]> Cypherpunks repositories - gostls13.git/log
gostls13.git
6 years agocmd/compile/internal/gc: minor code reorg (cleanup)
Robert Griesemer [Wed, 19 Sep 2018 23:19:05 +0000 (16:19 -0700)]
cmd/compile/internal/gc: minor code reorg (cleanup)

Found while tracking down #26855.

Change-Id: Ice137fe390820ba351e1c7439b6a9a1b3bdc966b
Reviewed-on: https://go-review.googlesource.com/136396
Run-TryBot: Robert Griesemer <gri@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
6 years agocmd/compile: fix error message for &T{} literal mismatch
Robert Griesemer [Wed, 19 Sep 2018 22:55:36 +0000 (15:55 -0700)]
cmd/compile: fix error message for &T{} literal mismatch

See the change and comment in typecheck.go for a detailed explanation.

Fixes #26855.

Change-Id: I7867f948490fc0873b1bd849048cda6acbc36e76
Reviewed-on: https://go-review.googlesource.com/136395
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
6 years agocmd/compile/internal/gc: better names for (b|c|f)conf (cleanup)
Robert Griesemer [Wed, 19 Sep 2018 04:55:35 +0000 (21:55 -0700)]
cmd/compile/internal/gc: better names for (b|c|f)conf (cleanup)

Use String and GoString methods instead of the xconf names
for the numeric conversion routines.

Also, fixed a couple of comments in fmt.go.

Change-Id: I1b8acdd95dbff3fc30273070fbb1ac4860031a3c
Reviewed-on: https://go-review.googlesource.com/136197
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
6 years agocmd/compiler/internal/gc: remove flag from bconv (cleanup)
Robert Griesemer [Wed, 19 Sep 2018 04:15:19 +0000 (21:15 -0700)]
cmd/compiler/internal/gc: remove flag from bconv (cleanup)

Change-Id: I863eb8ef491e1d51b83d8dd3061bf11cbdc74a3e
Reviewed-on: https://go-review.googlesource.com/136196
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
6 years agocmd/compiler/internal/gc: remove flag argument from fconv (cleanup)
Robert Griesemer [Wed, 19 Sep 2018 03:50:04 +0000 (20:50 -0700)]
cmd/compiler/internal/gc: remove flag argument from fconv (cleanup)

The fconv flag arguments were 0, FmtSharp, and FmtSharp|FmtSign.
The 0 value was used for binary representation only, which was
readily available via Mpflt.String. Otherwise, FmtSharp was always
passed. FmtSign was used to print the '+' sign in case of a positive
number and only needed for complex number formatting. Instead
implemented cconv and handled it there.

Change-Id: I1f77282f995be9cfda05efb71a0e027836a9da26
Reviewed-on: https://go-review.googlesource.com/136195
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
6 years agofmt: fix usage of sync.Pool
Joe Tsai [Tue, 18 Sep 2018 19:52:39 +0000 (12:52 -0700)]
fmt: fix usage of sync.Pool

The current usage of sync.Pool is leaky because it stores an arbitrary
sized buffer into the pool. However, sync.Pool assumes that all items in the
pool are interchangeable from a memory cost perspective. Due to the unbounded
size of a buffer that may be added, it is possible for the pool to eventually
pin arbitrarily large amounts of memory in a live-lock situation.

As a simple fix, we just set a maximum size that we permit back into the pool.

We do not need to fix the use of a sync.Pool in scan.go since the free method
has always enforced a maximum capacity since the first commit of the scan logic.

Fixes #27740
Updates #23199

Change-Id: I875278f7dba42625405df36df3e9b028252ce5e3
Reviewed-on: https://go-review.googlesource.com/136116
Reviewed-by: Bryan C. Mills <bcmills@google.com>
Run-TryBot: Bryan C. Mills <bcmills@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

6 years agocmd/go: write an hgrc file in TestMoveHG
Bryan C. Mills [Tue, 18 Sep 2018 18:00:25 +0000 (14:00 -0400)]
cmd/go: write an hgrc file in TestMoveHG

Some variants of Mercurial respond differently to “permission denied” errors
than to “file not found”, and we set HOME to point to an absolute path that may
produce the former instead of the latter.

To discourage Mercurial from trying HOME, give it an explicit (empty)
configuration in the working directory instead.

Change-Id: I82ae99a6892bba7fc3d41b77209ca181d24315e2
Reviewed-on: https://go-review.googlesource.com/136135
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

6 years agoclean: clean mod cache should respect "-n" option.
Tim Xu [Wed, 19 Sep 2018 03:39:46 +0000 (03:39 +0000)]
clean: clean mod cache should respect "-n" option.

Clean mod cache should print remove commands and not run them when with set "-n" option.
Fixes #27458.

Change-Id: I97242cb40c062b347784cdb61653c84a3a7eab44
GitHub-Last-Rev: 5a6f10cad8c5f2c3916a74ca5eea27b1fdd1dc38
GitHub-Pull-Request: golang/go#27710
Reviewed-on: https://go-review.googlesource.com/135695
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/compile/internal/gc: remove redundant for label
Iskander Sharipov [Mon, 17 Sep 2018 22:22:59 +0000 (01:22 +0300)]
cmd/compile/internal/gc: remove redundant for label

Since there are no nested loops and/or switches,
loop label can be removed and "bare continue" can be used.

Change-Id: Id642a0859299e4470af544d59884fec51dbb31ee
Reviewed-on: https://go-review.googlesource.com/135837
Reviewed-by: Keith Randall <khr@golang.org>
Run-TryBot: Iskander Sharipov <iskander.sharipov@intel.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

6 years agocmd/compile/internal/gc: handle arith ops in samesafeexpr
Iskander Sharipov [Mon, 17 Sep 2018 18:37:30 +0000 (21:37 +0300)]
cmd/compile/internal/gc: handle arith ops in samesafeexpr

Teach samesafeexpr to handle arithmetic unary and binary ops.

It makes map lookup optimization possible in

m[k+1] = append(m[k+1], ...)
m[-k] = append(m[-k], ...)
... etc

Does not cover "+" for strings (concatenation).

Change-Id: Ibbb16ac3faf176958da344be1471b06d7cf33a6c
Reviewed-on: https://go-review.googlesource.com/135795
Run-TryBot: Iskander Sharipov <iskander.sharipov@intel.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
6 years agocmd/compile/internal/ssa: fix `a == a` to `a == b`
Iskander Sharipov [Mon, 17 Sep 2018 22:28:58 +0000 (01:28 +0300)]
cmd/compile/internal/ssa: fix `a == a` to `a == b`

Change-Id: I4ee4f702e1bfc9ad9ea899c255104d5e18cf2c96
Reviewed-on: https://go-review.googlesource.com/135838
Reviewed-by: Keith Randall <khr@golang.org>
Run-TryBot: Iskander Sharipov <iskander.sharipov@intel.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

6 years agocmd/compile: optimize AMD64's bit wise operation
Ben Shi [Mon, 17 Sep 2018 02:05:22 +0000 (02:05 +0000)]
cmd/compile: optimize AMD64's bit wise operation

Currently "arr[idx] |= 0x80" is compiled to MOVLload->BTSL->MOVLstore.
And this CL optimizes it to a single BTSLconstmodify. Other bit wise
operations with a direct memory operand are also implemented.

1. The size of the executable bin/go decreases about 4KB, and the total size
of pkg/linux_amd64 (excluding cmd/compile) decreases about 0.6KB.

2. There a little improvement in the go1 benchmark test (excluding noise).
name                     old time/op    new time/op    delta
BinaryTree17-4              2.66s ± 4%     2.66s ± 3%    ~     (p=0.596 n=49+49)
Fannkuch11-4                2.38s ± 2%     2.32s ± 2%  -2.69%  (p=0.000 n=50+50)
FmtFprintfEmpty-4          42.7ns ± 4%    43.2ns ± 7%  +1.31%  (p=0.009 n=50+50)
FmtFprintfString-4         71.0ns ± 5%    72.0ns ± 3%  +1.33%  (p=0.000 n=50+50)
FmtFprintfInt-4            80.7ns ± 4%    80.6ns ± 3%    ~     (p=0.931 n=50+50)
FmtFprintfIntInt-4          125ns ± 3%     126ns ± 4%    ~     (p=0.051 n=50+50)
FmtFprintfPrefixedInt-4     158ns ± 1%     142ns ± 3%  -9.84%  (p=0.000 n=36+50)
FmtFprintfFloat-4           215ns ± 4%     212ns ± 4%  -1.23%  (p=0.002 n=50+50)
FmtManyArgs-4               519ns ± 3%     510ns ± 3%  -1.77%  (p=0.000 n=50+50)
GobDecode-4                6.49ms ± 6%    6.52ms ± 5%    ~     (p=0.866 n=50+50)
GobEncode-4                5.93ms ± 8%    6.01ms ± 7%    ~     (p=0.076 n=50+50)
Gzip-4                      222ms ± 4%     224ms ± 8%  +0.80%  (p=0.001 n=50+50)
Gunzip-4                   36.6ms ± 5%    36.4ms ± 4%    ~     (p=0.093 n=50+50)
HTTPClientServer-4         59.1µs ± 1%    58.9µs ± 2%  -0.24%  (p=0.039 n=49+48)
JSONEncode-4               9.23ms ± 4%    9.21ms ± 5%    ~     (p=0.244 n=50+50)
JSONDecode-4               48.8ms ± 4%    48.7ms ± 4%    ~     (p=0.653 n=50+50)
Mandelbrot200-4            3.81ms ± 4%    3.80ms ± 3%    ~     (p=0.834 n=50+50)
GoParse-4                  3.20ms ± 5%    3.19ms ± 5%    ~     (p=0.494 n=50+50)
RegexpMatchEasy0_32-4      78.1ns ± 2%    77.4ns ± 3%  -0.86%  (p=0.005 n=50+50)
RegexpMatchEasy0_1K-4       233ns ± 3%     233ns ± 3%    ~     (p=0.074 n=50+50)
RegexpMatchEasy1_32-4      74.2ns ± 3%    73.4ns ± 3%  -1.06%  (p=0.000 n=50+50)
RegexpMatchEasy1_1K-4       369ns ± 2%     364ns ± 4%  -1.41%  (p=0.000 n=36+50)
RegexpMatchMedium_32-4      109ns ± 4%     107ns ± 3%  -2.06%  (p=0.001 n=50+50)
RegexpMatchMedium_1K-4     31.5µs ± 3%    30.8µs ± 3%  -2.20%  (p=0.000 n=50+50)
RegexpMatchHard_32-4       1.57µs ± 3%    1.56µs ± 2%  -0.57%  (p=0.016 n=50+50)
RegexpMatchHard_1K-4       47.4µs ± 4%    47.0µs ± 3%  -0.82%  (p=0.008 n=50+50)
Revcomp-4                   414ms ± 7%     412ms ± 7%    ~     (p=0.285 n=50+50)
Template-4                 64.3ms ± 4%    62.7ms ± 3%  -2.44%  (p=0.000 n=50+50)
TimeParse-4                 316ns ± 3%     313ns ± 3%    ~     (p=0.122 n=50+50)
TimeFormat-4                291ns ± 3%     293ns ± 3%  +0.80%  (p=0.001 n=50+50)
[Geo mean]                 46.5µs         46.2µs       -0.81%

