]> Cypherpunks repositories - gostls13.git/log
gostls13.git
7 years agoruntime: forward crashing signals to late handlers
Elias Naur [Sat, 19 Aug 2017 14:59:19 +0000 (16:59 +0200)]
runtime: forward crashing signals to late handlers

CL 49590 made it possible for external signal handlers to catch
signals from a crashing Go process. This CL extends that support
to handlers registered after the Go runtime has initialized.

Updates #20392 (and possibly fix it).

Change-Id: I18eccd5e958a505f4d1782a7fc51c16bd3a4ff9c
Reviewed-on: https://go-review.googlesource.com/57291
Run-TryBot: Elias Naur <elias.naur@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
7 years agoruntime: fix memhash64 on big endian systems
Michael Munday [Tue, 29 Aug 2017 04:10:17 +0000 (05:10 +0100)]
runtime: fix memhash64 on big endian systems

Fixes #21677.

Change-Id: I869dee5f43df5d87d86922681726297e3024c562
Reviewed-on: https://go-review.googlesource.com/59810
Run-TryBot: Michael Munday <mike.munday@ibm.com>
Reviewed-by: Martin Möhrmann <moehrmann@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

7 years agocmd/link: make it compile again (fixes build)
Alex Brainman [Tue, 29 Aug 2017 02:21:15 +0000 (12:21 +1000)]
cmd/link: make it compile again (fixes build)

CL 59375 changed Reloc.Done to bool, but that change
got lost in pe.go while merging. Restore that change.

Change-Id: Ie5a89e85579cdc9282f504fefd56355cfeb49389
Reviewed-on: https://go-review.googlesource.com/59711
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Alex Brainman <alex.brainman@gmail.com>
7 years agocmd/vet: add test that copylock catches copying a sync.Map
Manish Goregaokar [Sat, 26 Aug 2017 06:10:50 +0000 (23:10 -0700)]
cmd/vet: add test that copylock catches copying a sync.Map

fixes #21635

Change-Id: I78716d47b6a930074e5394f7171d5f545dfba0bc
Reviewed-on: https://go-review.googlesource.com/59690
Reviewed-by: Bryan Mills <bcmills@google.com>
Reviewed-by: Rob Pike <r@golang.org>
Run-TryBot: Bryan Mills <bcmills@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

7 years agocmd/link: introduce and use peFile.writeOptionalHeader
Alex Brainman [Mon, 28 Aug 2017 03:52:11 +0000 (13:52 +1000)]
cmd/link: introduce and use peFile.writeOptionalHeader

Change-Id: I27b33f2425281bc1790528ae514d99a468ad7fce
Reviewed-on: https://go-review.googlesource.com/59429
Reviewed-by: Ian Lance Taylor <iant@golang.org>
7 years agocmd/link: introduce and use peFile.dataDirectory
Alex Brainman [Mon, 28 Aug 2017 03:49:15 +0000 (13:49 +1000)]
cmd/link: introduce and use peFile.dataDirectory

Change-Id: I61f64287a488ed82e71639540697b6ade5a0426a
Reviewed-on: https://go-review.googlesource.com/59428
Reviewed-by: Ian Lance Taylor <iant@golang.org>
7 years agocmd/link: introduce and use peFile.writeFileHeader
Alex Brainman [Mon, 28 Aug 2017 03:45:56 +0000 (13:45 +1000)]
cmd/link: introduce and use peFile.writeFileHeader

Change-Id: I1324b69ab9edb870589197fa601b5764634d42a4
Reviewed-on: https://go-review.googlesource.com/59427
Reviewed-by: Ian Lance Taylor <iant@golang.org>
7 years agocmd/link: introduce and use peFile.writeSymbolTableAndStringTable
Alex Brainman [Mon, 28 Aug 2017 03:42:06 +0000 (13:42 +1000)]
cmd/link: introduce and use peFile.writeSymbolTableAndStringTable

Change-Id: I506f5e146f3b5bf359d6932a85ac5572d3a3f103
Reviewed-on: https://go-review.googlesource.com/59426
Reviewed-by: Ian Lance Taylor <iant@golang.org>
7 years agocmd/link: introduce and use peFile.writeSymbols
Alex Brainman [Mon, 28 Aug 2017 03:39:19 +0000 (13:39 +1000)]
cmd/link: introduce and use peFile.writeSymbols

Change-Id: If629b19cebc6ae8dfbf603dcd7f2dd3d0046a935
Reviewed-on: https://go-review.googlesource.com/59425
Reviewed-by: Ian Lance Taylor <iant@golang.org>
7 years agocmd/link: simplify writePESymTableRecords
Alex Brainman [Mon, 28 Aug 2017 03:35:21 +0000 (13:35 +1000)]
cmd/link: simplify writePESymTableRecords

Change-Id: I2da982fdef826aaecb86431f7d80ffc0fb2c4337
Reviewed-on: https://go-review.googlesource.com/59424
Reviewed-by: Ian Lance Taylor <iant@golang.org>
7 years agocmd/link: introduce and use peFile.mapToPESection
Alex Brainman [Mon, 28 Aug 2017 03:32:19 +0000 (13:32 +1000)]
cmd/link: introduce and use peFile.mapToPESection

