 (17:24 -0400)]
cmd/go/internal/list: do not examine TestmainGo if the test fails to load

Fixes #25980.

Change-Id: I33c8e72871ffbb1cd1babbcd5dabc8cac0bbcaed
Reviewed-on: https://go-review.googlesource.com/123758
Reviewed-by: Russ Cox <rsc@golang.org>
6 years agocmd/compile: don't run gc/ssa_test/TestGenFlowGraph in short mode
David Chase [Thu, 19 Jul 2018 19:04:14 +0000 (15:04 -0400)]
cmd/compile: don't run gc/ssa_test/TestGenFlowGraph in short mode

The test runs far too long for -short mode (4 seconds).

Also removed useless test of now-disconnected knob
(GO_SSA_PHI_LOC_CUTOFF), which cuts 4 seconds to 2 seconds (which
is still too long), and finished removing the disconnected knob.

Updates #26469.

Change-Id: I6c594227c4a5aaffee46832049bdbbf570d86e60
Reviewed-on: https://go-review.googlesource.com/125075
Run-TryBot: David Chase <drchase@google.com>
Reviewed-by: Keith Randall <khr@golang.org>
6 years agocmd/go: skip vet when package cannot be build under "go test/vet"
Yury Smolsky [Sun, 15 Jul 2018 15:34:55 +0000 (18:34 +0300)]
cmd/go: skip vet when package cannot be build under "go test/vet"

If the the package cannot be built,
"go test" and "go vet" should not run the "vet" tool.
In that case only errors from the compilers will be displayed.

Fixes #26125

Change-Id: I5da6ba64bae5f44feaf5bd4e765eea85533cddd4
Reviewed-on: https://go-review.googlesource.com/123938
Run-TryBot: Yury Smolsky <yury@smolsky.by>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Bryan C. Mills <bcmills@google.com>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
6 years agocmd/go: fix TestScript on Plan 9
David du Colombier [Thu, 19 Jul 2018 08:41:54 +0000 (10:41 +0200)]
cmd/go: fix TestScript on Plan 9

CL 123577 added TestScript. The install_rebuild_gopath
test was failing on Plan 9 because it defines a GOPATH
using the ':' separator, while Plan 9 expects the '\000'
separator in environment variables.

This change fixes the script engine by defining a new
":" environment variable set to OS-specific path list
separator.

The install_rebuild_gopath test has been updated to use
"${:}" instead of ":".

Fixes #26421.

Change-Id: I58a97f882cdb48cc0836398b0d98a80ea58041ba
Reviewed-on: https://go-review.googlesource.com/124435
Run-TryBot: David du Colombier <0intro@gmail.com>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
6 years agosyscall/js: show goroutine stack traces on deadlock
Richard Musiol [Sat, 14 Jul 2018 10:19:36 +0000 (12:19 +0200)]
syscall/js: show goroutine stack traces on deadlock

When using callbacks, it is not necessarily a deadlock if there is no
runnable goroutine, since a callback might still be pending. If there
is no callback pending, Node.js simply exits with exit code zero,
which is not desired if the Go program is still considered running.
This is why an explicit check on exit is used to trigger the "deadlock"
error. This CL makes it so this is Go's normal "deadlock" error, which
includes the stack traces of all goroutines.

Updates #26382

Change-Id: If88486684d0517a64f570009a5ea0ad082679a54
Reviewed-on: https://go-review.googlesource.com/123936
Run-TryBot: Richard Musiol <neelance@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
6 years agocmd/go: warn about non-use of go.mod in legacy go get
Russ Cox [Thu, 19 Jul 2018 04:56:10 +0000 (00:56 -0400)]
cmd/go: warn about non-use of go.mod in legacy go get

It's important for a smooth transition for non-module users
not to change operation in GOPATH/src by default in Go 1.11,
even if go.mod exists in a downloaded dependency.

Even so, users create go.mod and then are confused about
why 'go get' commands seem to behave oddly, when in fact
they are getting the old 'go get'.

Try to split the difference by printing a warning in 'go get'
when run in a tree that would normally be considered a
module if only it were outside GOPATH/src.

Change-Id: I55a1cbef127f3f36de54a8d7b93e1fc64bf0a708
Reviewed-on: https://go-review.googlesource.com/124859
Reviewed-by: Bryan C. Mills <bcmills@google.com>
6 years agocmd/go/internal/get: fix "mod over non-mod" preference for meta tags
Russ Cox [Wed, 18 Jul 2018 19:45:13 +0000 (15:45 -0400)]
cmd/go/internal/get: fix "mod over non-mod" preference for meta tags

If there was a mod and non-mod meta tag for a given prefix,
the meta tag extractor was already dropping the non-mod meta tag.
But we might have mod and non-mod meta tags with different
prefixes, in which case the mod tag should prevail when both match.