name                     old speed      new speed      delta
GobDecode-4               118MB/s ± 6%   118MB/s ± 5%    ~     (p=0.863 n=50+50)
GobEncode-4               130MB/s ± 9%   128MB/s ± 8%    ~     (p=0.076 n=50+50)
Gzip-4                   87.4MB/s ± 4%  86.8MB/s ± 7%  -0.78%  (p=0.002 n=50+50)
Gunzip-4                  531MB/s ± 5%   533MB/s ± 4%    ~     (p=0.093 n=50+50)
JSONEncode-4              210MB/s ± 4%   211MB/s ± 5%    ~     (p=0.247 n=50+50)
JSONDecode-4             39.8MB/s ± 4%  39.9MB/s ± 4%    ~     (p=0.654 n=50+50)
GoParse-4                18.1MB/s ± 5%  18.2MB/s ± 5%    ~     (p=0.493 n=50+50)
RegexpMatchEasy0_32-4     410MB/s ± 2%   413MB/s ± 3%  +0.86%  (p=0.004 n=50+50)
RegexpMatchEasy0_1K-4    4.39GB/s ± 3%  4.38GB/s ± 3%    ~     (p=0.063 n=50+50)
RegexpMatchEasy1_32-4     432MB/s ± 3%   436MB/s ± 3%  +1.07%  (p=0.000 n=50+50)
RegexpMatchEasy1_1K-4    2.77GB/s ± 2%  2.81GB/s ± 4%  +1.46%  (p=0.000 n=36+50)
RegexpMatchMedium_32-4   9.16MB/s ± 3%  9.35MB/s ± 4%  +2.09%  (p=0.001 n=50+50)
RegexpMatchMedium_1K-4   32.5MB/s ± 3%  33.2MB/s ± 3%  +2.25%  (p=0.000 n=50+50)
RegexpMatchHard_32-4     20.4MB/s ± 3%  20.5MB/s ± 2%  +0.56%  (p=0.017 n=50+50)
RegexpMatchHard_1K-4     21.6MB/s ± 4%  21.8MB/s ± 3%  +0.83%  (p=0.008 n=50+50)
Revcomp-4                 613MB/s ± 4%   618MB/s ± 7%    ~     (p=0.152 n=48+50)
Template-4               30.2MB/s ± 4%  30.9MB/s ± 3%  +2.49%  (p=0.000 n=50+50)
[Geo mean]                127MB/s        128MB/s       +0.64%

Change-Id: If405198283855d75697f66cf894b2bef458f620e
Reviewed-on: https://go-review.googlesource.com/135422
Reviewed-by: Keith Randall <khr@golang.org>
6 years agocmd/compile/internal/gc: simplify `x = x <op> y` to `x <op>= y`
Iskander Sharipov [Mon, 17 Sep 2018 22:04:41 +0000 (01:04 +0300)]
cmd/compile/internal/gc: simplify `x = x <op> y` to `x <op>= y`

Change-Id: I5afba2c10372252be4b65dae7a95461722de904f
Reviewed-on: https://go-review.googlesource.com/135835
Reviewed-by: Keith Randall <khr@golang.org>
Reviewed-by: Robert Griesemer <gri@golang.org>
Run-TryBot: Iskander Sharipov <iskander.sharipov@intel.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

6 years agocmd/compile: simplify AMD64's assembly generator
Ben Shi [Tue, 18 Sep 2018 01:53:42 +0000 (01:53 +0000)]
cmd/compile: simplify AMD64's assembly generator

AMD64's ADDQconstmodify/ADDLconstmodify have similar logic with
other constmodify like operators, but seperated case statements.
This CL simplify them with a fallthrough.

Change-Id: Ia73ffeaddc5080182f68c06c9d9b48fe32a14e38
Reviewed-on: https://go-review.googlesource.com/135855
Run-TryBot: Ben Shi <powerman1st@163.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
6 years agoruntime: support windows/arm
Jordan Rhee [Wed, 8 Aug 2018 21:44:42 +0000 (14:44 -0700)]
runtime: support windows/arm

Updates #26148

Change-Id: I8f68b2c926c7b11dc86c9664ed7ff2d2f78b64b4
Reviewed-on: https://go-review.googlesource.com/128715
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
6 years agobuiltin: document when len and cap are constant
Rob Pike [Mon, 17 Sep 2018 04:25:06 +0000 (14:25 +1000)]
builtin: document when len and cap are constant

The rules are subtle, but under some circumstances the result
can be constant. Mention this and refer to the appropriate
section of the specification.

Fixes #27588.

Change-Id: I4beaad036db87501378fb2ef48d216742d096933
Reviewed-on: https://go-review.googlesource.com/135519
Reviewed-by: Robert Griesemer <gri@golang.org>
6 years agocmd/compile: don't do floating point optimization x+0 -> x
Keith Randall [Mon, 17 Sep 2018 21:46:07 +0000 (14:46 -0700)]
cmd/compile: don't do floating point optimization x+0 -> x

That optimization is not valid if x == -0.

The test is a bit tricky because 0 == -0. We distinguish
0 from -0 with 1/0 == inf, 1/-0 == -inf.

This has been a bug since CL 24790 in Go 1.8. Probably doesn't
warrant a backport.

Fixes #27718

Note: the optimization x-0 -> x is actually valid.
But it's probably best to take it out, so as to not confuse readers.

Change-Id: I99f16a93b45f7406ec8053c2dc759a13eba035fa
Reviewed-on: https://go-review.googlesource.com/135701
Reviewed-by: Cherry Zhang <cherryyz@google.com>
6 years agoruntime: ignore races between close and len/cap
Keith Randall [Mon, 17 Sep 2018 19:25:36 +0000 (12:25 -0700)]
runtime: ignore races between close and len/cap

They aren't really races, or at least they don't have any
observable effect. The spec is silent on whether these are actually
races or not.

Fix this problem by not using the address of len (or of cap)
as the location where channel operations are recorded to occur.
Use a random other field of hchan for that.

I'm not 100% sure we should in fact fix this. Opinions welcome.

Fixes #27070

Change-Id: Ib4efd4b62e0d1ef32fa51e373035ef207a655084
Reviewed-on: https://go-review.googlesource.com/135698
Reviewed-by: Dmitry Vyukov <dvyukov@google.com>
6 years agoruntime: use MADV_FREE on Linux if available
Tobias Klauser [Fri, 14 Sep 2018 07:57:06 +0000 (09:57 +0200)]
runtime: use MADV_FREE on Linux if available

On Linux, sysUnused currently uses madvise(MADV_DONTNEED) to signal the
kernel that a range of allocated memory contains unneeded data. After a
successful call, the range (but not the data it contained before the
call to madvise) is still available but the first access to that range
will unconditionally incur a page fault (needed to 0-fill the range).

A faster alternative is MADV_FREE, available since Linux 4.5. The
mechanism is very similar, but the page fault will only be incurred if
the kernel, between the call to madvise and the first access, decides to
reuse that memory for something else.

In sysUnused, test whether MADV_FREE is supported and fall back to
MADV_DONTNEED in case it isn't. This requires making the return value of
the madvise syscall available to the caller, so change runtime.madvise
to return it.

Fixes #23687

Change-Id: I962c3429000dd9f4a00846461ad128b71201bb04
Reviewed-on: https://go-review.googlesource.com/135395
Run-TryBot: Tobias Klauser <tobias.klauser@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
6 years agocmd/link: fix DWARF refs so that they always point to the typedef entry
Alessandro Arzilli [Mon, 10 Sep 2018 13:36:59 +0000 (15:36 +0200)]
cmd/link: fix DWARF refs so that they always point to the typedef entry

For types defined as:

type typename struct { ... }

the linker produces two DIEs: (1) a DW_TAG_structure_type DIE and (2) a
DW_TAG_typedef_type DIE having (1) as its type attribute.

All subsequent references to 'typename' should use the
DW_TAG_typedef_type DIE, not the DW_TAG_structure_type. Mostly this is
true but sometimes one reference will use the DW_TAG_structure_type
directly. In particular, this happens to the 'first' reference to the
type in question (where 'first' means whatever happens first in the way
the linker scans its symbols).

This isn't only true of struct types: pointer types, array types, etc.
can also be affected.

This fix solves the problem by always returning the typedef DIE in
newtype, when one is created.

Fixes #27614

Change-Id: Ia65b4a1d8c2b752e33a4ebdb74ccd92faa69526e
Reviewed-on: https://go-review.googlesource.com/134555
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Than McIntosh <thanm@google.com>
6 years agoruntime: ignore EAGAIN from exec in TestCgoExecSignalMask
Ian Lance Taylor [Tue, 18 Sep 2018 14:58:11 +0000 (07:58 -0700)]
runtime: ignore EAGAIN from exec in TestCgoExecSignalMask

Fixes #27731

Change-Id: Ifb4d57923b1bba0210ec1f623d779d7b5f442812
Reviewed-on: https://go-review.googlesource.com/135995
Run-TryBot: Ian Lance Taylor <iant@golang.org>
Reviewed-by: Michael Munday <mike.munday@ibm.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

6 years agocmd/go: don't mention -mod=release
Mark Rushakoff [Thu, 30 Aug 2018 02:15:39 +0000 (02:15 +0000)]
cmd/go: don't mention -mod=release

The -mod=release flag is not supported, so this appears to be a
documentation mistake.

Fixes #27354.

Change-Id: I895e8d5b4918adcb1f605361773173f312fa7b65
GitHub-Last-Rev: 42bfe0c11e38c90e76887771654ea81af98d50ec
GitHub-Pull-Request: golang/go#27358
Reviewed-on: https://go-review.googlesource.com/132116
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/compile: use bounded shift information on ppc64x
Lynn Boger [Thu, 13 Sep 2018 18:45:23 +0000 (14:45 -0400)]
cmd/compile: use bounded shift information on ppc64x

Makes use of bounded shift information to generate
more efficient shift instructions.

Updates #25167 for ppc64x

Change-Id: I7fc8d49a3fb3e0f273cc51bc767470b239cbdca7
Reviewed-on: https://go-review.googlesource.com/135380
Run-TryBot: Lynn Boger <laboger@linux.vnet.ibm.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Michael Munday <mike.munday@ibm.com>
6 years agonet/http: fix minor typos in Request godoc
Ivy Evans [Thu, 13 Sep 2018 02:16:27 +0000 (02:16 +0000)]
net/http: fix minor typos in Request godoc

Fixes missing commas where it wasn't immediately apparent whether
"requests" was being used as a verb or a noun.

Change-Id: Ic8c99b4f46475f40a6160d26a3cd11c215940dd5
GitHub-Last-Rev: 1becf6fabeb6f928e37526e96297dd60397ccf9b
GitHub-Pull-Request: golang/go#27649
Reviewed-on: https://go-review.googlesource.com/135135
Reviewed-by: Emmanuel Odeke <emm.odeke@gmail.com>
6 years agocmd/compile/internal/gc: simplify bool expression
Iskander Sharipov [Mon, 17 Sep 2018 22:13:04 +0000 (01:13 +0300)]
cmd/compile/internal/gc: simplify bool expression

Change-Id: Idcd79788e64947a927af662b6394ac7218e62ba8
Reviewed-on: https://go-review.googlesource.com/135836
Reviewed-by: Keith Randall <khr@golang.org>
Reviewed-by: Robert Griesemer <gri@golang.org>
6 years agocmd/compile: fix wrong comment message in AMD64Ops.go
Ben Shi [Sun, 16 Sep 2018 03:05:35 +0000 (03:05 +0000)]
cmd/compile: fix wrong comment message in AMD64Ops.go

According to AMD64.rules, BTS&BTR&BTC use arg1 as the bit index,
while BT uses arg0. This CL fixes the wrong comment message in
AMD64Ops.go, which indicates all bit indexes are in arg0.