Change-Id: I598e9da5587908f39faa13c11d2b42054f5a632d
Reviewed-on: https://go-review.googlesource.com/59423
Reviewed-by: Ian Lance Taylor <iant@golang.org>
7 years agocmd/link: introduce and use peFile.writeSymbol
Alex Brainman [Mon, 28 Aug 2017 03:29:28 +0000 (13:29 +1000)]
cmd/link: introduce and use peFile.writeSymbol

Change-Id: I9fcae7bfd647b52d9b4f36f04710ba7921609c02
Reviewed-on: https://go-review.googlesource.com/59422
Reviewed-by: Ian Lance Taylor <iant@golang.org>
7 years agocmd/link: move perelocsect into peFile.emitRelocations (no code changed)
Alex Brainman [Mon, 28 Aug 2017 03:25:45 +0000 (13:25 +1000)]
cmd/link: move perelocsect into peFile.emitRelocations (no code changed)

Change-Id: I807412bd743d544195aaf9d720f81b128b4636b3
Reviewed-on: https://go-review.googlesource.com/59421
Reviewed-by: Ian Lance Taylor <iant@golang.org>
7 years agocmd/link: introduce peFile.ctorsSect and use it in peFile.emitRelocations
Alex Brainman [Mon, 28 Aug 2017 03:18:03 +0000 (13:18 +1000)]
cmd/link: introduce peFile.ctorsSect and use it in peFile.emitRelocations

Change-Id: I65058c150a19aa4e3083fd8754d08ea23721844b
Reviewed-on: https://go-review.googlesource.com/59420
Reviewed-by: Ian Lance Taylor <iant@golang.org>
7 years agocmd/link: introduce and use peFile.emitRelocations
Alex Brainman [Mon, 28 Aug 2017 03:14:43 +0000 (13:14 +1000)]
cmd/link: introduce and use peFile.emitRelocations

Change-Id: I2458d1391af063d99ba1e446bcf12f7c41ae5f6b
Reviewed-on: https://go-review.googlesource.com/59419
Reviewed-by: Ian Lance Taylor <iant@golang.org>
7 years agocmd/link: introduce and use peSection.emitRelocations
Alex Brainman [Mon, 28 Aug 2017 03:04:27 +0000 (13:04 +1000)]
cmd/link: introduce and use peSection.emitRelocations

Change-Id: I2bebee5566ee07786695f147c27661e69337a0f7
Reviewed-on: https://go-review.googlesource.com/59418
Reviewed-by: Ian Lance Taylor <iant@golang.org>
7 years agocmd/link: fix warning for buildmode=plugin on darwin/amd64
Hiroshi Ioka [Sun, 27 Aug 2017 10:09:18 +0000 (19:09 +0900)]
cmd/link: fix warning for buildmode=plugin on darwin/amd64

Without this CL, the system linker complains about absolute addressing
in type..eqfunc.*.

Updates #18190

Change-Id: I68db37a7f4c96b16a9c13baffc0f043a3048df6d
Reviewed-on: https://go-review.googlesource.com/59373
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

7 years agodebug/macho: add missing file flags
Kunpei Sakai [Mon, 28 Aug 2017 06:28:12 +0000 (15:28 +0900)]
debug/macho: add missing file flags

Fixes #21414

Change-Id: I8b25d416e6ead087b4a9965e8fa1a7f59ff07291
Reviewed-on: https://go-review.googlesource.com/59290
Reviewed-by: Hiroshi Ioka <hirochachacha@gmail.com>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

7 years agocmd/cgo: avoid using common names for sniffing
Hiroshi Ioka [Mon, 28 Aug 2017 22:39:57 +0000 (07:39 +0900)]
cmd/cgo: avoid using common names for sniffing

Current code uses names like "x" and "s" which can conflict with user's
code easily. Use cryptographic names.

Fixes #21668

Change-Id: Ib6d3d6327aa5b92d95c71503d42e3a79d96c8e15
Reviewed-on: https://go-review.googlesource.com/59710
Reviewed-by: Ian Lance Taylor <iant@golang.org>
7 years agocmd/compile: merge tempnamel into tempAt
Matthew Dempsky [Fri, 25 Aug 2017 22:24:14 +0000 (15:24 -0700)]
cmd/compile: merge tempnamel into tempAt

Passes toolstash-check.

Change-Id: I01ed1c04be5a23756742d461f13f1e587ea7ecb8
Reviewed-on: https://go-review.googlesource.com/59610
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
7 years agocmd/compile: add Type.NumResults and friends
Josh Bleecher Snyder [Tue, 2 May 2017 16:16:22 +0000 (09:16 -0700)]
cmd/compile: add Type.NumResults and friends

Passes toolstash-check.

Change-Id: Id62bacff13fbd30de62b925d97a4e7bee1c66120
Reviewed-on: https://go-review.googlesource.com/59331
Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
7 years agocmd/doc: search for packages in the two-arg case
Rob Pike [Mon, 28 Aug 2017 05:33:11 +0000 (15:33 +1000)]
cmd/doc: search for packages in the two-arg case

When given one argument, as in

go doc binary.BigEndian

doc would search for the package, but when given two, as in

go doc binary BigEndian

it would not. Fix the inconsistency.