Fixes #26200.

Change-Id: I17ab361338e270b9fa03999ad1954f2bbe0f5017
Reviewed-on: https://go-review.googlesource.com/124714
Reviewed-by: Bryan C. Mills <bcmills@google.com>
6 years agocrypto/elliptic: implement P256 for arm64
Vlad Krasnov [Fri, 9 Mar 2018 01:19:11 +0000 (01:19 +0000)]
crypto/elliptic: implement P256 for arm64

This patch ports the existing optimized P256 implementation to arm64.

name            old time/op    new time/op    delta
pkg:crypto/ecdsa goos:linux goarch:arm64
SignP256           539µs ±13%      43µs ± 2%  -91.95%  (p=0.000 n=20+20)
SignP384          13.2ms ± 1%    13.2ms ± 1%     ~     (p=0.739 n=10+10)
VerifyP256        1.57ms ± 0%    0.12ms ± 0%  -92.40%  (p=0.000 n=18+20)
KeyGeneration      391µs ± 0%      25µs ± 0%  -93.62%  (p=0.000 n=9+9)
pkg:crypto/elliptic goos:linux goarch:arm64
BaseMult          1.66ms ± 0%    1.65ms ± 1%     ~     (p=0.079 n=9+10)
BaseMultP256       389µs ± 0%      22µs ± 1%  -94.28%  (p=0.000 n=19+20)
ScalarMultP256    1.03ms ± 0%    0.09ms ± 0%  -91.25%  (p=0.000 n=19+20)

name            old alloc/op   new alloc/op   delta
pkg:crypto/ecdsa goos:linux goarch:arm64
SignP256          5.47kB ± 0%    3.20kB ± 0%  -41.50%  (p=0.000 n=20+20)
SignP384          2.32MB ± 0%    2.32MB ± 0%     ~     (p=0.739 n=10+10)
VerifyP256        7.65kB ± 4%    0.98kB ± 0%  -87.24%  (p=0.000 n=20+20)
KeyGeneration     1.41kB ± 0%    0.69kB ± 0%  -51.05%  (p=0.000 n=9+10)
pkg:crypto/elliptic goos:linux goarch:arm64
BaseMult            224B ± 0%      224B ± 0%     ~     (all equal)
BaseMultP256      1.12kB ± 0%    0.29kB ± 0%  -74.29%  (p=0.000 n=20+20)
ScalarMultP256    1.59kB ± 7%    0.26kB ± 0%  -83.91%  (p=0.000 n=20+20)

name            old allocs/op  new allocs/op  delta
pkg:crypto/ecdsa goos:linux goarch:arm64
SignP256            67.0 ± 0%      35.0 ± 0%  -47.76%  (p=0.000 n=20+20)
SignP384           17.5k ± 0%     17.5k ± 0%     ~     (p=0.725 n=10+10)
VerifyP256          97.2 ± 3%      17.0 ± 0%  -82.52%  (p=0.000 n=20+20)
KeyGeneration       21.0 ± 0%      13.0 ± 0%  -38.10%  (p=0.000 n=10+10)
pkg:crypto/elliptic goos:linux goarch:arm64
BaseMult            5.00 ± 0%      5.00 ± 0%     ~     (all equal)
BaseMultP256        16.0 ± 0%       6.0 ± 0%  -62.50%  (p=0.000 n=20+20)
ScalarMultP256      19.9 ± 6%       5.0 ± 0%  -74.87%  (p=0.000 n=20+20)

Fixes #22806

Change-Id: I12b343a27e6544189334f99c84242bb59db70a76
Reviewed-on: https://go-review.googlesource.com/121360
Run-TryBot: Vlad Krasnov <vlad@cloudflare.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Vlad Krasnov <vlad@cloudflare.com>
Reviewed-by: Filippo Valsorda <filippo@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
6 years agonet/http: update bundled http2
Brad Fitzpatrick [Thu, 19 Jul 2018 18:01:10 +0000 (18:01 +0000)]
net/http: update bundled http2

Updates http2 to x/net/http2 git rev a680a1efc54 for:

   http2: reject large SETTINGS frames or those with duplicates
   https://golang.org/cl/124735

Change-Id: I2168d1d1eef9c63b1a9c06b514b77fae16f920ed
Reviewed-on: https://go-review.googlesource.com/125036
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
6 years agocmd/go: generate alldocs.go
Russ Cox [Wed, 18 Jul 2018 17:30:33 +0000 (13:30 -0400)]
cmd/go: generate alldocs.go

Change-Id: I5a82aec66332f52e304f647758221b5f30b4e2b6
Reviewed-on: https://go-review.googlesource.com/124701
Reviewed-by: Bryan C. Mills <bcmills@google.com>
6 years agocmd/go/internal/par: actually make par.Work run things in parallel
Russ Cox [Thu, 19 Jul 2018 17:23:32 +0000 (13:23 -0400)]
cmd/go/internal/par: actually make par.Work run things in parallel

This was an unfortunate debugging print introduced
while working on the unfortunately large CL 123576.
At least now we're done with that awfulness.