Change-Id: Idb78f4d39f7ef5ea78065ad8bc651324597e2a8a
Reviewed-on: https://go-review.googlesource.com/135419
Reviewed-by: Keith Randall <khr@golang.org>
6 years agogo/types: don't report cycle error if clearer error follows
Robert Griesemer [Mon, 17 Sep 2018 20:43:35 +0000 (13:43 -0700)]
go/types: don't report cycle error if clearer error follows

If a cyclic declaration uses a non-type object where it expects
a type, don't report the cycle error in favor of the clearer and
more informative error about the missing type.

Fixes #25790.

Change-Id: If937078383def878efb4c69686e5b4b2a495fd5d
Reviewed-on: https://go-review.googlesource.com/135700
Reviewed-by: Alan Donovan <adonovan@google.com>
6 years agogo/types: simplify some internal code
Robert Griesemer [Mon, 17 Sep 2018 19:46:33 +0000 (12:46 -0700)]
go/types: simplify some internal code

Change-Id: Ia32d40cc272cb049c0a7c9d5f8ef4329bdefc7fe
Reviewed-on: https://go-review.googlesource.com/135699
Run-TryBot: Robert Griesemer <gri@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Alan Donovan <adonovan@google.com>
6 years agocmd/go: display correct options in "go help get" using modules
Eric Ponce [Thu, 30 Aug 2018 00:27:35 +0000 (02:27 +0200)]
cmd/go: display correct options in "go help get" using modules

Fixes: #27298
Change-Id: Icfc6992b470136bb25a77912f670a25883642316
Reviewed-on: https://go-review.googlesource.com/132095
Reviewed-by: Bryan C. Mills <bcmills@google.com>
Run-TryBot: Bryan C. Mills <bcmills@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

6 years agocmd/compile: optimize math.Float64(32)bits and math.Float64(32)frombits on arm64
fanzha02 [Mon, 16 Jul 2018 04:45:25 +0000 (04:45 +0000)]
cmd/compile: optimize math.Float64(32)bits and math.Float64(32)frombits on arm64

Use float <-> int register moves without conversion instead of stores
and loads to move float <-> int values.

Math package benchmark results.
name                 old time/op  new time/op  delta
Acosh                 153ns ± 0%   147ns ± 0%   -3.92%  (p=0.000 n=10+10)
Asinh                 183ns ± 0%   177ns ± 0%   -3.28%  (p=0.000 n=10+10)
Atanh                 157ns ± 0%   155ns ± 0%   -1.27%  (p=0.000 n=10+10)
Atan2                 118ns ± 0%   117ns ± 1%   -0.59%  (p=0.003 n=10+10)
Cbrt                  119ns ± 0%   114ns ± 0%   -4.20%  (p=0.000 n=10+10)
Copysign             7.51ns ± 0%  6.51ns ± 0%  -13.32%  (p=0.000 n=9+10)
Cos                  73.1ns ± 0%  70.6ns ± 0%   -3.42%  (p=0.000 n=10+10)
Cosh                  119ns ± 0%   121ns ± 0%   +1.68%  (p=0.000 n=10+9)
ExpGo                 154ns ± 0%   149ns ± 0%   -3.05%  (p=0.000 n=9+10)
Expm1                 101ns ± 0%    99ns ± 0%   -1.88%  (p=0.000 n=10+10)
Exp2Go                150ns ± 0%   146ns ± 0%   -2.67%  (p=0.000 n=10+10)
Abs                  7.01ns ± 0%  6.01ns ± 0%  -14.27%  (p=0.000 n=10+9)
Mod                   234ns ± 0%   212ns ± 0%   -9.40%  (p=0.000 n=9+10)
Frexp                34.5ns ± 0%  30.0ns ± 0%  -13.04%  (p=0.000 n=10+10)
Gamma                 112ns ± 0%   111ns ± 0%   -0.89%  (p=0.000 n=10+10)
Hypot                73.6ns ± 0%  68.6ns ± 0%   -6.79%  (p=0.000 n=10+10)
HypotGo              77.1ns ± 0%  72.1ns ± 0%   -6.49%  (p=0.000 n=10+10)
Ilogb                31.0ns ± 0%  28.0ns ± 0%   -9.68%  (p=0.000 n=10+10)
J0                    437ns ± 0%   434ns ± 0%   -0.62%  (p=0.000 n=10+10)
J1                    433ns ± 0%   431ns ± 0%   -0.46%  (p=0.000 n=10+10)
Jn                    927ns ± 0%   922ns ± 0%   -0.54%  (p=0.000 n=10+10)
Ldexp                41.5ns ± 0%  37.0ns ± 0%  -10.84%  (p=0.000 n=9+10)
Log                   124ns ± 0%   118ns ± 0%   -4.84%  (p=0.000 n=10+9)
Logb                 34.0ns ± 0%  32.0ns ± 0%   -5.88%  (p=0.000 n=10+10)
Log1p                 110ns ± 0%   108ns ± 0%   -1.82%  (p=0.000 n=10+10)
Log10                 136ns ± 0%   132ns ± 0%   -2.94%  (p=0.000 n=10+10)
Log2                 51.6ns ± 0%  47.1ns ± 0%   -8.72%  (p=0.000 n=10+10)
Nextafter32          33.0ns ± 0%  30.5ns ± 0%   -7.58%  (p=0.000 n=10+10)
Nextafter64          29.0ns ± 0%  26.5ns ± 0%   -8.62%  (p=0.000 n=10+10)
PowInt                169ns ± 0%   160ns ± 0%   -5.33%  (p=0.000 n=10+10)
PowFrac               375ns ± 0%   361ns ± 0%   -3.73%  (p=0.000 n=10+10)
RoundToEven          14.0ns ± 0%  12.5ns ± 0%  -10.71%  (p=0.000 n=10+10)
Remainder             206ns ± 0%   192ns ± 0%   -6.80%  (p=0.000 n=10+9)
Signbit              6.01ns ± 0%  5.51ns ± 0%   -8.32%  (p=0.000 n=10+9)
Sin                  70.1ns ± 0%  69.6ns ± 0%   -0.71%  (p=0.000 n=10+10)
Sincos               99.1ns ± 0%  99.6ns ± 0%   +0.50%  (p=0.000 n=9+10)
SqrtGoLatency         178ns ± 0%   146ns ± 0%  -17.70%  (p=0.000 n=8+10)
SqrtPrime            9.19µs ± 0%  9.20µs ± 0%   +0.01%  (p=0.000 n=9+9)
Tanh                  125ns ± 1%   127ns ± 0%   +1.36%  (p=0.000 n=10+10)
Y0                    428ns ± 0%   426ns ± 0%   -0.47%  (p=0.000 n=10+10)
Y1                    431ns ± 0%   429ns ± 0%   -0.46%  (p=0.000 n=10+9)
Yn                    906ns ± 0%   901ns ± 0%   -0.55%  (p=0.000 n=10+10)
Float64bits          4.50ns ± 0%  3.50ns ± 0%  -22.22%  (p=0.000 n=10+10)
Float64frombits      4.00ns ± 0%  3.50ns ± 0%  -12.50%  (p=0.000 n=10+9)
Float32bits          4.50ns ± 0%  3.50ns ± 0%  -22.22%  (p=0.002 n=8+10)
Float32frombits      4.00ns ± 0%  3.50ns ± 0%  -12.50%  (p=0.000 n=10+10)

Change-Id: Iba829e15d5624962fe0c699139ea783efeefabc2
Reviewed-on: https://go-review.googlesource.com/129715
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

6 years agocmd/trace: don't drop sweep slice details
Hana Kim [Mon, 17 Sep 2018 18:46:50 +0000 (14:46 -0400)]
cmd/trace: don't drop sweep slice details

For sweep events, we used to modify the ViewerEvent returned from
ctx.emitSlice later in order to embed more details about the sweep
operation. The trick no longer works after the change
https://golang.org/cl/92375 and caused a regression.

ctx.emit method encodes the ViewerEvent, so any modification to the
ViewerEvent object after ctx.emit returns will not be reflected.

Refactor ctx.emitSlice, so ctx.makeSlice can be used when producing
slices for SWEEP. ctx.emit* methods are meant to truely emit
ViewerEvents.

Fixes #27711

Change-Id: I0b733ebbbfd4facd8714db0535809ec3cab0833d
Reviewed-on: https://go-review.googlesource.com/135775
Reviewed-by: Austin Clements <austin@google.com>
Run-TryBot: Austin Clements <austin@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

6 years agogo/types: fix a couple of internal comments
Robert Griesemer [Mon, 17 Sep 2018 18:32:04 +0000 (11:32 -0700)]
go/types: fix a couple of internal comments

Change-Id: If0e8fbb05c09ee7c64e1aa6b0aa2ade35a70df8a
Reviewed-on: https://go-review.googlesource.com/135696
Reviewed-by: Alan Donovan <adonovan@google.com>
6 years agosyscall: enable TestSyscallNoError on all Linux 32-bit architectures
Tobias Klauser [Mon, 17 Sep 2018 15:22:48 +0000 (17:22 +0200)]
syscall: enable TestSyscallNoError on all Linux 32-bit architectures

Check the size of uintptr instead of listing GOARCHes explicitly. This
will make the test also run on linux/mips{,le}.

Change-Id: I649f15d293002afc1360b1913910202c3e5188b7
Reviewed-on: https://go-review.googlesource.com/135715
Run-TryBot: Tobias Klauser <tobias.klauser@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
6 years agocmd/compile: fix CMPconstload rule
Keith Randall [Sat, 15 Sep 2018 22:30:13 +0000 (15:30 -0700)]
cmd/compile: fix CMPconstload rule

The CMPconstload opcodes take a ValAndOff as their AuxInt, not just
an offset.

Originally introduced in CL 135379.

Change-Id: I244b2d56ef2e99d2975faa2e97f4291ec97c64b7
Reviewed-on: https://go-review.googlesource.com/135418
Run-TryBot: Keith Randall <khr@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ilya Tocar <ilya.tocar@intel.com>
6 years agocmd/go/testdata/script/mod_test: add missing test invocation for case e
Bryan C. Mills [Thu, 9 Aug 2018 21:56:59 +0000 (17:56 -0400)]
cmd/go/testdata/script/mod_test: add missing test invocation for case e

Change-Id: Ib0544adc1444a473f8edcb9dd92aefa9fcbc7330
Reviewed-on: https://go-review.googlesource.com/134656
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 agocmd/compile: avoid more float32 <-> float64 conversions in compiler
Michael Munday [Wed, 12 Sep 2018 11:16:50 +0000 (12:16 +0100)]
cmd/compile: avoid more float32 <-> float64 conversions in compiler

Use the new custom truncate/extension code when storing or extracting
float32 values from AuxInts to avoid the value being changed by the
host platform's floating point conversion instructions (e.g. sNaN ->
qNaN).

Updates #27516.

Change-Id: Id39650f1431ef74af088c895cf4738ea5fa87974
Reviewed-on: https://go-review.googlesource.com/134855
Run-TryBot: Michael Munday <mike.munday@ibm.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
6 years agocmd/compile/internal/gc: handle array slice self-assign in esc.go
Iskander Sharipov [Tue, 4 Sep 2018 20:14:53 +0000 (23:14 +0300)]
cmd/compile/internal/gc: handle array slice self-assign in esc.go

Instead of skipping all OSLICEARR, skip only ones with non-pointer
array type. For pointers to arrays, it's safe to apply the
self-assignment slicing optimizations.

Refactored the matching code into separate function for readability.

This is an extension to already existing optimization.

On its own, it does not improve any code under std, but
it opens some new optimization opportunities. One
of them is described in the referenced issue.

