]> Cypherpunks repositories - gostls13.git/log
gostls13.git
7 years agoRevert "net: Forget lookups for canceled contexts"
Brad Fitzpatrick [Fri, 17 Nov 2017 17:13:49 +0000 (17:13 +0000)]
Revert "net: Forget lookups for canceled contexts"

This reverts commit 6a3d4be3b80054b1d802b73d5a519e252e0f82ed.

Reason for revert: breaks various builds. See comments on CL 77670

Change-Id: Iaf3260319b560f49ace06af705a2114630f32063
Reviewed-on: https://go-review.googlesource.com/78515
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
7 years agocmd/compile: leave Pos unset for racewalk enter/exit
David Chase [Thu, 16 Nov 2017 02:25:01 +0000 (18:25 -0800)]
cmd/compile: leave Pos unset for racewalk enter/exit

The locations chosen for racewalking inserted code can
be wrong and thus cause unwanted next/step behavior in
debuggers.  Forcing the positions to be unset results in
better behavior.

Test added, and test harness corrected to deal with
changes to gdb's output caused by -racewalk.

Incidental changes in Delve (not part of the usual testing,
but provided because we care about Delve) also reflected
in this CL.

Fixes #22600.

Change-Id: Idd0218afed52ab8c68efd9eabbdff3c92ea2b996
Reviewed-on: https://go-review.googlesource.com/78336
Run-TryBot: David Chase <drchase@google.com>
Reviewed-by: Alessandro Arzilli <alessandro.arzilli@gmail.com>
Reviewed-by: Keith Randall <khr@golang.org>
7 years agocmd/link: enable c-shared and c-archive mode on s390x
Bill O'Farrell [Tue, 31 Oct 2017 23:18:48 +0000 (19:18 -0400)]
cmd/link: enable c-shared and c-archive mode on s390x

Adding s390x to the list of architectures that support c-shared and c-archive.
Required adding load-time initialization (via _rt0_s390x_linux_lib) and adding s390x
to the c-shared and c-archive tests.

Change-Id: I75883b2891c310fe8ce7f08c27b06895c074e123
Reviewed-on: https://go-review.googlesource.com/74910
Reviewed-by: Michael Munday <mike.munday@ibm.com>
7 years agoruntime: remove another TODO
Austin Clements [Wed, 15 Nov 2017 22:43:05 +0000 (14:43 -0800)]
runtime: remove another TODO

I experimented with having the compiler spill the two registers that
are clobbered by the write barrier fast path, but it slightly slows
down compilebench, which is a good write barrier benchmark:

name       old time/op     new time/op     delta
Template       175ms ± 0%      176ms ± 1%    ~           (p=0.393 n=10+10)
Unicode       83.6ms ± 1%     85.1ms ± 2%  +1.79%         (p=0.000 n=9+10)
GoTypes        585ms ± 0%      588ms ± 1%    ~            (p=0.173 n=8+10)
Compiler       2.78s ± 1%      2.81s ± 2%  +0.81%        (p=0.023 n=10+10)
SSA            7.11s ± 1%      7.15s ± 1%  +0.59%        (p=0.029 n=10+10)
Flate          115ms ± 1%      116ms ± 2%    ~           (p=0.853 n=10+10)
GoParser       144ms ± 2%      145ms ± 2%    ~           (p=1.000 n=10+10)
Reflect        389ms ± 1%      390ms ± 1%    ~           (p=0.481 n=10+10)
Tar            185ms ± 2%      185ms ± 2%    ~           (p=0.529 n=10+10)
XML            205ms ± 0%      207ms ± 2%    ~            (p=0.065 n=9+10)

Since this didn't pan out, remove the TODO.

Change-Id: I2186942c6d1ba10585a5da03cd7c1d26ce906273
Reviewed-on: https://go-review.googlesource.com/78034
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Reviewed-by: Rick Hudson <rlh@golang.org>
7 years agoruntime: remove TODO
Austin Clements [Wed, 15 Nov 2017 22:26:33 +0000 (14:26 -0800)]
runtime: remove TODO

I experimented with changing the write barrier to take the value in SI
rather than AX to improve register allocation. It had no effect on
performance and only made the "hello world" text 0.07% smaller, so
let's just remove the comment.

Change-Id: I6a261d14139b7a02a8467b31e74951dfb927ffb4
Reviewed-on: https://go-review.googlesource.com/78033
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Reviewed-by: Rick Hudson <rlh@golang.org>
7 years agodatabase/sql: allow OpenConnector in a driver.Driver interface
Daniel Theophanes [Tue, 14 Nov 2017 16:53:56 +0000 (08:53 -0800)]
database/sql: allow OpenConnector in a driver.Driver interface

While driver.Connector was previously added to allow non-string
connection arguments and access to the context, most users of
the sql package will continue to rely on a string DSN.

Allow drivers to implement a string DSN to Connector interface
that both allows a single parsing of the string DSN and uses
the Connector interface which passes available context to
the driver dialer.

Fixes #22713

Change-Id: Ia0b862262f4c4670effe2538d0d6d43733fea18d
Reviewed-on: https://go-review.googlesource.com/77550
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>

7 years agocmd/go: add import config debugging flag
Russ Cox [Thu, 16 Nov 2017 20:41:44 +0000 (15:41 -0500)]
cmd/go: add import config debugging flag

Change-Id: I11d83c8841f3de4ed4d9d014dec65d6f20464b11
Reviewed-on: https://go-review.googlesource.com/78396
Reviewed-by: David Crawshaw <crawshaw@golang.org>
7 years agocmd/go: simplify compiler import config preparation
Russ Cox [Thu, 16 Nov 2017 19:56:12 +0000 (14:56 -0500)]
cmd/go: simplify compiler import config preparation

I realized this simplification was possible when writing the vet loop
(just above the code being modified here) but never circled back
to make the compiler loop match.

Change-Id: Ic2277d2a4b6d94ea4897cc3615fc1a29f2fb243c
Reviewed-on: https://go-review.googlesource.com/78395
Run-TryBot: Russ Cox <rsc@golang.org>
Reviewed-by: David Crawshaw <crawshaw@golang.org>
7 years agoos/user: handle large 32-bit uid/gid values when stringifying User.Uid/Gid
Brad Fitzpatrick [Wed, 15 Nov 2017 17:30:53 +0000 (17:30 +0000)]
os/user: handle large 32-bit uid/gid values when stringifying User.Uid/Gid

Fixes #22739

Change-Id: I374c29d237c498c9e5ac848b01f6d49d7c41b31f
Reviewed-on: https://go-review.googlesource.com/77930
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
7 years agocmd/go/internal/work: do not write trivial.c when testing gcc flags
Russ Cox [Tue, 14 Nov 2017 15:57:14 +0000 (10:57 -0500)]
cmd/go/internal/work: do not write trivial.c when testing gcc flags

CL 61111 disabled the writing of trivial.c in -n mode, which
made -n mode at least inconsistent with regular mode in
how it was testing for flags. We think that both were getting
the same answer, so avoid creating the file in both modes
to make sure.

If this CL turns out to be wrong, then when we revert it we
should make sure that the empty file is written even in -n mode,
because this check affects the command-line flags printed
by other commands in that mode.

Change-Id: I0a050bfc148fe5a9d430a153d7816b2821277f0d
Reviewed-on: https://go-review.googlesource.com/78115
Run-TryBot: Russ Cox <rsc@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
7 years agocmd/go: fix swig support and run swig tests during run.bash
Russ Cox [Thu, 16 Nov 2017 02:14:21 +0000 (21:14 -0500)]
cmd/go: fix swig support and run swig tests during run.bash

(The tests only run when swig is already installed on the local system.)

Change-Id: I172d106a68cfc746a1058f5a4bcf6761bab88912
Reviewed-on: https://go-review.googlesource.com/78175
Reviewed-by: Ian Lance Taylor <iant@golang.org>
7 years agocmd/go: drop flaky part of TestGoTestJSON
Russ Cox [Thu, 16 Nov 2017 16:43:39 +0000 (11:43 -0500)]
cmd/go: drop flaky part of TestGoTestJSON

Still failing on builders. I give up.
(For example https://build.golang.org/log/4fa5ba031c2fab5df397ac894f8b81392a93728e.)

Change-Id: Ifa87813e27e1adfe9920c7e657b0ff129eb5bf2f
Reviewed-on: https://go-review.googlesource.com/78315
Reviewed-by: Russ Cox <rsc@golang.org>
7 years agocmd/go: fix cgo-disabled test failure from CL 77151
Russ Cox [Thu, 16 Nov 2017 16:56:45 +0000 (11:56 -0500)]
cmd/go: fix cgo-disabled test failure from CL 77151

Change-Id: I8ed4c6dd32a74d94565d77771bb773f006ff90ac
Reviewed-on: https://go-review.googlesource.com/78317
Run-TryBot: Russ Cox <rsc@golang.org>
Reviewed-by: Russ Cox <rsc@golang.org>
7 years agoarchive/tar: partially revert sparse file support
Joe Tsai [Wed, 15 Nov 2017 19:27:10 +0000 (11:27 -0800)]
archive/tar: partially revert sparse file support

This CL removes the following APIs:
type SparseEntry struct{ ... }
type Header struct{ SparseHoles []SparseEntry; ... }
func (*Header) DetectSparseHoles(f *os.File) error
func (*Header) PunchSparseHoles(f *os.File) error
func (*Reader) WriteTo(io.Writer) (int, error)
func (*Writer) ReadFrom(io.Reader) (int, error)

This API was added during the Go1.10 dev cycle, and are safe to remove.

The rationale for reverting is because Header.DetectSparseHoles and
Header.PunchSparseHoles are functionality that probably better belongs in
the os package itself.

The other API like Header.SparseHoles, Reader.WriteTo, and Writer.ReadFrom
perform no OS specific logic and only perform the actual business logic of
reading and writing sparse archives. Since we do know know what the API added to
package os may look like, we preemptively revert these non-OS specific changes
as well by simply commenting them out.

Updates #13548
Updates #22735

Change-Id: I77842acd39a43de63e5c754bfa1c26cc24687b70
Reviewed-on: https://go-review.googlesource.com/78030
Reviewed-by: Russ Cox <rsc@golang.org>
7 years agocmd/go/internal/cache: fix build failure introduced in CL 78176
Russ Cox [Thu, 16 Nov 2017 16:49:33 +0000 (11:49 -0500)]
cmd/go/internal/cache: fix build failure introduced in CL 78176

Maybe a bad git merge - not sure.
In any event, I do miss the trybots.