Change-Id: Ib83db59382a799f649832d22d3c6f039d2ef9d2c
Reviewed-on: https://go-review.googlesource.com/125015
Reviewed-by: Bryan C. Mills <bcmills@google.com>
6 years agocmd/go: scrub go.sum during go mod -sync
Russ Cox [Wed, 18 Jul 2018 19:36:04 +0000 (15:36 -0400)]
cmd/go: scrub go.sum during go mod -sync

go.sum accumulates cruft as modules are added and removed as
direct and indirect dependencies. Instead of exposing all that cruft,
let "go mod -sync" clean it out.

Fixes #26381.

Change-Id: I7c9534cf7cc4579f7f82646d00ff691c87a13c4a
Reviewed-on: https://go-review.googlesource.com/124713
Reviewed-by: Bryan C. Mills <bcmills@google.com>
6 years agocmd/go: diagnose 'go mod' in GOPATH/src better
Russ Cox [Wed, 18 Jul 2018 19:05:17 +0000 (15:05 -0400)]
cmd/go: diagnose 'go mod' in GOPATH/src better

People are (understandably) confused by creating go.mod files in GOPATH/src
and then having the go command not use modules in those directories.
We can't change that behavior (or we'll break non-module users of GOPATH)
but we can force 'go mod' (including 'go mod -init') to fail loudly in that case.

If this is not enough, the next step would be to print a warning every time
the go command is run in a GOPATH/src directory with a go.mod but
module mode hasn't triggered. But that will annoy all the non-module users.
Hopefully anyone confused will eventually run a 'go mod' command of
some kind, which will fail loudly.

Fixes #26365.

Change-Id: I8c5fe987fbc3f8d2eceb1138e6862a391ade150c
Reviewed-on: https://go-review.googlesource.com/124708
Reviewed-by: Bryan C. Mills <bcmills@google.com>
6 years agocmd/go: document $GOPROXY, other module adjustments
Russ Cox [Wed, 18 Jul 2018 18:40:20 +0000 (14:40 -0400)]
cmd/go: document $GOPROXY, other module adjustments

Also document module use of GOPATH including GOPATH/src/mod
and GOPATH/bin (unless GOBIN is set).

Fixes #26399.
Fixes #26406.

Change-Id: I7be8eaf110f4fa6fc76ea4cd39aea3dd8addf0b0
Reviewed-on: https://go-review.googlesource.com/124707
Reviewed-by: Bryan C. Mills <bcmills@google.com>
6 years agocmd/go/internal/modget: fix get pkg@none
Russ Cox [Wed, 18 Jul 2018 20:18:29 +0000 (16:18 -0400)]
cmd/go/internal/modget: fix get pkg@none

Now pkg@none actually removes the pkg instead of dying.

For #26342.

Change-Id: I9df7281ed8fd24480109b36f33a563f92e279244
Reviewed-on: https://go-review.googlesource.com/124796
Reviewed-by: Bryan C. Mills <bcmills@google.com>
6 years agocmd/go: convert final module tests to scripts
Russ Cox [Wed, 18 Jul 2018 17:23:17 +0000 (13:23 -0400)]
cmd/go: convert final module tests to scripts

Change-Id: Iba68b3aaf4a132bd4ca44edf4912a46549d2ef8f
Reviewed-on: https://go-review.googlesource.com/124700
Reviewed-by: Bryan C. Mills <bcmills@google.com>
6 years agocmd/go/internal/modfetch: move to new pseudo-version design
Russ Cox [Tue, 17 Jul 2018 14:16:38 +0000 (10:16 -0400)]
cmd/go/internal/modfetch: move to new pseudo-version design

The original pseudo-version design used versions of the form

v0.0.0-yyyymmddhhmmss-abcdef123456

These were intentionally chosen to be valid semantic versions
that sort below any explicitly-chosen semantic version (even v0.0.0),
so that they could be used before anything was tagged but after
that would essentially only be useful in replace statements
(because the max operation during MVS would always prefer
a tagged version).

Then we changed the go command to accept hashes on the
command line, so that you can say

go get github.com/my/proj@abcdef

and it will download and use v0.0.0-yyyymmddhhmmss-abcdef123456.