Updates #7921

Change-Id: I08ac660d3ef80eb15fd7933fb73cf53ded9333ad
Reviewed-on: https://go-review.googlesource.com/133375
Run-TryBot: Iskander Sharipov <iskander.sharipov@intel.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
6 years agocmd/compile/internal/gc: treat cap/len as safe in mayAffectMemory
Iskander Sharipov [Wed, 5 Sep 2018 15:49:52 +0000 (18:49 +0300)]
cmd/compile/internal/gc: treat cap/len as safe in mayAffectMemory

OLEN and OCAP can't affect memory state as long as their
arguments don't.

Re-organized case bodies to avoid duplicating same branches for
recursive invocations.

Change-Id: I30407143429f7dd1891badb70df88969ed267535
Reviewed-on: https://go-review.googlesource.com/133555
Run-TryBot: Iskander Sharipov <iskander.sharipov@intel.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
6 years agoimage/png: minor cleanup of a few tests
Ian Davis [Mon, 10 Sep 2018 08:27:47 +0000 (09:27 +0100)]
image/png: minor cleanup of a few tests

Removes a redundant err check and replaces some returns in a testing
loop with continue to prevent skipping unrelated test cases when
a failure is encountered.

Change-Id: Ic1a560751b95bb0ef8dfa957e057e0fa0c2b281d
Reviewed-on: https://go-review.googlesource.com/134236
Run-TryBot: Daniel Martí <mvdan@mvdan.cc>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Daniel Martí <mvdan@mvdan.cc>
6 years agodoc: add a go/golang entry to the FAQ
Rob Pike [Sun, 16 Sep 2018 01:14:19 +0000 (11:14 +1000)]
doc: add a go/golang entry to the FAQ

It's worth clarifying that the language is called "Go".

Fixes #27616.

Change-Id: Ie4a9cb5e7e6afa437c60e06914125ef7490f27d0
Reviewed-on: https://go-review.googlesource.com/135517
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
6 years agointernal/poll: improve Splice comments
Andrei Tudor Călin [Sat, 15 Sep 2018 16:31:49 +0000 (18:31 +0200)]
internal/poll: improve Splice comments

Clarify the behavior of splice on older kernels, merge comments so
control flow becomes more obvious, as discussed in CL 133575.

Change-Id: I95855991bd0b1fa1c78a900b39c4382f65d83468
Reviewed-on: https://go-review.googlesource.com/135436
Reviewed-by: Ian Lance Taylor <iant@golang.org>
6 years agotime: return ENOENT from androidLoadTzinfoFromTzdata if zone not found
Ian Lance Taylor [Fri, 14 Sep 2018 22:11:51 +0000 (15:11 -0700)]
time: return ENOENT from androidLoadTzinfoFromTzdata if zone not found

This makes Android consistent with the change in CL 121877.

Updates #20969

Change-Id: I1f114556fd1d4654c8e4e6a59513bddd5dc3d1a0
Reviewed-on: https://go-review.googlesource.com/135416
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Reviewed-by: Elias Naur <elias.naur@gmail.com>
6 years agoruntime: fix TODO comment filepath
Koichi Shiraishi [Sat, 15 Sep 2018 08:06:08 +0000 (17:06 +0900)]
runtime: fix TODO comment filepath

The cmd/compile/internal/ld/go.go file not exist, actually cmd/link/internal/ld/go.go.
Also, write line number is not good because it changes every commit of the file.

Change-Id: Id2b9f2c9904390adb011dab357716ee8e2fe84fc
Reviewed-on: https://go-review.googlesource.com/135516
Reviewed-by: Ian Lance Taylor <iant@golang.org>
6 years agointernal/poll: fall back on unsupported splice from unix socket
Ben Burkert [Wed, 5 Sep 2018 17:20:26 +0000 (10:20 -0700)]
internal/poll: fall back on unsupported splice from unix socket

Gracefully fallback to a userspace copy when the kernel does not support
splice(2) on a unix domain socket. EINVAL is returned by the splice
syscall if it does not support unix domain sockets. Keeping the handled
return value as false when the first splice call fails with EINVAL will
cause the caller to fall back to a userspace copy.

Fixes #27513

Change-Id: I4b10c1900ba3c096cb32edb7c8a6044f468efb52
Reviewed-on: https://go-review.googlesource.com/133575
Run-TryBot: Tobias Klauser <tobias.klauser@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Tobias Klauser <tobias.klauser@gmail.com>
6 years agodoc/go1.11, cmd/go: elaborate on new GOFLAGS environment variable
Dmitri Shuralyov [Wed, 12 Sep 2018 17:58:18 +0000 (13:58 -0400)]
doc/go1.11, cmd/go: elaborate on new GOFLAGS environment variable

In Go 1.11, cmd/go gained support for the GOFLAGS environment variable.
It was added and described in detail in CL 126656.
Mention it in the Go 1.11 release notes, link to the cmd/go documentation,
and add more details there.

Fixes #27282.

Change-Id: Ifc35bfe3e0886a145478d36dde8e80aedd8ec68e
Reviewed-on: https://go-review.googlesource.com/135035
Reviewed-by: Bryan C. Mills <bcmills@google.com>
Reviewed-by: Rob Pike <r@golang.org>
6 years agotime: improve error message for LoadLocation
Agniva De Sarker [Sat, 30 Jun 2018 20:05:38 +0000 (01:35 +0530)]
time: improve error message for LoadLocation

Currently, when a tz file was being checked inside a zoneInfo dir,
a syscall.ENOENT error was being returned, which caused it to look
in the zoneinfo.zip file and return an error for that case.

We return a syscall.ENOENT error for the zip file case too, so that
it falls through to the end of the loop and returns an uniform error
for both cases.

Fixes #20969

Change-Id: If1de068022ac7693caabb5cffd1c929878460140
Reviewed-on: https://go-review.googlesource.com/121877
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
6 years agoruntime: use CleanCmdEnv in TestRuntimePanic
Ian Lance Taylor [Fri, 14 Sep 2018 17:04:12 +0000 (10:04 -0700)]
runtime: use CleanCmdEnv in TestRuntimePanic

This makes TestRuntimePanic keep most of the existing environment,
just as the other runtime tests do.

Change-Id: I7944abfeee292d41716dca14483134a50d75f081
Reviewed-on: https://go-review.googlesource.com/135376
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
6 years agocmd/compile: fold address calculations into CMPload[const] ops
Keith Randall [Fri, 14 Sep 2018 17:57:57 +0000 (10:57 -0700)]
cmd/compile: fold address calculations into CMPload[const] ops

Makes go binary smaller by 0.2%.

I noticed this in autogenerated equal methods, and there are
probably a lot of those.

Change-Id: I4e04eb3653fbceb9dd6a4eee97ceab1fa4d10b72
Reviewed-on: https://go-review.googlesource.com/135379
Reviewed-by: Ilya Tocar <ilya.tocar@intel.com>
6 years agocmd/go: correct gccgo buildid file on ARM
Ian Lance Taylor [Thu, 13 Sep 2018 22:02:27 +0000 (15:02 -0700)]
cmd/go: correct gccgo buildid file on ARM

The GNU assembler for ARM treats @ as a comment character, so section
types must be written using % instead.

Fixes https://gcc.gnu.org/PR87260.

Change-Id: I5461e4bf5b20793db321f540c7f25a9e6e12b6f4
Reviewed-on: https://go-review.googlesource.com/135297
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Bryan C. Mills <bcmills@google.com>
6 years agocmd/compile: skip float32 constant folding test on 387 builder
Michael Munday [Thu, 13 Sep 2018 12:16:46 +0000 (13:16 +0100)]
cmd/compile: skip float32 constant folding test on 387 builder

The 387 unit always quietens float64 and float32 signaling NaNs,
even when just loading and storing them. This makes it difficult
to propagate such values in the compiler. This is a hard problem
to fix and it is also very obscure.

Updates #27516.

Change-Id: I03d88e31f14c86fa682fcea4b6d1fba18968aee8
Reviewed-on: https://go-review.googlesource.com/135195
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
6 years agopath/filepath: correct symlink eval for symlink at root
Ian Lance Taylor [Thu, 13 Sep 2018 19:18:09 +0000 (12:18 -0700)]
path/filepath: correct symlink eval for symlink at root

For a relative symlink in the root directory, such as /tmp ->
private/tmp, we were dropping the leading slash.

No test because we can't create a symlink in the root directory.
The test TestGZIPFilesHaveZeroMTimes was failing on the Darwin builders.

Updates #19922
Updates #20506

Change-Id: Ic83cb6d97ad0cb628fc551ac772a44fb3e20f038
Reviewed-on: https://go-review.googlesource.com/135295
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Robert Griesemer <gri@golang.org>
6 years agoruntime: don't build semasleep_test on Plan 9
David du Colombier [Thu, 13 Sep 2018 19:15:34 +0000 (21:15 +0200)]
runtime: don't build semasleep_test on Plan 9

CL 135015 added TestSpuriousWakeupsNeverHangSemasleep.
However, this test is failing on Plan 9 because
syscall.SIGIO is not defined.

This change excludes semasleep_test.go on Plan 9

Fixes #27662.

Change-Id: I52f9f0fe9ec3c70da5d2f586a95debbc1fe568a1
Reviewed-on: https://go-review.googlesource.com/135315
Run-TryBot: David du Colombier <0intro@gmail.com>
Reviewed-by: Emmanuel Odeke <emm.odeke@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

6 years agocmd/go/internal/modfetch: stop cutting the last character of versions
Roberto Selbach [Mon, 27 Aug 2018 15:57:32 +0000 (15:57 +0000)]
cmd/go/internal/modfetch: stop cutting the last character of versions

When a zip archive for a module contains an unexpected file, the error
message removes the last character in the version number, e.g. an invalid
archive for "somemod@v1.2.3" would generate the following error:
"zip for somemod@1.2. has unexpected file..."

Change-Id: I366622df16a71fa7467a4bc62cb696e3e83a2942
GitHub-Last-Rev: f172283bcdffd45b485b1e8fd99795eb93fef726
GitHub-Pull-Request: golang/go#27279
Reviewed-on: https://go-review.googlesource.com/131635
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/compile: improve rules for PPC64.rules
Lynn Boger [Wed, 15 Aug 2018 21:34:06 +0000 (17:34 -0400)]
cmd/compile: improve rules for PPC64.rules

This adds some improvements to the rules for PPC64 to eliminate
unnecessary zero or sign extends, and fix some rule for truncates
which were not always using the correct sign instruction.

This reduces of size of many functions by 1 or 2 instructions and
can improve performance in cases where the execution time depends
on small loops where at least 1 instruction was removed and where that
loop contributes a significant amount of the total execution time.

Included is a testcase for codegen to verify the sign/zero extend
instructions are omitted.