Noticed while fixing: change print-to-stderr+panic
to pure panic, just so that the test (which catches the panic)
does not print any errors before passing.

Change-Id: If25153ea64e81066455401110ae7a79c36f2f712
Reviewed-on: https://go-review.googlesource.com/78316
Run-TryBot: Russ Cox <rsc@golang.org>
Reviewed-by: Russ Cox <rsc@golang.org>
7 years agocmd/go: implement go clean -testcache
Russ Cox [Sun, 12 Nov 2017 00:50:19 +0000 (19:50 -0500)]
cmd/go: implement go clean -testcache

Ian suggested that since test caching is not expected to be perfect
in all cases, we should allow users to clear the test cache separately
from clearing the entire build cache.

This CL adds 'go clean -testcache' to do that. The implementation
does not actually delete files (for that, use 'go clean -cache').
Instead, it writes down the current time, and future go tests will
ignore any cached test results written before that time.

Change-Id: I4f84065d7dfc2499fa3f203e9ab62e68d7f367c5
Reviewed-on: https://go-review.googlesource.com/78176
Reviewed-by: Ian Lance Taylor <iant@golang.org>
7 years agocmd/go: run cover before cgo
Russ Cox [Fri, 10 Nov 2017 20:58:16 +0000 (15:58 -0500)]
cmd/go: run cover before cgo

If we're running coverage on a package using cgo, we need to
apply both cmd/cover and cmd/cgo as source transformers.
To date we've applied cgo, then cover.

Cover is very sensitive to the exact character position of
expressions in its input, though, and cgo is not, so swap
them, applying first cover and then cgo.

The only drawback here is that coverage formerly applied
to SWIG-generated cgo files, and now it does not.
I am not convinced anyone depended critically on that,
and probably the later analysis with go tool cover would
have tried to parse the original .swig file as a Go file and
gotten very confused.

Fixes #8726.
Fixes #9212.
Fixes #9479.

Change-Id: I777c8b64f7726cb117d59e03073954abc6dfa34d
Reviewed-on: https://go-review.googlesource.com/77155
Reviewed-by: Ian Lance Taylor <iant@golang.org>
7 years agocmd/go: make file:line for cgo files look like non-cgo files
Russ Cox [Fri, 10 Nov 2017 20:46:56 +0000 (15:46 -0500)]
cmd/go: make file:line for cgo files look like non-cgo files

Passing the absolute path to cgo puts the absolute path in the
generated file's //line directives, which then shows that path
in the compiler output, which the go command can then
make relative to the current directory, same as it does for
other compiler output.

Change-Id: Ia2064fea40078c46fd97e3a3b8c9fa1488f913e3
Reviewed-on: https://go-review.googlesource.com/77154
Reviewed-by: Ian Lance Taylor <iant@golang.org>
7 years agocmd/cgo: modify source as text, not as AST
Russ Cox [Fri, 10 Nov 2017 16:45:16 +0000 (11:45 -0500)]
cmd/cgo: modify source as text, not as AST

Cgo has always operated by rewriting the AST and invoking go/printer.
This CL converts it to use the AST to make decisions but then apply
its edits directly to the underlying source text. This approach worked
better in rsc.io/grind (used during the C to Go conversion) and also
more recently in cmd/cover. It guarantees that all comments and
line numbers are preserved exactly.

This eliminates a lot of special concern about comments and
problems with cgo not preserving meaningful comments.
Combined with the CL changing cmd/cover to use the same
approach, it means that the combination of applying cgo and
applying cover still guarantees all comments and line numbers
are preserved exactly.

This sets us up to fix some cgo vs cover bugs by swapping
the order in which they run during the go command.

This also sets up #16623 a bit: the edit list being
accumulated here is nearly exactly what you'd want
to pass to the compiler for that issue.

Change-Id: I7611815be22e7c5c0d4fc3fa11832c42b32c4eb3
Reviewed-on: https://go-review.googlesource.com/77153
Reviewed-by: Ian Lance Taylor <iant@golang.org>
7 years agoencoding/csv: restore Go 1.9 quoted \r\n handling in Reader
Russ Cox [Thu, 16 Nov 2017 15:29:50 +0000 (10:29 -0500)]
encoding/csv: restore Go 1.9 quoted \r\n handling in Reader

CL 52810 changed Reader to interpret a quoted \r\n as a raw \r\n
when reading fields. This seems likely to break existing users, and
discussion on both #21201 (the original issue that triggered the change)
and #22746 (discussing whether to revert the change) failed to identify
a single motivating example for this change. To avoid breaking existing
users for no clear reason, revert the change.

The Reader has been rewritten in the interim so this is not a git revert
but instead and adjustment (and slight simplification) of the new Reader.

Fixes #22746.

Change-Id: Ie857b2f4b1359a207d085b6d3c3a6d440a997d12
Reviewed-on: https://go-review.googlesource.com/78295
Reviewed-by: Joe Tsai <thebrokentoaster@gmail.com>
7 years agocmd/internal/obj, cmd/trace: restore bounds checks dropped in CL 56950
Russ Cox [Tue, 14 Nov 2017 03:32:19 +0000 (22:32 -0500)]
cmd/internal/obj, cmd/trace: restore bounds checks dropped in CL 56950

CL 56950 correctly identified code with checks that were impossible.
But instead of correcting the checks it deleted them.
This CL corrects the code to check what was meant.

Change-Id: Ic89222184ee4fa5cacccae12d750601a9438ac8d
Reviewed-on: https://go-review.googlesource.com/78113
Run-TryBot: Russ Cox <rsc@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
7 years agocmd/go: add dependencies for implicit SWIG imports
Ian Lance Taylor [Wed, 8 Nov 2017 19:06:09 +0000 (11:06 -0800)]
cmd/go: add dependencies for implicit SWIG imports

This fixes the misc/swig tests.

Change-Id: I60c87bbd361fe8b4f69e4507b25dc99a226da3d7
Reviewed-on: https://go-review.googlesource.com/76610
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Russ Cox <rsc@golang.org>
7 years agoRevert "math/rand: make Perm match Shuffle"
Russ Cox [Thu, 16 Nov 2017 02:29:04 +0000 (02:29 +0000)]
Revert "math/rand: make Perm match Shuffle"

This reverts CL 55972.

Reason for revert: this changes Perm's behavior unnecessarily.
I asked for this change originally but I now regret it.
Reverting so that I don't have to justify it in Go 1.10 release notes.

Edited to keep the change to rand_test.go, which seems to have
been mostly unrelated.

Fixes #22744.

Change-Id: If8bb1bcde3ced0db2fdcd0aa65ab128613686c66
Reviewed-on: https://go-review.googlesource.com/78195
Run-TryBot: Russ Cox <rsc@golang.org>
Reviewed-by: Emmanuel Odeke <emm.odeke@gmail.com>
Reviewed-by: Rob Pike <r@golang.org>
Reviewed-by: Joe Tsai <thebrokentoaster@gmail.com>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
7 years agoRevert "time: enable ZONEINFO tzdata file support and error reporting"
Russ Cox [Wed, 15 Nov 2017 17:53:30 +0000 (17:53 +0000)]
Revert "time: enable ZONEINFO tzdata file support and error reporting"

This reverts commit 630d176e7d5e42e21d176d1b9d48de0e03e7cec8.

Reason for revert: the CL moves a parser for what appears to be an
Android-specific file format into the main code and makes it available
on all platforms. Android-specific file formats should be limited to
Android.

Change-Id: I3f19fe03673d65ed1446a0dcf95e5986053e10c0
Reviewed-on: https://go-review.googlesource.com/77950
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Run-TryBot: Russ Cox <rsc@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

7 years agocmd/compile: fix buglet/typo in DWARF x86 setup
Than McIntosh [Wed, 15 Nov 2017 18:45:31 +0000 (13:45 -0500)]
cmd/compile: fix buglet/typo in DWARF x86 setup

Fix typo in DWARF register config for GOOARCH=x86; was
picking up the AMD64 set, should have been selecting
x86 set.

Change-Id: I9a4c6f1378baf3cb2f0ad8d60f3ee2f24cd5dc91
Reviewed-on: https://go-review.googlesource.com/77990
Run-TryBot: Than McIntosh <thanm@google.com>
Reviewed-by: Heschi Kreinick <heschi@google.com>
7 years agoio: eliminate full copy of copy loop in CopyN
Russ Cox [Tue, 14 Nov 2017 19:35:39 +0000 (14:35 -0500)]
io: eliminate full copy of copy loop in CopyN

CL 60630 claimed to and did “improve performance of CopyN”
but in doing so introduced a second copy of the I/O copying loop.
This code is subtle and easy to get wrong and the last thing we
need is of two copies that can drift out of sync. Even the newly
introduced copy contains various subtle changes that are not
obviously semantically equivalent to the original. (They probably
are, but it's not obvious.)

Although the CL description does not explain further what the
important optimization was, it appears that the most critical
one was not allocating a 32kB buffer for CopyN(w, r, 512).

This CL deletes the forked copy of copy and instead applies
the buffer size restriction optimization directly to copy itself.

CL 60630 reported:

name          old time/op    new time/op    delta
CopyNSmall-4    5.09µs ± 1%    2.25µs ±86%  -55.91%  (p=0.000 n=11+14)
CopyNLarge-4     114µs ±73%     121µs ±72%     ~     (p=0.701 n=14+14)

Starting with that CL as the baseline, this CL does not change a ton:

name          old time/op  new time/op  delta
CopyNSmall-8   370ns ± 1%   411ns ± 1%  +11.18%  (p=0.000 n=16+14)
CopyNLarge-8  18.2µs ± 1%  18.3µs ± 1%   +0.63%  (p=0.000 n=19+20)

It does give up a small amount of the win of 60630 but preserves
the bulk of it, with the benefit that we will not need to debug these
two copies drifting out of sync in the future.

Change-Id: I05b1a5a7115390c5867847cba606b75d513eb2e2
Reviewed-on: https://go-review.googlesource.com/78122
Run-TryBot: Russ Cox <rsc@golang.org>
Reviewed-by: Joe Tsai <thebrokentoaster@gmail.com>
7 years agocmd/cover: do not report coverage for assembly functions
Russ Cox [Fri, 10 Nov 2017 04:58:42 +0000 (23:58 -0500)]
cmd/cover: do not report coverage for assembly functions

cover -func mode was reporting a coverage for function
declarations without bodies - assembly functions.
Since we are not annotating their code, we have no data
for those functions and should not report them at all.

Fixes #6880.