If you were using v1.10.1 before and this commit is just little bit
newer than that commit, calling it v0.0.0-xxx is confusing but
also harmful: the go command sees the change from v1.10.1 to
the v0.0.0 pseudoversion as a downgrade, and it downgrades other
modules in the build. In particular if some other module has
a requirement of github.com/my/proj v1.9.0 (or later), the
pseudo-version appears to be before that, so go get would
downgrade that module too. It might even remove it entirely,
if every available version needs a post-v0.0.0 version of my/proj.

This CL introduces new pseudo-version forms that can be used
to slot in after the most recent explicit tag before the commit.
If the most recent tagged commit before abcdef is v1.10.1,
then now we will use

v1.10.2-0.yyyymmddhhmmss-abcdef123456

This has the right properties for downgrades and the like,
since it is after v1.10.1 but before almost any possible
successor, such as v1.10.2, v1.10.2-1, or v1.10.2-pre.

This CL also uses those pseudo-version forms as appropriate
when mapping a hash to a pseudo-version. This fixes the
downgrade problem.

Overall, this CL reflects our growing recognition of pseudo-versions
as being like "untagged prereleases".

Issue #26150 was about documenting best practices for how
to work around this kind of accidental downgrade problem
with additional steps. Now there are no additional steps:
the problem is avoided by default.

Fixes #26150.

Change-Id: I402feeccb93e8e937bafcaa26402d88572e9b14c
Reviewed-on: https://go-review.googlesource.com/124515
Reviewed-by: Bryan C. Mills <bcmills@google.com>
6 years agocmd/go/internal/module: add new +incompatible version build annotation
Russ Cox [Mon, 16 Jul 2018 15:38:57 +0000 (11:38 -0400)]
cmd/go/internal/module: add new +incompatible version build annotation

Repos written before the introduction of semantic import versioning
introduced tags like v2.0.0, v3.0.0, and so on, expecting that
(1) the import path would remain unchanged, and perhaps also
(2) there would be at most one copy of the package in a build.

We've always accommodated these by mapping them into the
v0/v1 version range, so that if you ran

    go get k8s.io/client-go@v8.0.0

it would not complain about v8.x.x being a non-v1 version and
instead would map that version to a pseudo-version in go.mod:

    require k8s.io/client-go v0.0.0-20180628043050-7d04d0e2a0a1

The pseudo-version fails to capture two important facts: first,
that this really is the v8.0.0 tag, and second, that it should be
preferred over any earlier v1 tags.

A related problem is that running "go get k8s.io/client-go"
with no version will choose the latest v1 tag (v1.5.1), which
is obsolete.

This CL introduces a new version suffix +incompatible that
indicates that the tag should be considered an (incompatible)
extension of the v1 version sequence instead of part of its
own major version with its own versioned module path.
The requirement above can now be written:

    require k8s.io/client-go v8.0.0+incompatible

(The +metadata suffix is a standard part of semantic versioning,
and that suffix is ignored when comparing two versions for
precedence or equality. As part of canonicalizing versions
recorded in go.mod, the go command has always stripped all
such suffixes. It still strips nearly all: only +incompatible is
preserved now.)

In addition to recognizing the +incompatible, the code that
maps a commit hash to a version will use that form when
appropriate, so that

    go get k8s.io/client-go@7d04d0

will choose k8s.io/client-go@v8.0.0+incompatible.

Also, the code that computes the list of available versions from
a given source code repository also maps old tags to +incompatible
versions, for any tagged commit in which a go.mod file does not exist.
Therefore

    go list -m -versions k8s.io/client-go@latest

will show

    k8s.io/client-go v1.4.0 v1.5.0 v1.5.1 v2.0.0-alpha.0+incompatible ... v8.0.0+incompatible

and similarly

    go get k8s.io/client-go

will now choose v8.0.0+incompatible as the meaning of "latest tagged version".

The extraction of +incompatible versions from source code repos
depends on a codehost.Repo method ReadFileRevs, to do a bulk read
of multiple revisions of a file. That method is only implemented for git in this CL.
Future CLs will need to add support for that method to the other repository
implementations.

Documentation for this change is in CL 124515.

Fixes #26238.

Change-Id: I5bb1d7a46b5fffde34a3c0e6f8d19d9608188cea
Reviewed-on: https://go-review.googlesource.com/124384
Run-TryBot: Russ Cox <rsc@golang.org>
Reviewed-by: Bryan C. Mills <bcmills@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

6 years agocmd/go: preserve %SYSTEMROOT% in TestScript on Windows
Russ Cox [Thu, 19 Jul 2018 13:31:55 +0000 (09:31 -0400)]
cmd/go: preserve %SYSTEMROOT% in TestScript on Windows

Windows networking doesn't work without this environment variable (#25210).

Re-enable TestScript on Windows, and fix two minor failures.

Fixes #26457.