An example of the improvement (strings):
IndexAnyASCII/256:1-16     392ns ± 0%   369ns ± 0%  -5.79%  (p=0.000 n=1+10)
IndexAnyASCII/256:2-16     397ns ± 0%   376ns ± 0%  -5.23%  (p=0.000 n=1+9)
IndexAnyASCII/256:4-16     405ns ± 0%   384ns ± 0%  -5.19%  (p=1.714 n=1+6)
IndexAnyASCII/256:8-16     427ns ± 0%   403ns ± 0%  -5.57%  (p=0.000 n=1+10)
IndexAnyASCII/256:16-16    441ns ± 0%   418ns ± 1%  -5.33%  (p=0.000 n=1+10)
IndexAnyASCII/4096:1-16   5.62µs ± 0%  5.27µs ± 1%  -6.31%  (p=0.000 n=1+10)
IndexAnyASCII/4096:2-16   5.67µs ± 0%  5.29µs ± 0%  -6.67%  (p=0.222 n=1+8)
IndexAnyASCII/4096:4-16   5.66µs ± 0%  5.28µs ± 1%  -6.66%  (p=0.000 n=1+10)
IndexAnyASCII/4096:8-16   5.66µs ± 0%  5.31µs ± 1%  -6.10%  (p=0.000 n=1+10)
IndexAnyASCII/4096:16-16  5.70µs ± 0%  5.33µs ± 1%  -6.43%  (p=0.182 n=1+10)

Change-Id: I739a6132b505936d39001aada5a978ff2a5f0500
Reviewed-on: https://go-review.googlesource.com/129875
Reviewed-by: David Chase <drchase@google.com>
6 years agogo/types: be more precise in API comment
Robert Griesemer [Thu, 13 Sep 2018 00:11:33 +0000 (17:11 -0700)]
go/types: be more precise in API comment

Change-Id: I24c4b08091bf3b8734f5dcdb9eac1a3582a4daa8
Reviewed-on: https://go-review.googlesource.com/135116
Reviewed-by: Alan Donovan <adonovan@google.com>
6 years agogo/types: fix scope printing (debugging support)
Robert Griesemer [Wed, 12 Sep 2018 23:45:10 +0000 (16:45 -0700)]
go/types: fix scope printing (debugging support)

Printing of scopes was horribly wrong: If a scope contained
no object declarations, it would abort printing even if the
scope had children scopes. Also, the line breaks were not
inserted consistently. The actual test case (ExampleScope)
was incorrect as well.

Fixed and simplified printing, and adjusted example which
tests the printing output.

Change-Id: If21c1d4ad71b15a517d4a54da16de5e6228eb4b5
Reviewed-on: https://go-review.googlesource.com/135115
Run-TryBot: Robert Griesemer <gri@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Alan Donovan <adonovan@google.com>
6 years agoruntime: skip TestGcSys on arm64
erifan01 [Wed, 12 Sep 2018 10:31:04 +0000 (10:31 +0000)]
runtime: skip TestGcSys on arm64

This failure occurs randomly on arm64.

13:10:32 --- FAIL: TestGcSys (0.06s)
13:10:32 gc_test.go:30: expected "OK\n", but got "using too much memory: 71401472 bytes\n"
13:10:32 FAIL

Updates #27636

Change-Id: Ifd4cfce167d8054dc6f037bd34368d63c7f68ed4
Reviewed-on: https://go-review.googlesource.com/135155
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
6 years agopath/filepath: rewrite walkSymlinks
Ian Lance Taylor [Fri, 29 Jun 2018 17:39:44 +0000 (10:39 -0700)]
path/filepath: rewrite walkSymlinks

Rather than try to work around Clean and Join on intermediate steps,
which can remove ".." components unexpectedly, just do everything in
walkSymlinks. Use a single loop over path components.

Fixes #23444

Change-Id: I4f15e50d0df32349cc4fd55e3d224ec9ab064379
Reviewed-on: https://go-review.googlesource.com/121676
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Alex Brainman <alex.brainman@gmail.com>
6 years agoregexp: fix incorrect name in Match doc comment
Alan Donovan [Thu, 13 Sep 2018 15:15:30 +0000 (11:15 -0400)]
regexp: fix incorrect name in Match doc comment

Change-Id: I628aad9a3abe9cc0c3233f476960e53bd291eca9
Reviewed-on: https://go-review.googlesource.com/135235
Reviewed-by: Ralph Corderoy <ralph@inputplus.co.uk>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
6 years agostrings, bytes: optimize function Index
erifan01 [Thu, 19 Jul 2018 09:55:17 +0000 (09:55 +0000)]
strings, bytes: optimize function Index

This change compares the first two characters instead of the first one,
and if they match, the entire string is compared. Comparing the first two
characters helps to filter out the case where the first character matches
but the subsequent characters do not match, thereby improving the substring
search speed in this case.

Benchmarks with no effect or minimal impact (less than 5%) is not listed,
the following are improved benchmarks:
On arm64:
strings:
IndexPeriodic/IndexPeriodic16-8   172890.00ns +- 2%   124156.20ns +- 0%  -28.19%  (p=0.008 n=5+5)
IndexPeriodic/IndexPeriodic32-8    78092.80ns +- 0%    65138.60ns +- 0%  -16.59%  (p=0.008 n=5+5)
IndexPeriodic/IndexPeriodic64-8    42322.20ns +- 0%    34661.60ns +- 0%  -18.10%  (p=0.008 n=5+5)
bytes:
IndexPeriodic/IndexPeriodic16-8     183468.20ns +- 6%     123759.00ns +- 0%  -32.54%  (p=0.008 n=5+5)
IndexPeriodic/IndexPeriodic32-8      84776.40ns +- 0%      63907.80ns +- 0%  -24.62%  (p=0.008 n=5+5)
IndexPeriodic/IndexPeriodic64-8      45835.60ns +- 0%      34194.20ns +- 0%  -25.40%  (p=0.008 n=5+5)

On amd64:
strings:
IndexPeriodic/IndexPeriodic8-16    219499.00ns +- 0%   178123.40ns +- 0%  -18.85%  (p=0.008 n=5+5)
IndexPeriodic/IndexPeriodic16-16   109760.20ns +- 0%    88957.80ns +- 0%  -18.95%  (p=0.008 n=5+5)
IndexPeriodic/IndexPeriodic32-16    54943.00ns +- 0%    44573.80ns +- 0%  -18.87%  (p=0.008 n=5+5)
IndexPeriodic/IndexPeriodic64-16    29804.80ns +- 0%    24417.80ns +- 0%  -18.07%  (p=0.008 n=5+5)
bytes:
IndexPeriodic/IndexPeriodic8-16     226592.60ns +- 0%    181183.20ns +- 0%  -20.04%  (p=0.008 n=5+5)
IndexPeriodic/IndexPeriodic16-16    111432.60ns +- 0%     90634.60ns +- 0%  -18.66%  (p=0.008 n=5+5)
IndexPeriodic/IndexPeriodic32-16     55640.60ns +- 0%     45433.00ns +- 0%  -18.35%  (p=0.008 n=5+5)
IndexPeriodic/IndexPeriodic64-16     30833.00ns +- 0%     24784.20ns +- 0%  -19.62%  (p=0.008 n=5+5)

Change-Id: I2d9e7e138d29e960d20a203eb74dc2ec976a9d71
Reviewed-on: https://go-review.googlesource.com/131177
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
6 years agocmd/vet: avoid internal error for implicitly declared type switch vars
Robert Griesemer [Thu, 13 Sep 2018 01:07:51 +0000 (18:07 -0700)]
cmd/vet: avoid internal error for implicitly declared type switch vars

For type switches using a short variable declaration of the form

   switch t := x.(type) {
   case T1:
   ...

go/types doesn't declare the symbolic variable (t in this example)
with the switch; thus such variables are not found in types.Info.Defs.

Instead they are implicitly declared with each type switch case,
and can be found in types.Info.Implicits.

Adjust the shadowing code accordingly.

Added a test case to verify that the issue is fixed, and a test
case verifying that the shadowing code now considers implicitly
declared variables introduces in type switch cases.

While at it, also fixed the (internal) error reporting to provide
more accurate information.

Fixe #26725.

Change-Id: If408ed9e692bf47c640f81de8f46bf5eb43415b0
Reviewed-on: https://go-review.googlesource.com/135117
Run-TryBot: Robert Griesemer <gri@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Daniel Martí <mvdan@mvdan.cc>
Reviewed-by: Alan Donovan <adonovan@google.com>
6 years agocmd/compile: intrinsify math.RoundToEven and math.Abs on arm64
erifan01 [Tue, 22 May 2018 06:58:32 +0000 (06:58 +0000)]
cmd/compile: intrinsify math.RoundToEven and math.Abs on arm64

math.RoundToEven can be done by one arm64 instruction FRINTND, intrinsify it to improve performance.
The current pure Go implementation of the function Abs is translated into five instructions on arm64:
str, ldr, and, str, ldr. The intrinsic implementation requires only one instruction, so in terms of
performance, intrinsify it is worthwhile.

Benchmarks:
name           old time/op  new time/op  delta
Abs-8          3.50ns ± 0%  1.50ns ± 0%  -57.14%  (p=0.000 n=10+10)
RoundToEven-8  9.26ns ± 0%  1.50ns ± 0%  -83.80%  (p=0.000 n=10+10)

Change-Id: I9456b26ab282b544dfac0154fc86f17aed96ac3d
Reviewed-on: https://go-review.googlesource.com/116535
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

6 years agoos: add ModeCharDevice to ModeType
Tobias Klauser [Wed, 12 Sep 2018 19:30:09 +0000 (21:30 +0200)]
os: add ModeCharDevice to ModeType

When masking FileInfo.Mode() from a character device with the ModeType
mask, ModeCharDevice cannot be recovered.

ModeCharDevice was added https://golang.org/cl/5531052, but nothing
indicates why it was omitted from ModeType. Add it now.

Fixes #27640

Change-Id: I52f56108b88b1b0a5bc6085c66c3c67e10600619
Reviewed-on: https://go-review.googlesource.com/135075
Run-TryBot: Tobias Klauser <tobias.klauser@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
6 years agoruntime: regression test for semasleep indefinite hang
Emmanuel T Odeke [Wed, 12 Sep 2018 18:22:42 +0000 (12:22 -0600)]
runtime: regression test for semasleep indefinite hang

A regression test in which: for a program that invokes semasleep,
we send non-terminal signals such as SIGIO.
Since the signal wakes up pthread_cond_timedwait_relative_np,
after CL 133655, we should only re-spin for the amount of
time left, instead of re-spinning with the original duration
which would cause an indefinite spin.

Updates #27520

Change-Id: I744a6d04cf8923bc4e13649446aff5e42b7de5d8
Reviewed-on: https://go-review.googlesource.com/135015
Run-TryBot: Emmanuel Odeke <emm.odeke@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
6 years agointernal/bytealg: optimize Equal on arm64
erifan01 [Mon, 7 May 2018 08:08:30 +0000 (08:08 +0000)]
internal/bytealg: optimize Equal on arm64

Currently the 16-byte loop chunk16_loop is implemented with NEON instructions LD1, VMOV and VCMEQ.
Using scalar instructions LDP and CMP to achieve this loop can reduce the number of clock cycles.
For cases where the length of strings are between 4 to 15 bytes, loading the last 8 or 4 bytes at
a time to reduce the number of comparisons.

Benchmarks:
name                 old time/op    new time/op    delta
Equal/0-8              5.51ns ± 0%    5.84ns ±14%     ~     (p=0.246 n=7+8)
Equal/1-8              10.5ns ± 0%    10.5ns ± 0%     ~     (all equal)
Equal/6-8              14.0ns ± 0%    12.5ns ± 0%  -10.71%  (p=0.000 n=8+8)
Equal/9-8              13.5ns ± 0%    12.5ns ± 0%   -7.41%  (p=0.000 n=8+8)
Equal/15-8             15.5ns ± 0%    12.5ns ± 0%  -19.35%  (p=0.000 n=8+8)
Equal/16-8             14.0ns ± 0%    13.0ns ± 0%   -7.14%  (p=0.000 n=8+8)
Equal/20-8             16.5ns ± 0%    16.0ns ± 0%   -3.03%  (p=0.000 n=8+8)
Equal/32-8             16.5ns ± 0%    15.3ns ± 0%   -7.27%  (p=0.000 n=8+8)
Equal/4K-8              552ns ± 0%     553ns ± 0%     ~     (p=0.315 n=8+8)
Equal/4M-8             1.13ms ±23%    1.20ms ±27%     ~     (p=0.442 n=8+8)
Equal/64M-8            32.9ms ± 0%    32.6ms ± 0%   -1.15%  (p=0.000 n=8+8)
CompareBytesEqual-8    12.0ns ± 0%    12.0ns ± 0%     ~     (all equal)

Change-Id: If317ecdcc98e31883d37fd7d42b113b548c5bd2a
Reviewed-on: https://go-review.googlesource.com/112496
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Run-TryBot: Cherry Zhang <cherryyz@google.com>

6 years agocmd/go: avoid type names in __debug__modinfo__ variable injected in package main
Bryan C. Mills [Wed, 12 Sep 2018 14:41:29 +0000 (10:41 -0400)]
cmd/go: avoid type names in __debug__modinfo__ variable injected in package main

If we use the name 'string' to refer to the built-in type, that name
can be shadowed by a local declaration. Use a string constant instead,
but keep the init function to populate it so that //go:linkname will
still work properly.