Change-Id: I4b8cd90805accf61f54e3ee167f54f4dc10c7c59
Reviewed-on: https://go-review.googlesource.com/77152
Reviewed-by: Ian Lance Taylor <iant@golang.org>
7 years agocmd/cover: add //line comment pointing to original file
Russ Cox [Fri, 10 Nov 2017 04:55:15 +0000 (23:55 -0500)]
cmd/cover: add //line comment pointing to original file

Now that cover does not modify the formatting of the original file
or add any newline characters, we can make it print a //line comment
pointing back at the original, and compiler errors and panics will
report accurate line numbers.

Fixes #6329.
Fixes #15757.

Change-Id: I7b0e386112c69beafe69e0d47c5f9e9abc87c0f5
Reviewed-on: https://go-review.googlesource.com/77151
Run-TryBot: Russ Cox <rsc@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
7 years agoarchive/zip: add documentation about compression methods
Joe Kyo [Wed, 15 Nov 2017 10:24:21 +0000 (10:24 +0000)]
archive/zip: add documentation about compression methods

Change-Id: I491c5ddd1a5d8e55f8e6bb9377bc3811e42773f8
Reviewed-on: https://go-review.googlesource.com/77870
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
7 years agocmd/go: deflake TestGoTestJSON further
Russ Cox [Mon, 13 Nov 2017 15:44:32 +0000 (10:44 -0500)]
cmd/go: deflake TestGoTestJSON further

Even after disabling on 1-CPU systems, builders are still flaking too often.
Unless there are at least 4 CPUs, don't require test interlacing at all.

Fixes #22665 (again).

Change-Id: Ief792c496c1ee70939532e6ca8bef012fe78178e
Reviewed-on: https://go-review.googlesource.com/77310
Run-TryBot: Russ Cox <rsc@golang.org>
Reviewed-by: Russ Cox <rsc@golang.org>
7 years agodatabase/sql/driver: document that Execer, Queryer need not be implemented
Russ Cox [Wed, 15 Nov 2017 20:19:31 +0000 (15:19 -0500)]
database/sql/driver: document that Execer, Queryer need not be implemented

CL 21663 allowed drivers to implement ExecerContext without
also implementing Execer, and similarly QueryerContext without
Queryer, but it did not make that clear in the documentation.

This CL updates the documentation.

Change-Id: I9a4accaac32edfe255fe7c0b0907d4c1014322b4
Reviewed-on: https://go-review.googlesource.com/78129
Reviewed-by: Daniel Theophanes <kardianos@gmail.com>
7 years agoencoding/hex: make Decode, Decoder, DecodeString agree about partial results and...
Russ Cox [Tue, 14 Nov 2017 18:43:02 +0000 (13:43 -0500)]
encoding/hex: make Decode, Decoder, DecodeString agree about partial results and errors

CL 70210 added Decoder for #21590, and in doing so it changed
the existing func Decode to return partial results for decoding errors.
That seems like a good change to make to Decode, but it was
untested (except as used by Decoder), inconsistent with DecodeString
in all error cases, and inconsistent with Decoder in not returning
partial results for odd-length input strings.

This CL makes Decode, DecodeString, and Decoder all agree about
the handling of partial results (they are returned) and error
precedence (the error earliest in the input is reported),
and it documents and tests this.

Change-Id: Ifb7d1e100ecb66fe2ed5ba34a621084d480f16db
Reviewed-on: https://go-review.googlesource.com/78120
Run-TryBot: Russ Cox <rsc@golang.org>
Reviewed-by: Joe Tsai <thebrokentoaster@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

7 years agocmd/vet: shorten diagnostic about suspicious struct tag spaces
Russ Cox [Tue, 14 Nov 2017 16:27:03 +0000 (11:27 -0500)]
cmd/vet: shorten diagnostic about suspicious struct tag spaces

Change-Id: I112d0164df6abd9ca1df287376cf3605268385df
Reviewed-on: https://go-review.googlesource.com/78116
Run-TryBot: Russ Cox <rsc@golang.org>
Reviewed-by: Rob Pike <r@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

7 years agocrypto/cipher: use raw bytes for keys in docs
Kevin Burke [Fri, 14 Jul 2017 22:20:27 +0000 (16:20 -0600)]
crypto/cipher: use raw bytes for keys in docs

Using ASCII values for keys is a bad idea since it makes them vastly
easier to guess. Instead, use the same method as the examples in the
golang.org/x/crypto/nacl package to load keys from a hex value.

Changing the key required updating the ciphertext in many of the
examples.

I am still worried about the fact the examples ask the user to
authenticate messages; authentication isn't trivial, and to be honest
it may be better to steer people to a higher level primitive like
secretbox, unless people really need AES.

Fixes #21012.

Change-Id: I8d918cf194694cd380b06c2d561178167ca61adb
Reviewed-on: https://go-review.googlesource.com/48596
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
7 years agoencoding/asn1: add MarshalWithParams
Hiroshi Ioka [Wed, 24 May 2017 23:57:03 +0000 (08:57 +0900)]
encoding/asn1: add MarshalWithParams

Fixes #18873

Change-Id: Idb9750f739f91ebca34efcbc177254d412b4d90d
Reviewed-on: https://go-review.googlesource.com/44111
Reviewed-by: Adam Langley <agl@golang.org>
Run-TryBot: Adam Langley <agl@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

7 years agonet: Forget lookups for canceled contexts
Troels Thomsen [Tue, 14 Nov 2017 22:22:19 +0000 (23:22 +0100)]
net: Forget lookups for canceled contexts

A sequential lookup using any non-canceled context has a risk of
returning the result of the previous lookup for a canceled context (i.e.
an error).

This is already prevented for timed out context by forgetting the host
immediately and extending this to also compare the error to
`context.Canceled` resolves this issue.

Fixes #22724

Change-Id: I7aafa1459a0de4dc5c4332988fbea23cbf4dba07
Reviewed-on: https://go-review.googlesource.com/77670
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
7 years agoarchive/zip: fix handling of replacement rune in UTF8 check
Russ Cox [Wed, 15 Nov 2017 16:47:31 +0000 (11:47 -0500)]
archive/zip: fix handling of replacement rune in UTF8 check

The replacement rune is a valid rune and can appear as itself in valid UTF8
(it encodes as three bytes). To check for invalid UTF8 it is necessary to
look for utf8.DecodeRune returning the replacement rune and size==1.

Change-Id: I169be8d1fe61605c921ac13cc2fde94f80f3463c
Reviewed-on: https://go-review.googlesource.com/78126
Run-TryBot: Russ Cox <rsc@golang.org>
Reviewed-by: Joe Tsai <thebrokentoaster@gmail.com>
7 years agoapi: update next.txt
Russ Cox [Wed, 15 Nov 2017 21:06:00 +0000 (16:06 -0500)]
api: update next.txt

Change-Id: I540bdabe8ffda4697315fa6f09ad710c05b4a94d
Reviewed-on: https://go-review.googlesource.com/78134
Run-TryBot: Russ Cox <rsc@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
7 years agodatabase/sql: fix ctx.Done() == nil check
Russ Cox [Wed, 15 Nov 2017 20:14:07 +0000 (15:14 -0500)]
database/sql: fix ctx.Done() == nil check

ctx.Done() == ctx.Background().Done() is just
a long way to write ctx.Done() == nil.
Use the short way.

Change-Id: I7b3198b5dc46b8b40086243aa61882bc8c268eac
Reviewed-on: https://go-review.googlesource.com/78128
Run-TryBot: Russ Cox <rsc@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
7 years agodatabase/sql/driver: rename ResetSessioner to SessionResetter
Russ Cox [Wed, 15 Nov 2017 20:07:44 +0000 (15:07 -0500)]
database/sql/driver: rename ResetSessioner to SessionResetter

Originally we tried the strict -er suffix as the rule in this case
but eventually we decided it was too awkward: io.WriteByter
became io.ByteWriter. By analogy, here the interface should be
named SessionResetter instead of the awkward ResetSessioner.

This change should not affect any drivers that have already
implemented the interface, because the method name is not changing.

(This was added during the Go 1.10 cycle and has not been
released yet, so we can change it.)

Change-Id: Ie50e4e090d3811f85965da9da37d966e9f45e79d
Reviewed-on: https://go-review.googlesource.com/78127
Run-TryBot: Russ Cox <rsc@golang.org>
Reviewed-by: Daniel Theophanes <kardianos@gmail.com>
7 years agonet/mail: clarify doc comment
Russ Cox [Tue, 14 Nov 2017 20:33:17 +0000 (15:33 -0500)]
net/mail: clarify doc comment

Rewrite the text added in CL 50911, which I did not understand.

Change-Id: Id6271ffe2f7c8833dd7733fe0254fa4927fac150
Reviewed-on: https://go-review.googlesource.com/78124
Run-TryBot: Russ Cox <rsc@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
7 years agoio: document and test MultiWriter error behavior
Russ Cox [Tue, 14 Nov 2017 19:54:10 +0000 (14:54 -0500)]
io: document and test MultiWriter error behavior

MultiWriter(w1, w2) only writes to w2 if w1.Write succeeds.
I did not know this, and it was not documented.
Document and test.

Change-Id: Idec2e8444d5a7aca0b95d07814a28daa454eb1d3
Reviewed-on: https://go-review.googlesource.com/78123
Run-TryBot: Russ Cox <rsc@golang.org>
Reviewed-by: Joe Tsai <thebrokentoaster@gmail.com>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
7 years agoencoding/xml: move unexported const out of exported const block
Russ Cox [Tue, 14 Nov 2017 19:01:11 +0000 (14:01 -0500)]
encoding/xml: move unexported const out of exported const block

CL 58210 introduced this constant for reasons I don't understand.
It should not be in the exported const block, which will pollute
godoc output with a "... unexported" notice.

Also since we already have a constant named xmlnsPrefix for "xmlns",
it is very confusing to also have xmlNamespacePrefix for "xml".
If we must have the constant at all, rename it to xmlPrefix.

Change-Id: I15f937454d730005816fcd32b1acca703acf1e51
Reviewed-on: https://go-review.googlesource.com/78121
Run-TryBot: Russ Cox <rsc@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
7 years agoencoding/csv: rename ParseError.RecordLine to .StartLine
Russ Cox [Tue, 14 Nov 2017 18:23:10 +0000 (13:23 -0500)]
encoding/csv: rename ParseError.RecordLine to .StartLine

A record can span multiple lines (the whole reason for the extra field),
so the important fact is that it's the _start_ of the record.
Make that clear in the name.

(This API was added during the Go 1.10 cycle so it can still be cleaned up.)