Change-Id: Id9bea49dfb58403195c29c3d831a532ef0f9a233
Reviewed-on: https://go-review.googlesource.com/124858
Run-TryBot: Russ Cox <rsc@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
6 years agoruntime: don't create heap hints outside TSAN's supported heap
Austin Clements [Fri, 13 Jul 2018 21:29:39 +0000 (17:29 -0400)]
runtime: don't create heap hints outside TSAN's supported heap

TSAN for Go only supports heap address in the range [0x00c000000000,
0x00e000000000). However, we currently create heap hints of the form
0xXXc000000000 for XX between 0x00 and 0x7f. Even for XX=0x01, this
hint is outside TSAN's supported heap address range.

Fix this by creating a slightly different set of hints in race mode,
all of which fall inside TSAN's heap address range.

This should fix TestArenaCollision flakes. That test forces the
runtime to use later heap hints. Currently, this always results in
TSAN "failed to allocate" failures on Windows (which happens to have a
slightly more constrained TSAN layout than non-Windows). Most of the
time we don't notice these failures, but sometimes it crashes TSAN,
leading to a test failure.

Fixes #25698.

Change-Id: I8926cd61f0ee5ee00efa77b283f7b809c555be46
Reviewed-on: https://go-review.googlesource.com/123780
Run-TryBot: Austin Clements <austin@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Rick Hudson <rlh@golang.org>
6 years agocmd/compile: add test for OPmodify ops clobbering flags
Keith Randall [Tue, 17 Jul 2018 22:07:26 +0000 (15:07 -0700)]
cmd/compile: add test for OPmodify ops clobbering flags

Code fix was in CL 122556.  This is a corresponding test case.

Fixes #26426

Change-Id: Ib8769f367aed8bead029da0a8d2ddccee1d1dccb
Reviewed-on: https://go-review.googlesource.com/124535
Run-TryBot: Keith Randall <khr@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
6 years agodoc: remove paragraph about text/template/parse
Daniel Martí [Thu, 19 Jul 2018 10:06:56 +0000 (11:06 +0100)]
doc: remove paragraph about text/template/parse

The backwards incompatible changes were undone in CL 120355, while still
preserving the additions needed for assignments in templates to work.

Change-Id: Ie76a798916ef36509c88e171a04bb2cf2a3d7e8e
Reviewed-on: https://go-review.googlesource.com/124917
Reviewed-by: Andrew Bonventre <andybons@golang.org>
6 years agocmd/compile: avoid compressed dwarf when testing for gdb on OSX
David Chase [Wed, 18 Jul 2018 15:23:56 +0000 (11:23 -0400)]
cmd/compile: avoid compressed dwarf when testing for gdb on OSX

Until we figure out how to deal with gdb on Darwin (doesn't
read compressed DWARF from binaries), avoid compressing
DWARF in that case so that the test will still yield meaningful
results.

This is also reported to be a problem for Windows.

Problem also exists for lldb, but this test doesn't check
lldb.

Updates #25925

Change-Id: I85c0e5db75f3329957290500626a3ac7f078f608
Reviewed-on: https://go-review.googlesource.com/124712
Run-TryBot: David Chase <drchase@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Heschi Kreinick <heschi@google.com>
6 years agocmd/go: skip TestScript/mod_* on Windows
Russ Cox [Thu, 19 Jul 2018 05:37:21 +0000 (01:37 -0400)]
cmd/go: skip TestScript/mod_* on Windows

I don't know why it's failing.
Filed #26457.

Change-Id: I84833293a572c5a1a25135bd01cb88518fc7441e
Reviewed-on: https://go-review.googlesource.com/124857
Run-TryBot: Russ Cox <rsc@golang.org>
Reviewed-by: Russ Cox <rsc@golang.org>
6 years agocmd/go/internal/str: simplify HasPathPrefix by epsilon
Russ Cox [Wed, 18 Jul 2018 20:20:19 +0000 (16:20 -0400)]
cmd/go/internal/str: simplify HasPathPrefix by epsilon

Pointed out in CL 122396.
An empty prefix has already been handled above.

Change-Id: Ib94df0a9c8c0517f932b90126232111caa9ad289
Reviewed-on: https://go-review.googlesource.com/124797
Reviewed-by: Bryan C. Mills <bcmills@google.com>
6 years agocmd/go: convert still more module tests to scripts
Russ Cox [Wed, 18 Jul 2018 13:08:45 +0000 (09:08 -0400)]
cmd/go: convert still more module tests to scripts

Change-Id: I249bb848c9911948dbd84cd88ad043a61ed6ea6b
Reviewed-on: https://go-review.googlesource.com/124699
Reviewed-by: Bryan C. Mills <bcmills@google.com>
6 years agocmd/go: convert even more module tests to scripts
Russ Cox [Wed, 18 Jul 2018 03:46:12 +0000 (23:46 -0400)]
cmd/go: convert even more module tests to scripts