Fixes #27584.

Change-Id: I850cad6663e566f70fd123107d2e4e742c93b450
Reviewed-on: https://go-review.googlesource.com/134915
Reviewed-by: Ian Lance Taylor <iant@golang.org>
6 years agoruntime: use functions from mem_bsd.go on Darwin
Tobias Klauser [Wed, 12 Sep 2018 12:18:45 +0000 (14:18 +0200)]
runtime: use functions from mem_bsd.go on Darwin

The implementations of the functions in mem_darwin.go is identical to
the ones defined in mem_bsd.go for all other BSD-like GOOSes. Also use
them on Darwin.

Change-Id: Ie7c170c1a50666475e79599471081cd85f0837ad
Reviewed-on: https://go-review.googlesource.com/134875
Run-TryBot: Tobias Klauser <tobias.klauser@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
6 years agocmd/dist: make raceDetectorSupported an exact copy of cmd/internal/sys.RaceDetectorSu...
Tobias Klauser [Tue, 11 Sep 2018 12:27:41 +0000 (14:27 +0200)]
cmd/dist: make raceDetectorSupported an exact copy of cmd/internal/sys.RaceDetectorSupported

The comment states that cmd/internal/sys.RaceDetectorSupported is a copy,
so make the two identical. No functional difference, since ppce64le is
only supported on linux anyway.

Change-Id: Id3e4d445fb700b9b3bb53bf15ea05b8911b4f95e
Reviewed-on: https://go-review.googlesource.com/134595
Run-TryBot: Tobias Klauser <tobias.klauser@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
6 years agoruntime: convert initial timediv quotient increments to bitsets
Emmanuel T Odeke [Mon, 10 Sep 2018 07:05:09 +0000 (01:05 -0600)]
runtime: convert initial timediv quotient increments to bitsets

At the very beginning of timediv, inside a for loop,
we reduce the base value by at most (1<<31)-1, while
incrementing the quotient result by 1<<uint(bit).
However, since the quotient value was 0 to begin with,
we are essentially just doing bitsets.

This change is in the hot path of various concurrency and
scheduling operations that require sleeping, waiting
on mutexes and futexes etc. On the following OSes:
* Dragonfly
* FreeBSD
* Linux
* NetBSD
* OpenBSD
* Plan9
* Windows

and paired with architectures that provide the BTS instruction, this
change shaves off a couple of nanoseconds per invocation of timediv.

Fixes #27529

Change-Id: Ia2fea5022c1109e02d86d1f962a3b0bd70967aa6
Reviewed-on: https://go-review.googlesource.com/134231
Run-TryBot: Emmanuel Odeke <emm.odeke@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
6 years agointernal/bytealg: implement bytes.Count in asm for ppc64x
Lynn Boger [Mon, 27 Aug 2018 20:34:53 +0000 (16:34 -0400)]
internal/bytealg: implement bytes.Count in asm for ppc64x

This adds an asm implementation for the Count function in ppc64x.
The Go code that manipulates a byte at a time is especially
inefficient on ppc64x, so an asm implementation is a significant
improvement.

bytes:
name               old time/op   new time/op    delta
CountSingle/10-8    23.1ns ± 0%    18.6ns ± 0%    -19.48%  (p=1.000 n=1+1)
CountSingle/32-8    60.4ns ± 0%    19.0ns ± 0%    -68.54%  (p=1.000 n=1+1)
CountSingle/4K-8    7.29µs ± 0%    0.45µs ± 0%    -93.80%  (p=1.000 n=1+1)
CountSingle/4M-8    7.49ms ± 0%    0.45ms ± 0%    -93.97%  (p=1.000 n=1+1)
CountSingle/64M-8    127ms ± 0%       9ms ± 0%    -92.53%  (p=1.000 n=1+1)

html:
name              old time/op  new time/op  delta
Escape-8          57.5µs ± 0%  36.1µs ± 0%  -37.13%  (p=1.000 n=1+1)
EscapeNone-8      20.0µs ± 0%   2.0µs ± 0%  -90.14%  (p=1.000 n=1+1)

Change-Id: Iadbf422c0e9a37b47d2d95fb8c778420f3aabb58
Reviewed-on: https://go-review.googlesource.com/131695
Run-TryBot: Lynn Boger <laboger@linux.vnet.ibm.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Michael Munday <mike.munday@ibm.com>
6 years agocmd/internal/obj/arm64: add error report for invalid base register
fanzha02 [Mon, 10 Sep 2018 02:22:48 +0000 (02:22 +0000)]
cmd/internal/obj/arm64: add error report for invalid base register

The current assembler accepts the non-integer register as the base register,
which should be an illegal combination.

Add the test cases.

Change-Id: Ia21596bbb5b1e212e34bd3a170748ae788860422
Reviewed-on: https://go-review.googlesource.com/134575
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

6 years agotest: fix the wrong test of math.Copysign(c, -1) for arm64
fanzha02 [Wed, 12 Sep 2018 01:43:09 +0000 (01:43 +0000)]
test: fix the wrong test of math.Copysign(c, -1) for arm64

The CL 132915 added the wrong codegen test for math.Copysign(c, -1),
it should test that AND is not emitted. This CL fixes this error.

Change-Id: Ida1d3d54ebfc7f238abccbc1f70f914e1b5bfd91
Reviewed-on: https://go-review.googlesource.com/134815
Reviewed-by: Giovanni Bajo <rasky@develer.com>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Run-TryBot: Giovanni Bajo <rasky@develer.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

6 years agoencoding/json: use panics for phase errors
Daniel Martí [Wed, 12 Sep 2018 07:26:31 +0000 (09:26 +0200)]
encoding/json: use panics for phase errors

Having these panic-like errors used to be ok, since they were used in
the internal decoder state instead of passed around via return
parameters.

Recently, the decoder was rewritten to use explicit error returns
instead. This error is a terrible fit for error returns; a handful of
functions must return an error because of it, and their callers must
check for an error that should never happen.

This is precisely what panics are for, so use them. The test coverage of
the package goes up from 91.3% to 91.6%, and performance is unaffected.
We can also get rid of some unnecessary verbosity in the code.

name           old time/op    new time/op    delta
CodeDecoder-4    27.5ms ± 1%    27.5ms ± 1%   ~     (p=0.937 n=6+6)

Change-Id: I01033b3f5b7c0cf0985082fa272754f96bf6353c
Reviewed-on: https://go-review.googlesource.com/134835
Run-TryBot: Daniel Martí <mvdan@mvdan.cc>
Run-TryBot: Joe Tsai <thebrokentoaster@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Joe Tsai <thebrokentoaster@gmail.com>
6 years agoencoding/json: more tests to cover decoding edge cases
Daniel Martí [Tue, 11 Sep 2018 20:09:00 +0000 (22:09 +0200)]
encoding/json: more tests to cover decoding edge cases

The overall coverage of the json package goes up from 90.8% to 91.3%.

While at it, apply two minor code simplifications found while inspecting
the HTML coverage report.

Change-Id: I0fba968afeedc813b1385e4bde72d93b878854d7
Reviewed-on: https://go-review.googlesource.com/134735
Run-TryBot: Daniel Martí <mvdan@mvdan.cc>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
6 years agocmd/link: don't pass all linker args when testing flag
Ian Lance Taylor [Fri, 7 Sep 2018 19:53:52 +0000 (12:53 -0700)]
cmd/link: don't pass all linker args when testing flag

Some linker flags can actually be input files, which can cause
misleading errors when doing the trial link, which can cause the
linker to incorrectly decide that the flag is not supported, which can
cause the link to fail.

Fixes #27510
Updates #27110
Updates #27293

Change-Id: I70c1e913cee3c813e7b267bf779bcff26d4d194a
Reviewed-on: https://go-review.googlesource.com/134057
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Lynn Boger <laboger@linux.vnet.ibm.com>
Reviewed-by: Damien Neil <dneil@google.com>
6 years agocmd/go/testdata/script: fix typos in test comments
Bryan C. Mills [Thu, 9 Aug 2018 21:54:43 +0000 (17:54 -0400)]
cmd/go/testdata/script: fix typos in test comments

Change-Id: I2a55cda76f1074b997349dfd6e001dc7277faade
Reviewed-on: https://go-review.googlesource.com/134655
Reviewed-by: Daniel Martí <mvdan@mvdan.cc>
6 years agomath/bits: add extended precision Add, Sub, Mul, Div
Brian Kessler [Tue, 14 Aug 2018 22:39:13 +0000 (16:39 -0600)]
math/bits: add extended precision Add, Sub, Mul, Div

Port math/big pure go versions of add-with-carry, subtract-with-borrow,
full-width multiply, and full-width divide.

Updates #24813

Change-Id: Ifae5d2f6ee4237137c9dcba931f69c91b80a4b1c
Reviewed-on: https://go-review.googlesource.com/123157
Reviewed-by: Robert Griesemer <gri@golang.org>
Run-TryBot: Robert Griesemer <gri@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

6 years agocmd/compile: use yyerrorl(n.Pos, ...) in typecheckdef
Yury Smolsky [Mon, 28 May 2018 14:15:11 +0000 (17:15 +0300)]
cmd/compile: use  yyerrorl(n.Pos, ...) in typecheckdef

n.Pos.IsKnown() is not needed because it is performed in setlineno.

toolstash-check passed.

Updates #19683.

Change-Id: I34d6a0e6dc9970679d99e8f3424f289ebf1e86ba
Reviewed-on: https://go-review.googlesource.com/114915
Run-TryBot: Yury Smolsky <yury@smolsky.by>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Robert Griesemer <gri@golang.org>
6 years agocmd/link,compress/zip,image/png: use binary.{Big,Little}Endian methods
Lynn Boger [Mon, 10 Sep 2018 19:07:09 +0000 (15:07 -0400)]
cmd/link,compress/zip,image/png: use binary.{Big,Little}Endian methods

Use the binary.{Big,Little}Endian integer encoding methods rather than
variations found in local implementations. The functions in
the binary package have been tested to ensure they inline correctly and
don't add unnecessary bounds checking.

Change-Id: Ie10111ca6edb7c11e8e5e21c58a5748ae99b7f87
Reviewed-on: https://go-review.googlesource.com/134375
Run-TryBot: Lynn Boger <laboger@linux.vnet.ibm.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Michael Munday <mike.munday@ibm.com>
6 years agocmd/compile: optimize arm's bit operation
Ben Shi [Mon, 10 Sep 2018 08:29:52 +0000 (08:29 +0000)]
cmd/compile: optimize arm's bit operation

BFC (Bit Field Clear) was introduced in ARMv7, which can simplify
ANDconst and BICconst. And this CL implements that optimization.