Change-Id: Id95b3ceb7cdfc4aa0ed5a053cb84da8945fa5496
Reviewed-on: https://go-review.googlesource.com/78119
Run-TryBot: Russ Cox <rsc@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Reviewed-by: Joe Tsai <thebrokentoaster@gmail.com>
7 years agoencoding/binary: make new example a bit more idiomatic
Russ Cox [Tue, 14 Nov 2017 17:38:11 +0000 (12:38 -0500)]
encoding/binary: make new example a bit more idiomatic

Mainly get rid of the weird zero-value struct literal,
but while we're here also group and order things a bit better:
first the reader, then the data, then the call (which takes reader then data).

Change-Id: I901b0661d85d8eaa0807e4482aac66500ca996c7
Reviewed-on: https://go-review.googlesource.com/78118
Run-TryBot: Russ Cox <rsc@golang.org>
Reviewed-by: Joe Tsai <thebrokentoaster@gmail.com>
7 years agodoc/contrib.html: add Go 1.10 to release list
Russ Cox [Tue, 14 Nov 2017 17:26:50 +0000 (12:26 -0500)]
doc/contrib.html: add Go 1.10 to release list

Apparently we maintain this list by hand (for example, CL 52351).

Change-Id: I0a0b346cf2b7b547729cb1d0fa1642de447f7bba
Reviewed-on: https://go-review.googlesource.com/78117
Run-TryBot: Russ Cox <rsc@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
7 years agocmd/dist: adjust package doc
Russ Cox [Tue, 14 Nov 2017 04:06:54 +0000 (23:06 -0500)]
cmd/dist: adjust package doc

Mainly capitalize the first letter.
(Followup to CL 54351.)

Change-Id: I2d5c3d72c53d3468de7a9d4af8bd009182ff3d38
Reviewed-on: https://go-review.googlesource.com/78114
Run-TryBot: Russ Cox <rsc@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
7 years agobytes, strings: restore O(1) behavior of IndexAny(s, "") and LastIndexAny(s, "")
Russ Cox [Tue, 14 Nov 2017 02:30:22 +0000 (21:30 -0500)]
bytes, strings: restore O(1) behavior of IndexAny(s, "") and LastIndexAny(s, "")

CL 65851 (bytes) and CL 65910 (strings) “improve[d] readability”
by removing the special case that bypassed the whole function body
when chars == "". In doing so, yes, the function was unindented a
level, which is nice, but the runtime of that case went from O(1) to O(n)
where n = len(s).

I don't know if anyone's code depends on the O(1) behavior in this case,
but quite possibly someone's does.

This CL adds the special case back, with a comment to prevent future
deletions, and without reindenting each function body in full.

Change-Id: I5aba33922b304dd1b8657e6d51d6c937a7f95c81
Reviewed-on: https://go-review.googlesource.com/78112
Run-TryBot: Russ Cox <rsc@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
7 years agobytes: make ExampleTrimLeft and ExampleTrimRight match
Russ Cox [Tue, 14 Nov 2017 02:21:59 +0000 (21:21 -0500)]
bytes: make ExampleTrimLeft and ExampleTrimRight match

ExampleTrimLeft was inexplicably complex.

Change-Id: I13ca81bdeba728bdd632acf82e3a1101d29b9f39
Reviewed-on: https://go-review.googlesource.com/78111
Run-TryBot: Russ Cox <rsc@golang.org>
Reviewed-by: Joe Tsai <thebrokentoaster@gmail.com>
7 years agobytes: change ExampleReader_Len to use a non-ASCII string
Russ Cox [Tue, 14 Nov 2017 02:20:23 +0000 (21:20 -0500)]
bytes: change ExampleReader_Len to use a non-ASCII string

This should help make clear that Len is not counting runes.
Also delete empty string, which doesn't add much.

Change-Id: I1602352df1897fef6e855e9db0bababb8ab788ca
Reviewed-on: https://go-review.googlesource.com/78110
Run-TryBot: Russ Cox <rsc@golang.org>
Reviewed-by: Joe Tsai <thebrokentoaster@gmail.com>
7 years agodoc/go1.10: first draft of release notes
Russ Cox [Wed, 15 Nov 2017 20:44:09 +0000 (15:44 -0500)]
doc/go1.10: first draft of release notes

Change-Id: If7ec07be4ecb0c1d6a1eb5c0740f150473aea6fa
Reviewed-on: https://go-review.googlesource.com/78130
Run-TryBot: Russ Cox <rsc@golang.org>
Reviewed-by: Russ Cox <rsc@golang.org>
7 years agoarchive/tar: change error prefix
Joe Tsai [Wed, 15 Nov 2017 18:49:35 +0000 (10:49 -0800)]
archive/tar: change error prefix

Change error message prefix from "tar:" to "archive/tar:" to maintain
backwards compatibility with Go1.9 and earlier in the unfortunate event
that someone is relying on string parsing of errors.

Fixes #22740

Change-Id: I59039c59818a0599e9d3b06bb5a531aa22a389b8
Reviewed-on: https://go-review.googlesource.com/77933
Reviewed-by: roger peppe <rogpeppe@gmail.com>
7 years agogo/printer: indent lone comments in composite lits
Daniel Martí [Sun, 29 Oct 2017 21:04:01 +0000 (21:04 +0000)]
go/printer: indent lone comments in composite lits

If a composite literal contains any comments on their own lines without
any elements, the printer would unindent the comments.

The comments in this edge case are written when the closing '}' is
written. Indent and outdent first so that the indentation is
interspersed before the comment is written.

Also note that the go/printer golden tests don't show the exact same
behaviour that gofmt does. Added a TODO to figure this out in a separate
CL.

While at it, ensure that the tree conforms to gofmt. The changes are
unrelated to this indentation fix, however.

Fixes #22355.

Change-Id: I5ac25ac6de95a236f1e123479127cc4dd71e93fe
Reviewed-on: https://go-review.googlesource.com/74232
Run-TryBot: Daniel Martí <mvdan@mvdan.cc>
Reviewed-by: Robert Griesemer <gri@golang.org>
7 years agoruntime: fix gctrace STW CPU time and CPU fraction
Austin Clements [Tue, 14 Nov 2017 23:08:32 +0000 (15:08 -0800)]
runtime: fix gctrace STW CPU time and CPU fraction

The CPU time reported in the gctrace for STW phases is simply
work.stwprocs times the wall-clock duration of these phases. However,
work.stwprocs is set to gcprocs(), which is wrong for multiple
reasons:

1. gcprocs is intended to limit the number of Ms used for mark
   termination based on how well the garbage collector actually
   scales, but the gctrace wants to report how much CPU time is being
   stolen from the application. During STW, that's *all* of the CPU,
   regardless of how many the garbage collector can actually use.

2. gcprocs assumes it's being called during STW, so it limits its
   result to sched.nmidle+1. However, we're not calling it during STW,
   so sched.nmidle is typically quite small, even if GOMAXPROCS is
   quite large.

Fix this by setting work.stwprocs to min(ncpu, GOMAXPROCS). This also
fixes the overall GC CPU fraction, which is based on the computed CPU
times.

Fixes #22725.

Change-Id: I64b5ce87e28dbec6870aa068ce7aecdd28c058d1
Reviewed-on: https://go-review.googlesource.com/77710
Run-TryBot: Austin Clements <austin@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Rick Hudson <rlh@golang.org>
7 years agobytes,strings: in generic Index, use mix of IndexByte and Rabin-Karp
Keith Randall [Sat, 4 Nov 2017 17:19:53 +0000 (10:19 -0700)]
bytes,strings: in generic Index, use mix of IndexByte and Rabin-Karp

Use IndexByte first, as it allows us to skip lots of bytes quickly.
If IndexByte is generating a lot of false positives, switch over to Rabin-Karp.

Experiments for ppc64le
bytes:
name                             old time/op  new time/op  delta
IndexPeriodic/IndexPeriodic2-2   1.12ms ± 0%  0.18ms ± 0%  -83.54%  (p=0.000 n=10+9)
IndexPeriodic/IndexPeriodic4-2    635µs ± 0%   184µs ± 0%  -71.06%  (p=0.000 n=9+9)
IndexPeriodic/IndexPeriodic8-2    289µs ± 0%   184µs ± 0%  -36.51%  (p=0.000 n=10+9)
IndexPeriodic/IndexPeriodic16-2   133µs ± 0%   183µs ± 0%  +37.68%  (p=0.000 n=10+9)
IndexPeriodic/IndexPeriodic32-2  68.3µs ± 0%  70.2µs ± 0%   +2.76%  (p=0.000 n=10+10)
IndexPeriodic/IndexPeriodic64-2  35.8µs ± 0%  36.6µs ± 0%   +2.17%  (p=0.000 n=8+10)

strings:
name                             old time/op  new time/op  delta
IndexPeriodic/IndexPeriodic2-2    184µs ± 0%   184µs ± 0%   +0.11%  (p=0.029 n=4+4)
IndexPeriodic/IndexPeriodic4-2    184µs ± 0%   184µs ± 0%     ~     (p=0.886 n=4+4)
IndexPeriodic/IndexPeriodic8-2    184µs ± 0%   184µs ± 0%     ~     (p=0.486 n=4+4)
IndexPeriodic/IndexPeriodic16-2   185µs ± 1%   184µs ± 0%     ~     (p=0.343 n=4+4)
IndexPeriodic/IndexPeriodic32-2   184µs ± 0%    69µs ± 0%  -62.37%  (p=0.029 n=4+4)
IndexPeriodic/IndexPeriodic64-2   184µs ± 0%    37µs ± 0%  -80.17%  (p=0.029 n=4+4)

Fixes #22578

Change-Id: If2a4d8554cb96bfd699b58149d13ac294615f8b8
Reviewed-on: https://go-review.googlesource.com/76070
Reviewed-by: Alberto Donizetti <alb.donizetti@gmail.com>
7 years agocrypto/sha1, crypto/sha256: add go:noescape annotations
Brad Fitzpatrick [Tue, 14 Nov 2017 19:22:43 +0000 (19:22 +0000)]
crypto/sha1, crypto/sha256: add go:noescape annotations

Additions to:
https://go-review.googlesource.com/c/go/+/61570
https://go-review.googlesource.com/c/go/+/61550

Change-Id: Id89e1119333a8721cb9720a04a01dab1f2705fa9
Reviewed-on: https://go-review.googlesource.com/77591
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
7 years agoruntime/trace: fix a typo in doc
Hana Kim [Wed, 30 Aug 2017 16:24:37 +0000 (12:24 -0400)]
runtime/trace: fix a typo in doc

Change-Id: I63f3d2edb09801c99957a1f744639523fb6d0b62
Reviewed-on: https://go-review.googlesource.com/60331
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
7 years agoarchive/tar: remove useless type conversions
Awn [Mon, 3 Jul 2017 00:36:52 +0000 (01:36 +0100)]
archive/tar: remove useless type conversions

Change-Id: I259a6ed6a1abc63d2dc39eca7e85f94cf38001cc
Reviewed-on: https://go-review.googlesource.com/47342
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

7 years agocrypto/sha1, crypto/sha256: fix typo fisrt -> first
Marko Mudrinic [Wed, 15 Nov 2017 12:56:01 +0000 (12:56 +0000)]
crypto/sha1, crypto/sha256: fix typo fisrt -> first

Change-Id: I9bb7568ca16b420254796eb38dfb86c5ca3399d7
Reviewed-on: https://go-review.googlesource.com/77890
Reviewed-by: Ian Lance Taylor <iant@golang.org>
7 years agocmd/cgo: permit passing string values directly between Go and C
Ian Lance Taylor [Sat, 14 Oct 2017 01:26:10 +0000 (18:26 -0700)]
cmd/cgo: permit passing string values directly between Go and C

Permit the C preamble to use the _GoString_ type. Permit Go code to
pass string values directly to those C types. Add accessors for C
code to retrieve sizes and pointers.