Change-Id: Iba185e00e9df2462e9089566053f6c64e24a6a92
Reviewed-on: https://go-review.googlesource.com/124698
Reviewed-by: Bryan C. Mills <bcmills@google.com>
6 years agocmd/go: convert more module tests to scripts
Russ Cox [Wed, 18 Jul 2018 02:59:51 +0000 (22:59 -0400)]
cmd/go: convert more module tests to scripts

Change-Id: I8a36fad061bdf9a19f40531511f3f5717db13b60
Reviewed-on: https://go-review.googlesource.com/124697
Run-TryBot: Russ Cox <rsc@golang.org>
Reviewed-by: Bryan C. Mills <bcmills@google.com>
6 years agocmd/go: convert module tests to scripts
Russ Cox [Wed, 18 Jul 2018 02:35:10 +0000 (22:35 -0400)]
cmd/go: convert module tests to scripts

Change-Id: If0976d15027db795f1383ef709c49c838cbb6953
Reviewed-on: https://go-review.googlesource.com/124696
Run-TryBot: Russ Cox <rsc@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Bryan C. Mills <bcmills@google.com>
6 years agonet/http: deflake TestClientTimeoutKillsConn_AfterHeaders
Brad Fitzpatrick [Wed, 18 Jul 2018 21:37:56 +0000 (21:37 +0000)]
net/http: deflake TestClientTimeoutKillsConn_AfterHeaders

It was flaky on slower machines.

Per report at https://github.com/golang/go/issues/23399#issuecomment-405792381

Change-Id: I7cab02821f78b5ce02ea51089d7eb51723f9705f
Reviewed-on: https://go-review.googlesource.com/124835
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
6 years agodoc: update the Origins section of the FAQ
Rob Pike [Wed, 18 Jul 2018 05:10:04 +0000 (15:10 +1000)]
doc: update the Origins section of the FAQ

Completely replace the opener, which had become not only stale
but bad, expand the discussion of the gopher, and generally provide
prose more connected to the present than to the programming world
of 2007.

Fixes #26107

Change-Id: I5e72f0c81e71d1237fe142dc26114991329a6996
Reviewed-on: https://go-review.googlesource.com/124616
Reviewed-by: Ian Lance Taylor <iant@golang.org>
6 years agodoc: fill in final standard library TODOs in go1.11.html
Ian Lance Taylor [Wed, 18 Jul 2018 22:29:32 +0000 (15:29 -0700)]
doc: fill in final standard library TODOs in go1.11.html

Change-Id: Ic1ff580573711a6c91c1d5e3eb019a298a2fec49
Reviewed-on: https://go-review.googlesource.com/124837
Reviewed-by: Ian Lance Taylor <iant@golang.org>
6 years agodoc: update go1.11 release notes
Brad Fitzpatrick [Wed, 18 Jul 2018 21:57:36 +0000 (21:57 +0000)]
doc: update go1.11 release notes

Change-Id: I9008afdc8c38c440ea083a4f2bed0d2253e112f0
Reviewed-on: https://go-review.googlesource.com/124836
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
6 years agodoc: update go1.11 release notes
Andrew Bonventre [Wed, 18 Jul 2018 20:49:23 +0000 (16:49 -0400)]
doc: update go1.11 release notes

Change-Id: I3e2a19fe90334f0c1ed1593c7a9a3a458f15d8e8
Reviewed-on: https://go-review.googlesource.com/124799
Reviewed-by: Andrew Bonventre <andybons@golang.org>
6 years agodoc: update go1.11 release notes
Andrew Bonventre [Wed, 18 Jul 2018 20:44:07 +0000 (16:44 -0400)]
doc: update go1.11 release notes

Change-Id: Ib488a78802ad730e7c6b3618eab24c259f4bebd1
Reviewed-on: https://go-review.googlesource.com/124798
Reviewed-by: Andrew Bonventre <andybons@golang.org>
6 years agodoc/go1.11: add some links to text/scanner, remove parens
Brad Fitzpatrick [Wed, 18 Jul 2018 20:20:56 +0000 (20:20 +0000)]
doc/go1.11: add some links to text/scanner, remove parens

Change-Id: I30d2b4b94f26300f2cf7b4ecd328a4875d69db51
Reviewed-on: https://go-review.googlesource.com/124777
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
6 years agodoc/go1.11: update runtime/trace to use region instead of span
Hana (Hyang-Ah) Kim [Wed, 18 Jul 2018 20:26:13 +0000 (16:26 -0400)]
doc/go1.11: update runtime/trace to use region instead of span

The final API uses 'region' instead of 'span' from the proposal.