1. The total size of pkg/android_arm decreases about 3KB, excluding
cmd/compile/.

2. There is no regression in the go1 benchmark result, and some
cases (FmtFprintfEmpty-4 and RegexpMatchMedium_32-4) even get
slight improvement.

name                     old time/op    new time/op    delta
BinaryTree17-4              25.3s ± 1%     25.2s ± 1%    ~     (p=0.072 n=30+29)
Fannkuch11-4                13.3s ± 0%     13.3s ± 0%  +0.13%  (p=0.000 n=30+26)
FmtFprintfEmpty-4           407ns ± 0%     394ns ± 0%  -3.19%  (p=0.000 n=26+28)
FmtFprintfString-4          664ns ± 0%     662ns ± 0%  -0.22%  (p=0.000 n=30+30)
FmtFprintfInt-4             712ns ± 0%     706ns ± 0%  -0.79%  (p=0.000 n=30+30)
FmtFprintfIntInt-4         1.06µs ± 0%    1.05µs ± 0%  -0.38%  (p=0.000 n=30+30)
FmtFprintfPrefixedInt-4    1.16µs ± 0%    1.16µs ± 0%  -0.13%  (p=0.000 n=30+29)
FmtFprintfFloat-4          2.24µs ± 0%    2.23µs ± 0%  -0.51%  (p=0.000 n=29+21)
FmtManyArgs-4              4.09µs ± 0%    4.06µs ± 0%  -0.83%  (p=0.000 n=28+30)
GobDecode-4                55.0ms ± 5%    55.4ms ± 5%    ~     (p=0.307 n=30+30)
GobEncode-4                51.2ms ± 1%    51.9ms ± 1%  +1.23%  (p=0.000 n=29+30)
Gzip-4                      2.64s ± 0%     2.60s ± 0%  -1.35%  (p=0.000 n=30+29)
Gunzip-4                    309ms ± 0%     308ms ± 0%  -0.27%  (p=0.000 n=30+30)
HTTPClientServer-4         1.03ms ± 5%    1.02ms ± 4%    ~     (p=0.117 n=30+29)
JSONEncode-4                101ms ± 2%     101ms ± 2%    ~     (p=0.338 n=29+29)
JSONDecode-4                383ms ± 2%     382ms ± 2%    ~     (p=0.751 n=26+30)
Mandelbrot200-4            18.4ms ± 0%    18.4ms ± 0%  -0.10%  (p=0.000 n=29+29)
GoParse-4                  22.6ms ± 0%    22.5ms ± 0%  -0.39%  (p=0.000 n=30+30)
RegexpMatchEasy0_32-4       761ns ± 0%     750ns ± 0%  -1.47%  (p=0.000 n=26+29)
RegexpMatchEasy0_1K-4      4.33µs ± 0%    4.34µs ± 0%  +0.27%  (p=0.000 n=25+28)
RegexpMatchEasy1_32-4       809ns ± 0%     795ns ± 0%  -1.74%  (p=0.000 n=27+25)
RegexpMatchEasy1_1K-4      5.54µs ± 0%    5.53µs ± 0%  -0.18%  (p=0.000 n=29+29)
RegexpMatchMedium_32-4     1.11µs ± 0%    1.08µs ± 0%  -2.78%  (p=0.000 n=27+29)
RegexpMatchMedium_1K-4      255µs ± 0%     255µs ± 0%  -0.02%  (p=0.029 n=30+30)
RegexpMatchHard_32-4       14.7µs ± 0%    14.7µs ± 0%  -0.28%  (p=0.000 n=30+29)
RegexpMatchHard_1K-4        439µs ± 0%     439µs ± 0%    ~     (p=0.907 n=23+27)
Revcomp-4                  41.9ms ± 1%    41.9ms ± 1%    ~     (p=0.230 n=28+30)
Template-4                  522ms ± 1%     528ms ± 1%  +1.25%  (p=0.000 n=30+30)
TimeParse-4                3.34µs ± 0%    3.35µs ± 0%  +0.23%  (p=0.000 n=30+27)
TimeFormat-4               6.06µs ± 0%    6.13µs ± 0%  +1.08%  (p=0.000 n=29+29)
[Geo mean]                  384µs          382µs       -0.37%

name                     old speed      new speed      delta
GobDecode-4              14.0MB/s ± 5%  13.9MB/s ± 5%    ~     (p=0.308 n=30+30)
GobEncode-4              15.0MB/s ± 1%  14.8MB/s ± 1%  -1.22%  (p=0.000 n=29+30)
Gzip-4                   7.36MB/s ± 0%  7.46MB/s ± 0%  +1.35%  (p=0.000 n=30+30)
Gunzip-4                 62.8MB/s ± 0%  63.0MB/s ± 0%  +0.27%  (p=0.000 n=30+30)
JSONEncode-4             19.2MB/s ± 2%  19.2MB/s ± 2%    ~     (p=0.312 n=29+29)
JSONDecode-4             5.05MB/s ± 3%  5.08MB/s ± 2%    ~     (p=0.356 n=29+30)
GoParse-4                2.56MB/s ± 0%  2.57MB/s ± 0%  +0.39%  (p=0.000 n=23+27)
RegexpMatchEasy0_32-4    42.0MB/s ± 0%  42.6MB/s ± 0%  +1.50%  (p=0.000 n=26+28)
RegexpMatchEasy0_1K-4     236MB/s ± 0%   236MB/s ± 0%  -0.27%  (p=0.000 n=25+28)
RegexpMatchEasy1_32-4    39.6MB/s ± 0%  40.2MB/s ± 0%  +1.73%  (p=0.000 n=27+27)
RegexpMatchEasy1_1K-4     185MB/s ± 0%   185MB/s ± 0%  +0.18%  (p=0.000 n=29+29)
RegexpMatchMedium_32-4    900kB/s ± 0%   920kB/s ± 0%  +2.22%  (p=0.000 n=29+29)
RegexpMatchMedium_1K-4   4.02MB/s ± 0%  4.02MB/s ± 0%  +0.07%  (p=0.004 n=30+27)
RegexpMatchHard_32-4     2.17MB/s ± 0%  2.18MB/s ± 0%  +0.46%  (p=0.000 n=30+26)
RegexpMatchHard_1K-4     2.33MB/s ± 0%  2.33MB/s ± 0%    ~     (all equal)
Revcomp-4                60.6MB/s ± 1%  60.7MB/s ± 1%    ~     (p=0.207 n=28+30)
Template-4               3.72MB/s ± 1%  3.67MB/s ± 1%  -1.23%  (p=0.000 n=30+30)
[Geo mean]               12.9MB/s       12.9MB/s       +0.29%

Change-Id: I07f497f8bb476c950dc555491d00c9066fb64a4e
Reviewed-on: https://go-review.googlesource.com/134232
Run-TryBot: Ben Shi <powerman1st@163.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
6 years agonet/url: remove an allocation for short strings in escape
Ian Davis [Mon, 10 Sep 2018 11:13:09 +0000 (12:13 +0100)]
net/url: remove an allocation for short strings in escape

Use a 64 byte array to avoid an allocation on the assumption that
most url escaping is performed on short strings. Also adds a fast
path for escaping strings whose only replacements are spaces which
is common in query components.

Adds benchmarks for QueryEscape, PathEscape, QueryUnescape and
PathUnescape but no optimizations are include for the unescape functions
so I don't include those benchmark results here.

Reduces allocations by 10% in the existing String benchmark with a
modest performance increase.

name               old time/op    new time/op    delta
QueryEscape/#00-8    64.6ns ± 1%    43.8ns ± 0%  -32.14%  (p=0.000 n=9+9)
QueryEscape/#01-8     276ns ± 3%     249ns ± 0%   -9.62%  (p=0.000 n=10+7)
QueryEscape/#02-8     176ns ± 2%     155ns ± 3%  -12.21%  (p=0.000 n=10+10)
QueryEscape/#03-8     388ns ± 1%     362ns ± 0%   -6.55%  (p=0.000 n=10+8)
QueryEscape/#04-8    2.32µs ± 2%    2.27µs ± 2%   -2.26%  (p=0.001 n=10+10)
PathEscape/#00-8     78.0ns ± 3%    63.4ns ± 1%  -18.69%  (p=0.000 n=10+10)
PathEscape/#01-8      276ns ± 2%     260ns ± 0%   -6.01%  (p=0.000 n=10+10)
PathEscape/#02-8      175ns ± 0%     153ns ± 0%  -12.53%  (p=0.000 n=8+10)
PathEscape/#03-8      389ns ± 2%     361ns ± 0%   -7.21%  (p=0.000 n=10+9)
PathEscape/#04-8     2.30µs ± 2%    2.27µs ± 1%   -1.33%  (p=0.001 n=9+10)
String-8             3.56µs ± 4%    3.42µs ± 7%   -4.00%  (p=0.003 n=10+10)

name               old alloc/op   new alloc/op   delta
QueryEscape/#00-8     16.0B ± 0%      8.0B ± 0%  -50.00%  (p=0.000 n=10+10)
QueryEscape/#01-8      128B ± 0%       64B ± 0%  -50.00%  (p=0.000 n=10+10)
QueryEscape/#02-8     64.0B ± 0%     32.0B ± 0%  -50.00%  (p=0.000 n=10+10)
QueryEscape/#03-8      128B ± 0%       64B ± 0%  -50.00%  (p=0.000 n=10+10)
QueryEscape/#04-8      832B ± 0%      832B ± 0%     ~     (all equal)
PathEscape/#00-8      32.0B ± 0%     16.0B ± 0%  -50.00%  (p=0.000 n=10+10)
PathEscape/#01-8       128B ± 0%       64B ± 0%  -50.00%  (p=0.000 n=10+10)
PathEscape/#02-8      64.0B ± 0%     32.0B ± 0%  -50.00%  (p=0.000 n=10+10)
PathEscape/#03-8       128B ± 0%       64B ± 0%  -50.00%  (p=0.000 n=10+10)
PathEscape/#04-8       704B ± 0%      704B ± 0%     ~     (all equal)
String-8             1.84kB ± 0%    1.66kB ± 0%   -9.57%  (p=0.000 n=10+10)

name               old allocs/op  new allocs/op  delta
QueryEscape/#00-8      2.00 ± 0%      1.00 ± 0%  -50.00%  (p=0.000 n=10+10)
QueryEscape/#01-8      2.00 ± 0%      1.00 ± 0%  -50.00%  (p=0.000 n=10+10)
QueryEscape/#02-8      2.00 ± 0%      1.00 ± 0%  -50.00%  (p=0.000 n=10+10)
QueryEscape/#03-8      2.00 ± 0%      1.00 ± 0%  -50.00%  (p=0.000 n=10+10)
QueryEscape/#04-8      2.00 ± 0%      2.00 ± 0%     ~     (all equal)
PathEscape/#00-8       2.00 ± 0%      1.00 ± 0%  -50.00%  (p=0.000 n=10+10)
PathEscape/#01-8       2.00 ± 0%      1.00 ± 0%  -50.00%  (p=0.000 n=10+10)
PathEscape/#02-8       2.00 ± 0%      1.00 ± 0%  -50.00%  (p=0.000 n=10+10)
PathEscape/#03-8       2.00 ± 0%      1.00 ± 0%  -50.00%  (p=0.000 n=10+10)
PathEscape/#04-8       2.00 ± 0%      2.00 ± 0%     ~     (all equal)
String-8               69.0 ± 0%      61.0 ± 0%  -11.59%  (p=0.000 n=10+10)

Updates #17860

Change-Id: I45c5e9d40b242f874c61f6ccc73bf94c494bb868
Reviewed-on: https://go-review.googlesource.com/134296
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
6 years agogo/token: add (*File).LineStart, which returns Pos for a given line
Alan Donovan [Fri, 7 Sep 2018 13:54:38 +0000 (09:54 -0400)]
go/token: add (*File).LineStart, which returns Pos for a given line

LineStart returns the position of the start of a given line.
Like MergeLine, it panics if the 1-based line number is invalid.