Fixes #6907

Change-Id: I190c88319ec88a3ef0ddb99f342a843ba69fcaa3
Reviewed-on: https://go-review.googlesource.com/70890
Run-TryBot: Ian Lance Taylor <iant@golang.org>
Reviewed-by: Austin Clements <austin@google.com>
7 years agoencoding/pem: add Encode example
Tim Cooper [Wed, 15 Nov 2017 01:32:00 +0000 (21:32 -0400)]
encoding/pem: add Encode example

Change-Id: Ib9ec3524b712e016a9dd2fbee5555362c1a0cb59
Reviewed-on: https://go-review.googlesource.com/77770
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

7 years agoruntime: IndexByte and memclr perf improvements on arm64
wei xiao [Mon, 28 Nov 2016 02:55:43 +0000 (10:55 +0800)]
runtime: IndexByte and memclr perf improvements on arm64

Update runtime asm_arm64.s and memclr_arm64.s to improve performance by using
SIMD instructions to do more in parallel. It shows improvement on bytes, html
and go1 benchmarks (particualrly regexp, which uses IndexByte frequently).

Benchmark results of bytes:

name                     old time/op   new time/op    delta
IndexByte/10-8            28.5ns ± 0%    19.5ns ± 0%   -31.58%  (p=0.000 n=10+10)
IndexByte/32-8            52.6ns ± 0%    19.0ns ± 0%   -63.88%  (p=0.000 n=10+10)
IndexByte/4K-8            4.12µs ± 0%    0.49µs ± 0%   -88.16%  (p=0.000 n=10+10)
IndexByte/4M-8            4.29ms ± 1%    0.70ms ±26%   -83.65%  (p=0.000 n=10+10)
IndexByte/64M-8           69.7ms ± 0%    16.0ms ± 0%   -76.97%  (p=0.000 n=9+10)
IndexBytePortable/10-8    34.0ns ± 0%    34.0ns ± 0%      ~     (all equal)
IndexBytePortable/32-8    66.1ns ± 0%    66.1ns ± 0%      ~     (p=0.471 n=9+9)
IndexBytePortable/4K-8    6.17µs ± 0%    6.17µs ± 0%      ~     (all equal)
IndexBytePortable/4M-8    6.33ms ± 0%    6.35ms ± 0%    +0.21%  (p=0.002 n=10+9)
IndexBytePortable/64M-8    103ms ± 0%     103ms ± 0%    +0.01%  (p=0.017 n=9+10)

name                     old speed     new speed      delta
IndexByte/10-8           351MB/s ± 0%   512MB/s ± 0%   +46.14%  (p=0.000 n=9+10)
IndexByte/32-8           609MB/s ± 0%  1683MB/s ± 0%  +176.40%  (p=0.000 n=10+10)
IndexByte/4K-8           994MB/s ± 0%  8378MB/s ± 0%  +742.75%  (p=0.000 n=10+10)
IndexByte/4M-8           977MB/s ± 1%  6149MB/s ±32%  +529.29%  (p=0.000 n=10+10)
IndexByte/64M-8          963MB/s ± 0%  4182MB/s ± 0%  +334.29%  (p=0.000 n=9+10)
IndexBytePortable/10-8   294MB/s ± 0%   294MB/s ± 0%    +0.17%  (p=0.000 n=8+8)
IndexBytePortable/32-8   484MB/s ± 0%   484MB/s ± 0%      ~     (p=0.877 n=9+9)
IndexBytePortable/4K-8   664MB/s ± 0%   664MB/s ± 0%      ~     (p=0.242 n=8+9)
IndexBytePortable/4M-8   662MB/s ± 0%   661MB/s ± 0%    -0.21%  (p=0.002 n=10+9)
IndexBytePortable/64M-8  652MB/s ± 0%   652MB/s ± 0%      ~     (p=0.065 n=10+10)

Benchmark results of html:

name              old time/op  new time/op  delta
Escape-8          62.0µs ± 1%  61.0µs ± 1%   -1.69%  (p=0.000 n=9+10)
EscapeNone-8      10.2µs ± 0%  10.2µs ± 0%   -0.09%  (p=0.022 n=9+10)
Unescape-8        71.9µs ± 0%  68.7µs ± 0%   -4.35%  (p=0.000 n=10+10)
UnescapeNone-8    4.03µs ± 0%  0.48µs ± 0%  -88.08%  (p=0.000 n=10+10)
UnescapeSparse-8  10.7µs ± 2%   7.1µs ± 3%  -33.91%  (p=0.000 n=10+10)
UnescapeDense-8   53.2µs ± 1%  53.5µs ± 1%     ~     (p=0.143 n=10+10)

Benchmark results of go1:

name                     old time/op    new time/op    delta
BinaryTree17-8              6.53s ± 0%     6.48s ± 2%      ~     (p=0.190 n=4+5)
Fannkuch11-8                6.35s ± 1%     6.35s ± 0%      ~     (p=1.000 n=5+5)
FmtFprintfEmpty-8           108ns ± 1%     101ns ± 2%    -6.32%  (p=0.008 n=5+5)
FmtFprintfString-8          172ns ± 1%     182ns ± 2%    +5.70%  (p=0.008 n=5+5)
FmtFprintfInt-8             207ns ± 0%     207ns ± 0%      ~     (p=0.444 n=5+5)
FmtFprintfIntInt-8          277ns ± 1%     276ns ± 1%      ~     (p=0.873 n=5+5)
FmtFprintfPrefixedInt-8     386ns ± 0%     382ns ± 1%    -1.04%  (p=0.024 n=5+5)
FmtFprintfFloat-8           492ns ± 0%     492ns ± 1%      ~     (p=0.571 n=4+5)
FmtManyArgs-8              1.32µs ± 1%    1.33µs ± 0%      ~     (p=0.087 n=5+5)
GobDecode-8                16.8ms ± 2%    16.7ms ± 1%      ~     (p=1.000 n=5+5)
GobEncode-8                14.1ms ± 1%    14.0ms ± 1%      ~     (p=0.056 n=5+5)
Gzip-8                      788ms ± 0%     802ms ± 0%    +1.71%  (p=0.008 n=5+5)
Gunzip-8                   83.6ms ± 0%    83.9ms ± 0%    +0.40%  (p=0.008 n=5+5)
HTTPClientServer-8          120µs ± 0%     120µs ± 1%      ~     (p=0.548 n=5+5)
JSONEncode-8               33.2ms ± 0%    33.0ms ± 1%    -0.71%  (p=0.008 n=5+5)
JSONDecode-8                152ms ± 1%     152ms ± 1%      ~     (p=1.000 n=5+5)
Mandelbrot200-8            10.0ms ± 0%    10.0ms ± 0%    -0.05%  (p=0.008 n=5+5)
GoParse-8                  7.97ms ± 0%    7.98ms ± 0%      ~     (p=0.690 n=5+5)
RegexpMatchEasy0_32-8       233ns ± 1%     206ns ± 0%   -11.44%  (p=0.016 n=5+4)
RegexpMatchEasy0_1K-8      1.86µs ± 0%    0.77µs ± 1%   -58.54%  (p=0.008 n=5+5)
RegexpMatchEasy1_32-8       250ns ± 0%     205ns ± 0%   -18.07%  (p=0.008 n=5+5)
RegexpMatchEasy1_1K-8      2.28µs ± 0%    1.11µs ± 0%   -51.09%  (p=0.029 n=4+4)
RegexpMatchMedium_32-8      332ns ± 1%     301ns ± 2%    -9.45%  (p=0.008 n=5+5)
RegexpMatchMedium_1K-8     85.5µs ± 2%    78.8µs ± 0%    -7.83%  (p=0.008 n=5+5)
RegexpMatchHard_32-8       4.34µs ± 1%    4.27µs ± 0%    -1.49%  (p=0.008 n=5+5)
RegexpMatchHard_1K-8        130µs ± 1%     127µs ± 0%    -2.53%  (p=0.008 n=5+5)
Revcomp-8                   1.35s ± 1%     1.13s ± 1%   -16.17%  (p=0.008 n=5+5)
Template-8                  160ms ± 2%     162ms ± 2%      ~     (p=0.222 n=5+5)
TimeParse-8                 795ns ± 2%     778ns ± 1%      ~     (p=0.095 n=5+5)
TimeFormat-8                782ns ± 0%     786ns ± 1%    +0.59%  (p=0.040 n=5+5)

name                     old speed      new speed      delta
GobDecode-8              45.8MB/s ± 2%  45.9MB/s ± 1%      ~     (p=1.000 n=5+5)
GobEncode-8              54.3MB/s ± 1%  55.0MB/s ± 1%      ~     (p=0.056 n=5+5)
Gzip-8                   24.6MB/s ± 0%  24.2MB/s ± 0%    -1.69%  (p=0.008 n=5+5)
Gunzip-8                  232MB/s ± 0%   231MB/s ± 0%    -0.40%  (p=0.008 n=5+5)
JSONEncode-8             58.4MB/s ± 0%  58.8MB/s ± 1%    +0.71%  (p=0.008 n=5+5)
JSONDecode-8             12.8MB/s ± 1%  12.8MB/s ± 1%      ~     (p=1.000 n=5+5)
GoParse-8                7.27MB/s ± 0%  7.26MB/s ± 0%      ~     (p=0.762 n=5+5)
RegexpMatchEasy0_32-8     137MB/s ± 1%   155MB/s ± 0%   +12.93%  (p=0.008 n=5+5)
RegexpMatchEasy0_1K-8     551MB/s ± 0%  1329MB/s ± 1%  +141.11%  (p=0.008 n=5+5)
RegexpMatchEasy1_32-8     128MB/s ± 0%   156MB/s ± 0%   +22.00%  (p=0.008 n=5+5)
RegexpMatchEasy1_1K-8     449MB/s ± 0%   920MB/s ± 0%  +104.68%  (p=0.016 n=4+5)
RegexpMatchMedium_32-8   3.00MB/s ± 0%  3.32MB/s ± 2%   +10.60%  (p=0.016 n=4+5)
RegexpMatchMedium_1K-8   12.0MB/s ± 2%  13.0MB/s ± 0%    +8.48%  (p=0.008 n=5+5)
RegexpMatchHard_32-8     7.38MB/s ± 1%  7.49MB/s ± 0%    +1.49%  (p=0.008 n=5+5)
RegexpMatchHard_1K-8     7.88MB/s ± 1%  8.08MB/s ± 0%    +2.59%  (p=0.008 n=5+5)
Revcomp-8                 188MB/s ± 1%   224MB/s ± 1%   +19.29%  (p=0.008 n=5+5)
Template-8               12.2MB/s ± 2%  12.0MB/s ± 2%      ~     (p=0.206 n=5+5)