Change-Id: I305da891a360596fff89b10bc6de3090289b5396
Reviewed-on: https://go-review.googlesource.com/124815
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
6 years agodoc: add a couple of links to go1.11 release notes
Ian Lance Taylor [Wed, 18 Jul 2018 20:30:25 +0000 (13:30 -0700)]
doc: add a couple of links to go1.11 release notes

Change-Id: Ie4e81b88cc8035fddf9c074363a1b35bcae3d470
Reviewed-on: https://go-review.googlesource.com/124778
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
6 years agocmd/compile: fix crash on invalid struct literal
Daniel Martí [Tue, 17 Jul 2018 13:38:52 +0000 (14:38 +0100)]
cmd/compile: fix crash on invalid struct literal

If one tries to use promoted fields in a struct literal, the compiler
errors correctly. However, if the embedded fields are of struct pointer
type, the field.Type.Sym.Name expression below panics.

This is because field.Type.Sym is nil in that case. We can simply use
field.Sym.Name in this piece of code though, as it only concerns
embedded fields, in which case what we are after is the field name.

Added a test mirroring fixedbugs/issue23609.go, but with pointer types.

Fixes #26416.

Change-Id: Ia46ce62995c9e1653f315accb99d592aff2f285e
Reviewed-on: https://go-review.googlesource.com/124395
Run-TryBot: Daniel Martí <mvdan@mvdan.cc>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Emmanuel Odeke <emm.odeke@gmail.com>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
6 years agodoc: update go1.11 release notes
Ian Lance Taylor [Wed, 18 Jul 2018 20:19:04 +0000 (13:19 -0700)]
doc: update go1.11 release notes

Change-Id: I806d411c075cdc66322112b6ee5e50f58462bc6b
Reviewed-on: https://go-review.googlesource.com/124776
Reviewed-by: Ian Lance Taylor <iant@golang.org>
6 years agodoc/go1.11: shorten, tidy, and promote user annotation API
Austin Clements [Wed, 18 Jul 2018 20:09:06 +0000 (16:09 -0400)]
doc/go1.11: shorten, tidy, and promote user annotation API

Text based on CL 124655.

Change-Id: I7c4866ce829cb28a4c60cd8ced3ef99047a38c54
Reviewed-on: https://go-review.googlesource.com/124711
Reviewed-by: Austin Clements <austin@google.com>
6 years agofilepath: updates doc to give case where WalkFunc info arg may be nil
Jack [Wed, 18 Jul 2018 19:12:14 +0000 (19:12 +0000)]
filepath: updates doc to give case where WalkFunc info arg may be nil

If a filepath.WalkFunc is called with an non-nil err argument, it's possible
that the info argument will be nil. The comment above filepath.WalkFunc now
reflects this.

Fixes #26425

Change-Id: Ib9963b3344587d2993f1698c5a801f2d1286856b
GitHub-Last-Rev: 553fc266b570d0c47efe12b3b670f88112e3b334
GitHub-Pull-Request: golang/go#26435
Reviewed-on: https://go-review.googlesource.com/124635
Reviewed-by: Ian Lance Taylor <iant@golang.org>
6 years agodoc/go1.11: mention major performance optimizations
Austin Clements [Wed, 18 Jul 2018 19:45:56 +0000 (15:45 -0400)]
doc/go1.11: mention major performance optimizations

Change-Id: I25b93a84996ab1c17d64089b4c2ffabdff3365ec
Reviewed-on: https://go-review.googlesource.com/124710
Reviewed-by: Austin Clements <austin@google.com>
6 years agodoc: update go1.11.html removing TODO about TLS offsets
Andrew Bonventre [Wed, 18 Jul 2018 19:35:12 +0000 (15:35 -0400)]
doc: update go1.11.html removing TODO about TLS offsets

Change-Id: Id06e5139f16cd7a85c59a3dcf2020cf647fcdea0
Reviewed-on: https://go-review.googlesource.com/124709
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Reviewed-by: Andrew Bonventre <andybons@golang.org>
6 years agodoc: update go1.11.html with user annotation API additions
Andrew Bonventre [Wed, 18 Jul 2018 19:16:26 +0000 (15:16 -0400)]
doc: update go1.11.html with user annotation API additions

Change-Id: I357eea0efb04392e1a4671d20190a2052bf548de
Reviewed-on: https://go-review.googlesource.com/124706
Reviewed-by: Andrew Bonventre <andybons@golang.org>
6 years agodoc: update go1.11 release notes
Andrew Bonventre [Wed, 18 Jul 2018 19:04:02 +0000 (15:04 -0400)]
doc: update go1.11 release notes

Update text/scanner and text/template sections.