This function is especially useful in programs that occasionally
handle non-Go files such as assembly but wish to use the token.Pos
mechanism to identify file positions.

Change-Id: I5f774c0690074059553cdb38c0f681f5aafc8da1
Reviewed-on: https://go-review.googlesource.com/134075
Reviewed-by: Robert Griesemer <gri@golang.org>
6 years agogo/types: fix swapped use of "uses" and "defines" in ObjectOf documentation
Dominik Honnef [Sun, 9 Sep 2018 09:50:56 +0000 (11:50 +0200)]
go/types: fix swapped use of "uses" and "defines" in ObjectOf documentation

Change-Id: I855a9c88c379978099ea53c7d28b87cefd7f5d73
Reviewed-on: https://go-review.googlesource.com/134295
Reviewed-by: Robert Griesemer <gri@golang.org>
6 years agoos/user: retrieve Current username from /etc/passwd, not $USER
Kevin Burke [Sat, 8 Sep 2018 19:35:03 +0000 (12:35 -0700)]
os/user: retrieve Current username from /etc/passwd, not $USER

Per golang/go#27524 there are situations where the username for the
uid does not match the value in the $USER environment variable and it
seems sensible to choose the value in /etc/passwd when they disagree.

This may make the Current() call slightly more expensive, since we
read /etc/passwd with cgo disabled instead of just checking the
environment. However, we cache the result of Current() calls, so we
only invoke this cost once in the lifetime of the process.

Fixes #14626.
Fixes #27524.
Updates #24884.

Change-Id: I0dcd224cf7f61dd5292f3fcc363aa2e9656a2cb1
Reviewed-on: https://go-review.googlesource.com/134218
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

6 years agointernal/poll: handle zero-byte write in FD.WriteTo
Alex Brainman [Sun, 9 Sep 2018 04:21:25 +0000 (14:21 +1000)]
internal/poll: handle zero-byte write in FD.WriteTo

Zero-byte write was fixed by CL 132781, that was submitted 3 days ago.
But I just submitted CL 129137, and the CL broken zero-byte write
functionality without me noticing. CL 129137 was based on old commit
(older than 3 days ago), and try-bots did not discover the breakage.

Fix zero-byte write again.

Fixes windows build.

Change-Id: Ib403b25fd25cb881963f25706eecca92b924aaa1
Reviewed-on: https://go-review.googlesource.com/134275
Run-TryBot: Alex Brainman <alex.brainman@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Alex Brainman <alex.brainman@gmail.com>
6 years agointernal/poll: cap reads and writes to 1GB on windows
Alex Brainman [Sun, 12 Aug 2018 01:16:26 +0000 (11:16 +1000)]
internal/poll: cap reads and writes to 1GB on windows

Fixes #26923

Change-Id: I62fec814220ccdf7acd8d79a133d1add3f24cf98
Reviewed-on: https://go-review.googlesource.com/129137
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

6 years agocmd/compile/internal/amd64: simplify assembly generator
Ben Shi [Sat, 8 Sep 2018 14:23:14 +0000 (14:23 +0000)]
cmd/compile/internal/amd64: simplify assembly generator

Merge two case-statements together, since they have similar logic.

1. That makes the assembly generator more clear.
2. The total size of cmd/compile decreases about 0.8KB.

Change-Id: I0144a07152202ee7b21e323bcd5dea80a351a6e3
Reviewed-on: https://go-review.googlesource.com/134215
Run-TryBot: Ben Shi <powerman1st@163.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
6 years agodoc: make golang-nuts discussion list more prominent
Ian Davis [Fri, 7 Sep 2018 23:57:31 +0000 (00:57 +0100)]
doc: make golang-nuts discussion list more prominent

The discussion list was buried beneath the developer mailing list.
This change puts the discussion list first and gives it a more
prominent heading.

Change-Id: I8dcb4af98e454ae3a0140f9758a5656909126983
Reviewed-on: https://go-review.googlesource.com/134136
Reviewed-by: Ian Lance Taylor <iant@golang.org>
6 years agocmd/compile: move v.Pos.line check to warnRule
Josh Bleecher Snyder [Mon, 28 May 2018 21:47:35 +0000 (14:47 -0700)]
cmd/compile: move v.Pos.line check to warnRule

This simplifies the rewrite rules.

Change-Id: Iff062297d42a23cb31ad55e8c733842ecbc07da2
Reviewed-on: https://go-review.googlesource.com/129377
Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
6 years agoruntime: use tgkill for raise
Michael Pratt [Fri, 7 Sep 2018 00:21:59 +0000 (17:21 -0700)]
runtime: use tgkill for raise

raise uses tkill to send a signal to the current thread. For this use,
tgkill is functionally equivalent to tkill expect that it also takes the
pid as the first argument.

Using tgkill makes it simpler to run a Go program in a strict sandbox.
With kill and tgkill, the sandbox policy (e.g., seccomp) can prevent the
program from sending signals to other processes by checking that the
first argument == getpid().

With tkill, the policy must whitelist all tids in the process, which is
effectively impossible given Go's dynamic thread creation.

Fixes #27548

Change-Id: I8ed282ef1f7215b02ef46de144493e36454029ea
Reviewed-on: https://go-review.googlesource.com/133975
Run-TryBot: Michael Pratt <mpratt@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
6 years agocmd/go: add -Wl,--export-dynamic to linker flag whitelist
Ian Lance Taylor [Fri, 7 Sep 2018 11:12:42 +0000 (04:12 -0700)]
cmd/go: add -Wl,--export-dynamic to linker flag whitelist

Fixes #27496

Change-Id: I53538c7697729294a9e50ace26a6a7183131e837
Reviewed-on: https://go-review.googlesource.com/134016
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Andrew Bonventre <andybons@golang.org>
6 years agotest: extend noescape bytes.Buffer test suite
Iskander Sharipov [Fri, 7 Sep 2018 17:36:56 +0000 (20:36 +0300)]
test: extend noescape bytes.Buffer test suite

Added some more cases that should be guarded against regression.

Change-Id: I9f1dda2fd0be9b6e167ef1cc018fc8cce55c066c
Reviewed-on: https://go-review.googlesource.com/134017
Run-TryBot: Iskander Sharipov <iskander.sharipov@intel.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
6 years agocrypto/x509: allow ":" in Common Name hostnames
Filippo Valsorda [Fri, 7 Sep 2018 16:58:14 +0000 (12:58 -0400)]
crypto/x509: allow ":" in Common Name hostnames

At least one popular service puts a hostname which contains a ":"
in the Common Name field. On the other hand, I don't know of any name
constrained certificates that only work if we ignore such CNs.

Updates #24151

Change-Id: I2d813e3e522ebd65ab5ea5cd83390467a869eea3
Reviewed-on: https://go-review.googlesource.com/134076
Run-TryBot: Filippo Valsorda <filippo@golang.org>
Reviewed-by: Adam Langley <agl@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

6 years agoruntime: in semasleep, subtract time spent so far from timeout
Keith Randall [Wed, 5 Sep 2018 21:36:20 +0000 (14:36 -0700)]
runtime: in semasleep, subtract time spent so far from timeout

When pthread_cond_timedwait_relative_np gets a spurious wakeup
(due to a signal, typically), we used to retry with the same
relative timeout. That's incorrect, we should lower the timeout
by the time we've spent in this function so far.

In the worst case, signals come in and cause spurious wakeups
faster than the timeout, causing semasleep to never time out.

Also fix nacl and netbsd while we're here. They have similar issues.

Fixes #27520

Change-Id: I6601e120e44a4b8ef436eef75a1e7c8cf1d39e39
Reviewed-on: https://go-review.googlesource.com/133655
Run-TryBot: Keith Randall <khr@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
6 years agocmd/compile: implement non-constant rotates using ROR on arm64
erifan01 [Sat, 30 Jun 2018 06:48:51 +0000 (06:48 +0000)]
cmd/compile: implement non-constant rotates using ROR on arm64

Add some rules to match the Go code like:
y &= 63
x << y | x >> (64-y)
or
y &= 63
x >> y | x << (64-y)
as a ROR instruction. Make math/bits.RotateLeft faster on arm64.

Extends CL 132435 to arm64.

Benchmarks of math/bits.RotateLeftxxN:
name            old time/op       new time/op       delta
RotateLeft-8    3.548750ns +- 1%  2.003750ns +- 0%  -43.54%  (p=0.000 n=8+8)
RotateLeft8-8   3.925000ns +- 0%  3.925000ns +- 0%     ~     (p=1.000 n=8+8)
RotateLeft16-8  3.925000ns +- 0%  3.927500ns +- 0%     ~     (p=0.608 n=8+8)
RotateLeft32-8  3.925000ns +- 0%  2.002500ns +- 0%  -48.98%  (p=0.000 n=8+8)
RotateLeft64-8  3.536250ns +- 0%  2.003750ns +- 0%  -43.34%  (p=0.000 n=8+8)

Change-Id: I77622cd7f39b917427e060647321f5513973232c
Reviewed-on: https://go-review.googlesource.com/122542
Run-TryBot: Ben Shi <powerman1st@163.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
6 years agocmd/dist, go/types: add support for GOARCH=sparc64
Tobias Klauser [Thu, 30 Aug 2018 08:23:54 +0000 (10:23 +0200)]
cmd/dist, go/types: add support for GOARCH=sparc64

This is needed in addition to CL 102555 in order to be able to generate
Go type definitions for linux/sparc64 in the golang.org/x/sys/unix
package.

Change-Id: I928185e320572fecb0c89396f871ea16cba8b9a6
Reviewed-on: https://go-review.googlesource.com/132155
Run-TryBot: Tobias Klauser <tobias.klauser@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
6 years agofmt: add example for Sprint
Thanabodee Charoenpiriyakij [Fri, 7 Sep 2018 12:02:51 +0000 (19:02 +0700)]
fmt: add example for Sprint

Updates #27376

Change-Id: I9ce6541a95b5ecd13f3932558427de1f597df07a
Reviewed-on: https://go-review.googlesource.com/134036
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
6 years agofmt: add example for Print
Thanabodee Charoenpiriyakij [Fri, 7 Sep 2018 11:52:41 +0000 (18:52 +0700)]
fmt: add example for Print

Updates #27376

Change-Id: I2fa63b0d1981a419626072d985e6f3326f6013ff
Reviewed-on: https://go-review.googlesource.com/134035
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
6 years agofmt: add example for Fprint
Thanabodee Charoenpiriyakij [Wed, 5 Sep 2018 06:40:29 +0000 (13:40 +0700)]
fmt: add example for Fprint

Updates #27376

Change-Id: I0ceb672a9fcd7bbf491be1577d7f135ef35b2561
Reviewed-on: https://go-review.googlesource.com/133455
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
6 years agodebug/elf: add R_RISCV_32_PCREL relocation
Tobias Klauser [Fri, 7 Sep 2018 10:45:49 +0000 (12:45 +0200)]
debug/elf: add R_RISCV_32_PCREL relocation

This were missed in CL 107339 as it is not documented (yet) in
https://github.com/riscv/riscv-elf-psabi-doc/blob/master/riscv-elf.md

But binutils already uses it. See
https://github.com/riscv/riscv-elf-psabi-doc/issues/36

Change-Id: I1b084cbf70eb6ac966136bed1bb654883a97b6a9
Reviewed-on: https://go-review.googlesource.com/134015
Run-TryBot: Tobias Klauser <tobias.klauser@gmail.com>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>