Change-Id: I94116620a287d173a6f60510684362e500f54887
Reviewed-on: https://go-review.googlesource.com/33597
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
7 years agonet/url: allow *User functions to work on a nil receiver.
OneOfOne [Sat, 8 Jul 2017 18:19:25 +0000 (20:19 +0200)]
net/url: allow *User functions to work on a nil receiver.

Fixes #20924

Change-Id: If89f31da63cbea38d7e615a428b7b07629770a45
Reviewed-on: https://go-review.googlesource.com/47851
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Tim Cooper <tim.cooper@layeh.com>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
7 years agoencoding/pem: add Encode, EncodeToMemory docs
Tim Cooper [Wed, 15 Nov 2017 01:54:06 +0000 (21:54 -0400)]
encoding/pem: add Encode, EncodeToMemory docs

Included in a warning that EncodeToMemory may return an incomplete PEM
encoded structure if invalid headers are supplied. Example:

pem.EncodeToMemory(&pem.Block{
Headers: map[string]string{
"a":   "test1",
"b:c": "test2",
},
})

Returns:

-----BEGIN -----
a: test1

Change-Id: Ia9cf0202f985e3cf210aabb6f07667e581ff081f
Reviewed-on: https://go-review.googlesource.com/77790
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
7 years agonet/http: don't set Content-Type with empty body automatically
Tw [Sat, 24 Jun 2017 06:23:17 +0000 (14:23 +0800)]
net/http: don't set Content-Type with empty body automatically

We set Content-Type to "text/plain; charset=utf-8" even with blank body
before. Let's strip this unnecessary header though it's harmless in most
cases.

Fixes #20784

Signed-off-by: Tw <tw19881113@gmail.com>
Change-Id: Ic58a410dcbc89f457c6ddd92961d9cbf545b2f4f
Reviewed-on: https://go-review.googlesource.com/46631
Run-TryBot: Emmanuel Odeke <emm.odeke@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
7 years agocrypto, hash: document marshal/unmarshal implementation
Roger Peppe [Sun, 12 Nov 2017 22:19:11 +0000 (22:19 +0000)]
crypto, hash: document marshal/unmarshal implementation

Unless you go back and read the hash package documentation, it's
not clear that all the hash packages implement marshaling and
unmarshaling. Document the behaviour specifically in each package
that implements it as it this is hidden behaviour and easy to miss.

Change-Id: Id9d3508909362f1a3e53872d0319298359e50a94
Reviewed-on: https://go-review.googlesource.com/77251
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Reviewed-by: Joe Tsai <thebrokentoaster@gmail.com>
7 years agoruntime: call amd64 VDSO entry points on large stack
Ian Lance Taylor [Fri, 10 Nov 2017 17:48:39 +0000 (09:48 -0800)]
runtime: call amd64 VDSO entry points on large stack

If the Linux kernel was built with CONFIG_OPTIMIZE_INLINING=n and was
built with hardening options turned on, GCC will insert a stack probe
in the VDSO function that requires a full page of stack space.
The stack probe can corrupt memory if another thread is using it.
Avoid sporadic crashes by calling the VDSO on the g0 or gsignal stack.

While we're at it, align the stack as C code expects. We've been
getting away with a misaligned stack, but it's possible that the VDSO
code will change in the future to break that assumption.

Benchmarks show a 11% hit on time.Now, but it's only 6ns.

name                      old time/op  new time/op  delta
AfterFunc-12              1.66ms ± 0%  1.66ms ± 1%     ~     (p=0.905 n=9+10)
After-12                  1.90ms ± 6%  1.86ms ± 0%   -2.05%  (p=0.012 n=10+8)
Stop-12                    113µs ± 3%   115µs ± 2%   +1.60%  (p=0.017 n=9+10)
SimultaneousAfterFunc-12   145µs ± 1%   144µs ± 0%   -0.68%  (p=0.002 n=10+8)
StartStop-12              39.5µs ± 3%  40.4µs ± 5%   +2.19%  (p=0.023 n=10+10)
Reset-12                  10.2µs ± 0%  10.4µs ± 0%   +2.45%  (p=0.000 n=10+9)
Sleep-12                   190µs ± 1%   190µs ± 1%     ~     (p=0.971 n=10+10)
Ticker-12                 4.68ms ± 2%  4.64ms ± 2%   -0.83%  (p=0.043 n=9+10)
Now-12                    48.4ns ±11%  54.0ns ±11%  +11.42%  (p=0.017 n=10+10)
NowUnixNano-12            48.5ns ±13%  56.9ns ± 8%  +17.30%  (p=0.000 n=10+10)
Format-12                  489ns ±11%   504ns ± 6%     ~     (p=0.289 n=10+10)
FormatNow-12               436ns ±23%   480ns ±13%  +10.25%  (p=0.026 n=9+10)
MarshalJSON-12             656ns ±14%   587ns ±24%     ~     (p=0.063 n=10+10)
MarshalText-12             647ns ± 7%   638ns ± 9%     ~     (p=0.516 n=10+10)
Parse-12                   348ns ± 8%   328ns ± 9%   -5.66%  (p=0.030 n=10+10)
ParseDuration-12           136ns ± 9%   140ns ±11%     ~     (p=0.425 n=10+10)
Hour-12                   14.8ns ± 6%  15.6ns ±11%     ~     (p=0.085 n=10+10)
Second-12                 14.0ns ± 6%  14.3ns ±12%     ~     (p=0.443 n=10+10)
Year-12                   32.4ns ±11%  33.4ns ± 6%     ~     (p=0.492 n=10+10)
Day-12                    41.5ns ± 9%  42.3ns ±12%     ~     (p=0.239 n=10+10)

Fixes #20427

Change-Id: Ia395cbb863215f4499b8e7ef95f4b99f51090911
Reviewed-on: https://go-review.googlesource.com/76990
Reviewed-by: Austin Clements <austin@google.com>
7 years agonet/mail: treat comment in address as display name
Michael Stapelberg [Tue, 14 Nov 2017 12:46:03 +0000 (04:46 -0800)]
net/mail: treat comment in address as display name

I verified this change on a corpus of > 200 GB of emails since the mid-90s. With
this change, more addresses parse than before, and anything which parsed before
still parses.

In said corpus, I came across the edge case of comments preceding an
addr-spec (with angle brackets!), e.g. “(John Doe) <john@example.com>”, which
does not satisfy the conditions to be treated as a fallback, as per my reading
of RFC2822.

This change does not parse quoted-strings within comments (a corresponding TODO
is in the code), but I have not seen that in the wild.

Fixes #22670

Change-Id: I526fcf7c6390aa1c219fdec1852f26c514506f76
Reviewed-on: https://go-review.googlesource.com/77474
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

7 years agodoc: correct supported gdb version info
Hana Kim [Tue, 14 Nov 2017 18:17:44 +0000 (13:17 -0500)]
doc: correct supported gdb version info

GDB 7.5 recognizes DWARF4 by default.

GDB 7.5 release note does not explicitly mention DWARF4 support
but according to GCC 4.8 release note
 https://gcc.gnu.org/gcc-4.8/changes.html

 "DWARF4 is now the default when generating DWARF debug
 information. ...
 GDB 7.5, Valgrind 3.8.0 and elfutils 0.154 debug information
 consumers support DWARF4 by default."

Change-Id: I56b011c7c38fbc103bbd366ceaea3b709c66ab7f
Reviewed-on: https://go-review.googlesource.com/77570
Reviewed-by: Alberto Donizetti <alb.donizetti@gmail.com>
7 years agointernal/cpu: detect cpu features in internal/cpu package
Fangming.Fang [Wed, 8 Nov 2017 02:17:51 +0000 (02:17 +0000)]
internal/cpu: detect cpu features in internal/cpu package

change hash/crc32 package to use cpu package instead of using
runtime internal variables to check crc32 instruction

Change-Id: I8f88d2351bde8ed4e256f9adf822a08b9a00f532
Reviewed-on: https://go-review.googlesource.com/76490
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>

7 years agointernal/cpu: fix arm64 build
Brad Fitzpatrick [Tue, 14 Nov 2017 18:27:15 +0000 (18:27 +0000)]
internal/cpu: fix arm64 build

I submitted two CLs which broke the build. Add temporary placeholder
with false bools to fix the build and restore old behavior.