Change-Id: I1a273e99ff400870053cca63cea68fb7a9b56764
Reviewed-on: https://go-review.googlesource.com/124705
Reviewed-by: Andrew Bonventre <andybons@golang.org>
6 years agodoc/go1.11: remove unnecessary absolute link
Brad Fitzpatrick [Wed, 18 Jul 2018 18:27:20 +0000 (18:27 +0000)]
doc/go1.11: remove unnecessary absolute link

This doesn't auto-deploy to golang.org, only tip.golang.org.

Change-Id: I112743ada2c1393e21edcc9075127f40da9e6270
Reviewed-on: https://go-review.googlesource.com/124755
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
6 years agodoc/go1.11: link to module overview
Russ Cox [Wed, 18 Jul 2018 17:30:41 +0000 (13:30 -0400)]
doc/go1.11: link to module overview

Change-Id: Id381ee73e678ff4b025c1c35512a755ba49d6f81
Reviewed-on: https://go-review.googlesource.com/124702
Run-TryBot: Russ Cox <rsc@golang.org>
Reviewed-by: Andrew Bonventre <andybons@golang.org>
6 years agodoc: fix a link in go1.11.html and flesh out a few sections
Brad Fitzpatrick [Wed, 18 Jul 2018 16:20:49 +0000 (16:20 +0000)]
doc: fix a link in go1.11.html and flesh out a few sections

Change-Id: Ic5b9ccb2772534cee77ffcaeee617c7d5edfb6fd
Reviewed-on: https://go-review.googlesource.com/124715
Reviewed-by: Ian Lance Taylor <iant@golang.org>
6 years agodoc/go1.11: update runtime/pprof change note
Hana (Hyang-Ah) Kim [Wed, 18 Jul 2018 07:06:21 +0000 (03:06 -0400)]
doc/go1.11: update runtime/pprof change note

Mention the change in the behavior of go test -memprofile.

Change-Id: I0384f058298bd8fcfd2d97996464d46b4e419938
Reviewed-on: https://go-review.googlesource.com/124656
Reviewed-by: Ian Lance Taylor <iant@golang.org>
6 years agodoc/go1.11: add space in Go version name
Alberto Donizetti [Wed, 18 Jul 2018 08:53:44 +0000 (10:53 +0200)]
doc/go1.11: add space in Go version name

Missed in CL 124516.

Change-Id: I6488196c8392987d69eca832ab4969aaafe1a26c
Reviewed-on: https://go-review.googlesource.com/124658
Reviewed-by: Ian Lance Taylor <iant@golang.org>
6 years agodoc/go1.11: mention the libSystem.so change for iOS
Elias Naur [Wed, 18 Jul 2018 07:17:19 +0000 (09:17 +0200)]
doc/go1.11: mention the libSystem.so change for iOS

The change to make the runtime use libSystem.so macOS instead of
direct kernel calls applies to iOS as well.

Change-Id: I97ea86452ac5f7433aea58bbd3ff53a2eb2835e0
Reviewed-on: https://go-review.googlesource.com/124657
Reviewed-by: Ian Lance Taylor <iant@golang.org>
6 years agocmd/compile: fix an arm64's comparison bug
Ben Shi [Wed, 18 Jul 2018 11:16:47 +0000 (11:16 +0000)]
cmd/compile: fix an arm64's comparison bug

The arm64 backend generates "TST" for "if uint32(a)&uint32(b) == 0",
which should be "TSTW".

fixes #26438

Change-Id: I7d64c30e3a840b43486bcd10eea2e3e75aaa4857
Reviewed-on: https://go-review.googlesource.com/124637
Run-TryBot: Ben Shi <powerman1st@163.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
6 years agodoc/go1.11: add time package changes to release notes
Alberto Donizetti [Wed, 18 Jul 2018 10:00:16 +0000 (12:00 +0200)]
doc/go1.11: add time package changes to release notes

Change-Id: I6f8d9cc8997128d0076a3a3e82fe5155d53db40d
Reviewed-on: https://go-review.googlesource.com/124659
Reviewed-by: Rob Pike <r@golang.org>
6 years agocmd/cgo: don't report inconsistency error for incomplete typedef
Ian Lance Taylor [Wed, 18 Jul 2018 00:36:15 +0000 (17:36 -0700)]
cmd/cgo: don't report inconsistency error for incomplete typedef

In CLs 122575 and 123177 the cgo tool started explicitly looking up
typedefs. When there are two Go files using import "C", and the first
one has an incomplete typedef and the second one has a complete
version of the same typedef, then we will now record a version of the
first typedef which will not match the recorded version of the second
typedef, producing an "inconsistent definitions" error. Fix this by
silently merging incomplete typedefs with complete ones.

Fixes #26430

Change-Id: I9e629228783b866dd29b5c3a31acd48f6e410a2d
Reviewed-on: https://go-review.googlesource.com/124575
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>