Updates golang/go#22718 (details of why it broke)

Change-Id: I1f30624e14f631a95f4eff5aae462f1091f723a2
Reviewed-on: https://go-review.googlesource.com/77590
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
7 years agocrypto/sha256: optimize arm64 sha256 implemention
fanzha02 [Fri, 28 Jul 2017 04:14:45 +0000 (04:14 +0000)]
crypto/sha256: optimize arm64 sha256 implemention

Optimize with ARMv8 SHA256 instructions.
Result (Cortex-A72)

name           old time/op    new time/op      delta
Hash8Bytes-64    1.54µs ± 1%      0.61µs ± 9%    -60.67%  (p=0.008 n=5+5)
Hash1K-64        17.2µs ± 1%       1.4µs ± 2%    -91.91%  (p=0.008 n=5+5)
Hash8K-64         127µs ± 0%         7µs ± 1%    -94.42%  (p=0.008 n=5+5)

name           old speed      new speed        delta
Hash8Bytes-64  5.20MB/s ± 1%   13.23MB/s ±10%   +154.58%  (p=0.008 n=5+5)
Hash1K-64      59.4MB/s ± 1%   735.1MB/s ± 2%  +1136.96%  (p=0.008 n=5+5)
Hash8K-64      64.5MB/s ± 0%  1156.3MB/s ± 1%  +1692.75%  (p=0.008 n=5+5)

Change-Id: I47eca6471b75cd07cb0c77477053a07d0de7494f
Reviewed-on: https://go-review.googlesource.com/61570
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

7 years agocrypto/sha1: optimize arm64 sha1 implemention
fanzha02 [Tue, 8 Aug 2017 06:13:43 +0000 (06:13 +0000)]
crypto/sha1: optimize arm64 sha1 implemention

Optimize with ARMv8 SHA1 instructions.
Results (Cortex-A72)

name             old time/op    new time/op     delta
Hash8Bytes-64      1.06µs ± 4%     0.56µs ± 4%   -47.19%  (p=0.008 n=5+5)
Hash320Bytes-64    3.92µs ± 1%     0.82µs ± 2%   -79.07%  (p=0.008 n=5+5)
Hash1K-64          10.2µs ± 2%      1.5µs ± 2%   -85.71%  (p=0.008 n=5+5)
Hash8K-64          73.9µs ± 1%      7.6µs ± 1%   -89.66%  (p=0.008 n=5+5)

name             old speed      new speed       delta
Hash8Bytes-64    7.55MB/s ± 4%  14.29MB/s ± 4%   +89.27%  (p=0.008 n=5+5)
Hash320Bytes-64  81.6MB/s ± 1%  390.0MB/s ± 2%  +377.64%  (p=0.008 n=5+5)
Hash1K-64         100MB/s ± 2%    701MB/s ± 2%  +599.65%  (p=0.008 n=5+5)
Hash8K-64         111MB/s ± 1%   1072MB/s ± 1%  +867.44%  (p=0.008 n=5+5)

Change-Id: I84397f980db9518f4150ac4c5ffa2c5a97a34444
Reviewed-on: https://go-review.googlesource.com/61550
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

7 years agomath/big: clarify comment on lehmerGCD overflow
Brian Kessler [Tue, 14 Nov 2017 06:05:45 +0000 (22:05 -0800)]
math/big: clarify comment on lehmerGCD overflow

A clarifying comment was added to indicate that overflow of a
single Word is not possible in the single digit calculation.
Lehmer's paper includes a proof of the bounds on the size of the
cosequences (u0, u1, u2, v0, v1, v2).

Change-Id: I98127a07aa8f8fe44814b74b2bc6ff720805194b
Reviewed-on: https://go-review.googlesource.com/77451
Reviewed-by: Robert Griesemer <gri@golang.org>
7 years agonet/http/httputil: allow ReverseProxy to call ModifyResponse on failed requests
Akhil Indurti [Tue, 14 Nov 2017 16:40:09 +0000 (11:40 -0500)]
net/http/httputil: allow ReverseProxy to call ModifyResponse on failed requests

Previously when RoundTrip returned a non-nil error, the proxy returned a
StatusBadGateway error, instead of first calling ModifyResponse. This
commit first calls ModifyResponse, whether or not the error returned
from RoundTrip is nil.

Also closes response body when ModifyResponse returns an error. See #22658.

Fixes #21255

Change-Id: I5b5bf23a69ae5608f87d4ece756a1b4985ccaa9c
Reviewed-on: https://go-review.googlesource.com/54030
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

7 years agonet/url: clarify QueryUnescape and PathUnescape doc
Alberto Donizetti [Fri, 10 Nov 2017 19:35:45 +0000 (20:35 +0100)]
net/url: clarify QueryUnescape and PathUnescape doc

In the doc for QueryUnescape and PathUnescape, clarify that by 0xAB we
means a substring with any two valid hexadecimal digits.

Fixes #18642

Change-Id: Ib65b130995ae5fcf07e25ee0fcc41fad520c5662
Reviewed-on: https://go-review.googlesource.com/77050
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
7 years agocmd/compile: always nil check before interface call
Keith Randall [Tue, 14 Nov 2017 03:03:31 +0000 (19:03 -0800)]
cmd/compile: always nil check before interface call

Fixes #22703

The fix was already done by Cherry for defer/go of an interface call (CL 23820).
We just need to do it everywhere.

Change-Id: I0115d22e443931fe1bcce44c93c4d0770b5fd268
Reviewed-on: https://go-review.googlesource.com/77450
Run-TryBot: Keith Randall <khr@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
7 years agobufio: Use maxConsecutiveEmptyReads instead of 100
Ryoichi KATO [Sun, 25 Jun 2017 06:39:29 +0000 (23:39 -0700)]
bufio: Use maxConsecutiveEmptyReads instead of 100

Use maxConsecutiveEmptyReads const instead of hardcoded
100 in scan.go too.

Change-Id: I993f353a3748f0d6bdefab38bf5cb224eea8a969
Reviewed-on: https://go-review.googlesource.com/46915
Reviewed-by: Rob Pike <r@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
7 years agoruntime: make TestWindowsStackMemory build even with CGO_ENABLED=0 set
Alex Brainman [Tue, 14 Nov 2017 01:51:58 +0000 (12:51 +1100)]
runtime: make TestWindowsStackMemory build even with CGO_ENABLED=0 set

Just copy some code to make TestWindowsStackMemory build
when CGO_ENABLED is set to 0.

Fixes #22680

Change-Id: I63f9b409a3a97b7718f5d37837ab706d8ed92e81
Reviewed-on: https://go-review.googlesource.com/77430
Reviewed-by: Chris Hines <chris.cs.guy@gmail.com>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

7 years agonet/http: add example to Server.Shutdown
Stephen McQuay (smcquay) [Sat, 15 Jul 2017 18:33:00 +0000 (12:33 -0600)]
net/http: add example to Server.Shutdown

Fixes #19579

Change-Id: Id99ca6de94d8d895dfaed1ed507e9d36c7f60670
Reviewed-on: https://go-review.googlesource.com/48869
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
7 years agodatabase/sql: do not leak the connectionResetter goroutine
Daniel Theophanes [Mon, 13 Nov 2017 22:25:19 +0000 (14:25 -0800)]
database/sql: do not leak the connectionResetter goroutine

Before terminating the connectionResetter goroutine the connection
pool processes all of the connections on the channel to unlock the
driverConn instances so everthing can shutdown cleanly. However
the channel was never closed so the goroutine hangs on the range.
Close the channel prior to ranging over it. Also prevent additional
connections from being sent to the resetter after the connection
pool has been closed.

Fixes #22699

Change-Id: I440d2b13cbedec2e04621557f5bd0b1526933dd7
Reviewed-on: https://go-review.googlesource.com/77390
Run-TryBot: Daniel Theophanes <kardianos@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
7 years agohash: document that the encoded state may contain input in plaintext
Joe Tsai [Mon, 13 Nov 2017 21:48:49 +0000 (13:48 -0800)]
hash: document that the encoded state may contain input in plaintext

The cryptographic checksums operate in blocks of 64 or 128 bytes,
which means that the last 128 bytes or so of the input may be encoded
in its original (plaintext) form as part of the state.
Document this so users do not falsely assume that the encoded state
carries no reversible information about the input.

Change-Id: I823dbb87867bf0a77aa20f6ed7a615dbedab3715
Reviewed-on: https://go-review.googlesource.com/77372
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
7 years agoruntime: don't elide wrapper functions that call panic or at TOS
Austin Clements [Thu, 9 Nov 2017 22:55:45 +0000 (17:55 -0500)]
runtime: don't elide wrapper functions that call panic or at TOS

CL 45412 started hiding autogenerated wrapper functions from call
stacks so that call stack semantics better matched language semantics.
This is based on the theory that the wrapper function will call the
"real" function and all the programmer knows about is the real
function.

However, this theory breaks down in two cases:

1. If the wrapper is at the top of the stack, then it didn't call
   anything. This can happen, for example, if the "stack" was actually
   synthesized by the user.

2. If the wrapper panics, for example by calling panicwrap or by
   dereferencing a nil pointer, then it didn't call the wrapped
   function and the user needs to see what panicked, even if we can't
   attribute it nicely.

This commit modifies the traceback logic to include the wrapper
function in both of these cases.

Fixes #22231.

Change-Id: I6e4339a652f73038bd8331884320f0b8edd86eb1
Reviewed-on: https://go-review.googlesource.com/76770
Run-TryBot: Austin Clements <austin@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
7 years agocrypto/elliptic: reduce allocations in s390x P256 code
Michael Munday [Tue, 26 Sep 2017 11:32:26 +0000 (07:32 -0400)]
crypto/elliptic: reduce allocations in s390x P256 code

The Go compiler assumes that pointers escape when passed into assembly
functions. To override this behavior we can annotate assembly functions
with go:noescape, telling the compiler that we know pointers do not
escape from it.

By annotating the assembly functions in the s390x P256 code in this way
we enable more variables to be allocated on the stack rather than
the heap, reducing the number of heap allocations required to execute
this code:

name        old alloc/op   new alloc/op   delta
SignP256      3.66kB ± 0%    2.64kB ± 0%  -27.95%  (p=0.008 n=5+5)
VerifyP256    4.46kB ± 0%    1.23kB ± 0%  -72.40%  (p=0.008 n=5+5)

name        old allocs/op  new allocs/op  delta
SignP256        40.0 ± 0%      31.0 ± 0%  -22.50%  (p=0.008 n=5+5)
VerifyP256      41.0 ± 0%      24.0 ± 0%  -41.46%  (p=0.008 n=5+5)

Change-Id: Id526c30c9b04b2ad79a55d76cab0e30cc8d60402
Reviewed-on: https://go-review.googlesource.com/66230
Run-TryBot: Michael Munday <mike.munday@ibm.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
7 years agocmd/compile: remove some more gotos in gc
Daniel Martí [Sun, 22 Oct 2017 16:18:51 +0000 (17:18 +0100)]
cmd/compile: remove some more gotos in gc

Split typecheckrange into two, separating the bigger chunk of code that
takes care of the range expression. It had to sometimes exit early,
which was done via a goto in the larger func. This lets us simplify many
declarations and the flow of the code. While at it, also replace the
toomany int with a bool.

In the case of walkselect, split it into two funcs too since using a
defer for all the trailing work would be a bit much. It also lets us
simplify the declarations and the flow of the code, since now
walkselectcases has a narrower scope and straightforward signature.

Also replace the gotos in typecheckaste with a lineno defer.

Passes toolstash -cmp on std cmd.

Change-Id: Iacfaa0a34c987c44f180a792c473558785cf6823
Reviewed-on: https://go-review.googlesource.com/72374
Run-TryBot: Daniel Martí <mvdan@mvdan.cc>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
7 years agodoc: mention the scratch repo
Kevin Burke [Wed, 19 Jul 2017 14:21:30 +0000 (07:21 -0700)]
doc: mention the scratch repo

If people are interested in contributing to Go, but not sure what
change they'd like to make just yet, we can point them to the scratch
repo, so they can go through the process of submitting and merging
something now, and make more useful contributions later.

My evidence that sending people to the scratch repo would encourage
future contributions is that a number of people who went through the
workshop at Gophercon have continued to send CL's after submitting to
the scratch repo, even though I doubt they planned to before going
through the workshop.

Change-Id: Ieb48415773c0ee7dc400f8bf6f57f752eca8eeb5
Reviewed-on: https://go-review.googlesource.com/49970
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
7 years agocmd/compile: add test for array decomposition
Keith Randall [Mon, 13 Nov 2017 18:52:35 +0000 (10:52 -0800)]
cmd/compile: add test for array decomposition

This test fails on 1.9.2, but is ok on tip.
CL 77331 has both the 1.9.2 fix and this test, and is on the 1.9 release branch.
This CL is just the test, and is on HEAD.  The buggy code doesn't exist on tip.

Update #22683

Change-Id: I04a24bd6c2d3068e18ca81da3347e2c1366f4447
Reviewed-on: https://go-review.googlesource.com/77332
Run-TryBot: Keith Randall <khr@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Reviewed-by: David Chase <drchase@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

7 years agotime: add example for Time.Unix
Adrian Hesketh [Fri, 1 Sep 2017 11:38:52 +0000 (12:38 +0100)]
time: add example for Time.Unix

Change-Id: Ie64eba5b57b609a343ddb381fe83c01f172c0bf4
Reviewed-on: https://go-review.googlesource.com/60890
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>

7 years agoencoding/json: always ignore embedded pointers to unexported struct types
Joe Tsai [Fri, 10 Nov 2017 04:14:47 +0000 (20:14 -0800)]
encoding/json: always ignore embedded pointers to unexported struct types

CL 60410 fixes a bug in reflect that allows assignments to an embedded
field of a pointer to an unexported struct type.
This breaks the json package because unmarshal is now unable to assign
a newly allocated struct to such fields.

In order to be consistent in the behavior for marshal and unmarshal,
this CL changes both marshal and unmarshal to always ignore
embedded pointers to unexported structs.

Fixes #21357

Change-Id: If62ea11155555e61115ebb9cfa5305caf101bde5
Reviewed-on: https://go-review.googlesource.com/76851
Run-TryBot: Joe Tsai <thebrokentoaster@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
7 years agocmd/vet: ignore unrecognized flags for fmt.Formatter
Chris Hines [Sun, 12 Nov 2017 05:31:52 +0000 (00:31 -0500)]
cmd/vet: ignore unrecognized flags for fmt.Formatter

Fixes #22608.

Change-Id: Id85eb86b0b262156646e55f102fe888b345b20cf
Reviewed-on: https://go-review.googlesource.com/77230
Run-TryBot: Russ Cox <rsc@golang.org>
Reviewed-by: Russ Cox <rsc@golang.org>
Reviewed-by: Rob Pike <r@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

7 years agodoc: add a link to the code of conduct
Leigh McCulloch [Mon, 13 Nov 2017 06:27:06 +0000 (06:27 +0000)]
doc: add a link to the code of conduct

GitHub has defined a set of standard meta files to include with projects
hosted on GitHub. According to the GitHub Insights Community page for
go project the only one missing is the code of conduct.

Go has a code of conduct on it's website and we should link to it in the
prescribed `.github/CODE_OF_CONDUCT.md` so that people can find it in
the standard location on GitHub. This would be consistent with the
contribution guidelines that are linked to in `.github/CONTRIBUTING.md`.

Ref: https://help.github.com/articles/adding-a-code-of-conduct-to-your-project/

Fixes #22685

Change-Id: Ie89aa39d3df741d7d6ed2dba5c8ba3d0e0dbf618
Reviewed-on: https://go-review.googlesource.com/77231
Reviewed-by: Gabriel Aszalos <gabriel.aszalos@gmail.com>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
7 years agocmd/compile: record original and absolute file names for line directives
griesemer [Fri, 10 Nov 2017 19:38:51 +0000 (11:38 -0800)]
cmd/compile: record original and absolute file names for line directives

Also, with this change, error locations don't print absolute positions
in [] brackets following positions relative to line directives. To get
the absolute positions as well, specify the -L flag.

Fixes #22660.

Change-Id: I9ecfa254f053defba9c802222874155fa12fee2c
Reviewed-on: https://go-review.googlesource.com/77090
Reviewed-by: David Crawshaw <crawshaw@golang.org>
Reviewed-by: Russ Cox <rsc@golang.org>
7 years agocmd/cover: modify source as text, not as AST
Russ Cox [Fri, 10 Nov 2017 04:39:12 +0000 (23:39 -0500)]
cmd/cover: modify source as text, not as AST

cmd/cover rewrites Go source code to add coverage annotations.
The approach to date has been to parse the code to AST, analyze it,
rewrite the AST, and print it back out. This approach fails to preserve
line numbers in the original code and has a very difficult time with
comments, because go/printer does as well.

This CL changes cmd/cover to decide what to modify based on the
AST but to apply the modifications as purely textual substitutions.
In this way, cmd/cover can be sure it never adds or removes a newline
character, nor a comment, so all line numbers and comments are
preserved.

This also allows us to emit a single //line comment at the beginning
of the translated file and have the compiler report errors with
correct line numbers in the original file.

Fixes #6329.
Fixes #15757.

Change-Id: Ia95f6f894bb498e80d1f91fde56cd4a8009d7f9b
Reviewed-on: https://go-review.googlesource.com/77150
Reviewed-by: Ian Lance Taylor <iant@golang.org>
7 years agocmd/go: more refined handling of cmd/compile magic environment variables
Russ Cox [Fri, 10 Nov 2017 19:50:03 +0000 (14:50 -0500)]
cmd/go: more refined handling of cmd/compile magic environment variables

Per discussion with David Chase, need to check GOSSAHASH$n
for increasing n until one is missing. Also if GSHS_LOGFILE is set,
the compiler writes to that file, so arrange never to cache in that case.

Change-Id: I3931b4e296251b99abab9bbbbbdcf94ae8c1e2a6
Reviewed-on: https://go-review.googlesource.com/77111
Reviewed-by: David Chase <drchase@google.com>
7 years agocmd/go: cache and replay command output during build
Russ Cox [Fri, 10 Nov 2017 19:44:31 +0000 (14:44 -0500)]
cmd/go: cache and replay command output during build

It's nice that

go build -gcflags=-m errors
go build -gcflags=-m errors

uses the cache for the second command.

Even nicer is to make the second command
print the same output as the first command.

Fixes #22587.

Change-Id: I64350839f01c86c9a095d9d22f6924cd7a0b9105
Reviewed-on: https://go-review.googlesource.com/77110
Reviewed-by: Ian Lance Taylor <iant@golang.org>
7 years agocmd/compile: use stringer on types.EType
Daniel Martí [Sun, 12 Nov 2017 16:55:11 +0000 (16:55 +0000)]
cmd/compile: use stringer on types.EType

Another one that is possible thanks to the new -trimprefix stringer
flag.

The only subtle difference is that, in the previous version, some values
such as TUNSAFEPTR were stringified as "TUNSAFEPTR" instead of
"UNSAFEPTR". The new String method is always consistent in removing the
"T" prefix.

Passes toolstash -cmp on std cmd.

Change-Id: I68407f391795403dfcbbfa68c813018c0235bbb5
Reviewed-on: https://go-review.googlesource.com/77250
Run-TryBot: Daniel Martí <mvdan@mvdan.cc>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Marvin Stenger <marvin.stenger94@gmail.com>
Reviewed-by: Robert Griesemer <gri@golang.org>
7 years agocmd/compile: replace classnames with Class.String
Daniel Martí [Fri, 10 Nov 2017 18:35:30 +0000 (18:35 +0000)]
cmd/compile: replace classnames with Class.String

Since the slice of names is almost exactly the same as what stringer is
already generating for us.

Passes toolstash -cmp on std cmd.

Change-Id: I3f1e95efc690c0108236689e721627f00f79a461
Reviewed-on: https://go-review.googlesource.com/77190
Run-TryBot: Daniel Martí <mvdan@mvdan.cc>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Dave Cheney <dave@cheney.net>
7 years agocrypto/x509: handle name constraints with cryptobyte
Adam Langley [Mon, 30 Oct 2017 02:20:33 +0000 (19:20 -0700)]
crypto/x509: handle name constraints with cryptobyte

This allows better precision and (the motivation) empty strings to
be handled correctly. With that in place tests for the behaviour of
empty name constraints can be added.

Also fixes a compatibility issue with NSS. See #22616.

Fixes #22616

Change-Id: I5139439bb58435d5f769828a4eebf8bed2d858e8
Reviewed-on: https://go-review.googlesource.com/74271
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
7 years agonet/http/httputil: make ReverseProxy close response body if ModifyResponse returns...
Edan B [Sat, 11 Nov 2017 08:10:14 +0000 (10:10 +0200)]
net/http/httputil: make ReverseProxy close response body if ModifyResponse returns an error

Fixes #22658

Change-Id: I00e2b007d77b6f54798f7755d0b08e4fea824392
Reviewed-on: https://go-review.googlesource.com/77170
Reviewed-by: Emmanuel Odeke <emm.odeke@gmail.com>
Run-TryBot: Emmanuel Odeke <emm.odeke@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

7 years agocmd/go: skip TestGoTestJSON on uniprocessor machines
David du Colombier [Fri, 10 Nov 2017 20:46:41 +0000 (21:46 +0100)]
cmd/go: skip TestGoTestJSON on uniprocessor machines

CL 76873 added TestGoTestJSON. However, this test
is only succeeding on SMP machines.

This change skips TestGoTestJSON on uniprocessor machines.

Fixes #22665.

Change-Id: I3989d3331fb71193a25a3f0bbb84ff3e1b730890
Reviewed-on: https://go-review.googlesource.com/77130
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>