Fixes #18697
Fixes #18664

Change-Id: Ib59dc483e8d4f91e6061c77a5ec24d0a50e115f0
Reviewed-on: https://go-review.googlesource.com/59413
Reviewed-by: Aliaksandr Valialkin <valyala@gmail.com>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
7 years agocmd/compile: avoid generating large offsets
Keith Randall [Mon, 28 Aug 2017 19:57:52 +0000 (12:57 -0700)]
cmd/compile: avoid generating large offsets

The assembler barfs on large offsets. Make sure that all the
instructions that need to have their offsets in an int32
  1) check on any rule that computes offsets for such instructions
  2) change their aux fields so the check builder checks it.

The assembler also silently misassembled offsets between 1<<31
and 1<<32. Add a check in the assembler to barf on those as well.

Fixes #21655

Change-Id: Iebf24bf10f9f37b3ea819ceb7d588251c0f46d7d
Reviewed-on: https://go-review.googlesource.com/59630
Run-TryBot: Keith Randall <khr@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: David Chase <drchase@google.com>
7 years agocmd/trace: add -d that prints parsed traces
Hana Kim [Fri, 25 Aug 2017 16:56:29 +0000 (12:56 -0400)]
cmd/trace: add -d that prints parsed traces

This is useful when debugging the tool.

Some tweaks on logging: log the webserver address, log.Print instead
of log.Printf when possible.

Change-Id: Iaf71b6523b40dc13795511784d48eacf0f5a396a
Reviewed-on: https://go-review.googlesource.com/59570
Reviewed-by: Heschi Kreinick <heschi@google.com>
Reviewed-by: Dmitry Vyukov <dvyukov@google.com>
7 years agoruntime: specialize memhash32 and memhash64
Martin Möhrmann [Sun, 27 Aug 2017 12:05:11 +0000 (14:05 +0200)]
runtime: specialize memhash32 and memhash64

AMD64 with AES support disabled:
name                old time/op    new time/op    delta
MapPopulate/1         78.0ns ± 1%    75.5ns ± 1%   -3.17%  (p=0.000 n=10+9)
MapPopulate/10         764ns ± 2%     673ns ± 2%  -11.91%  (p=0.000 n=10+10)
MapPopulate/100       9.52µs ± 1%    8.54µs ± 1%  -10.37%  (p=0.000 n=8+10)
MapPopulate/1000       116µs ± 2%     103µs ± 1%  -10.40%  (p=0.000 n=10+8)
MapPopulate/10000     1.01ms ± 1%    0.90ms ± 1%  -10.70%  (p=0.000 n=10+10)
MapPopulate/100000    9.81ms ± 1%    8.67ms ± 2%  -11.54%  (p=0.000 n=10+10)

386 with AES support disabled:
name                old time/op    new time/op    delta
MapPopulate/1         95.3ns ± 1%    90.6ns ± 1%  -4.95%  (p=0.000 n=10+9)
MapPopulate/10         983ns ± 2%     912ns ± 1%  -7.18%  (p=0.000 n=10+10)
MapPopulate/100       11.9µs ± 2%    11.2µs ± 1%  -6.01%  (p=0.000 n=10+10)
MapPopulate/1000       140µs ± 1%     131µs ± 1%  -6.19%  (p=0.000 n=10+10)
MapPopulate/10000     1.26ms ± 2%    1.18ms ± 1%  -5.93%  (p=0.000 n=9+10)
MapPopulate/100000    12.1ms ± 2%    11.4ms ± 1%  -5.48%  (p=0.000 n=10+10)

Fixes #21539

Change-Id: Ice128c947c9a6a294800d6a5250d82045eb70b55
Reviewed-on: https://go-review.googlesource.com/59352
Run-TryBot: Martin Möhrmann <moehrmann@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
7 years agodoc/1.9: add mention of net/http.LocalAddrContextKey
Tom Bergan [Mon, 28 Aug 2017 18:09:37 +0000 (11:09 -0700)]
doc/1.9: add mention of net/http.LocalAddrContextKey

Fixes #21603

Change-Id: I42fb7ea2dd7f6d6a201171055beaeda68c26b823
Reviewed-on: https://go-review.googlesource.com/59530
Reviewed-by: Ian Lance Taylor <iant@golang.org>
7 years agoruntime: only clear key string's pointer in mapdelete_faststr
Josh Bleecher Snyder [Sat, 26 Aug 2017 13:55:34 +0000 (06:55 -0700)]
runtime: only clear key string's pointer in mapdelete_faststr

Change-Id: I0360d294868ec4423e4ae036009fac4e72425c9c
Reviewed-on: https://go-review.googlesource.com/59152
Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
7 years agoruntime: remove t.indirectvalue handling in fast evacuation routines
Josh Bleecher Snyder [Tue, 22 Aug 2017 13:27:48 +0000 (06:27 -0700)]
runtime: remove t.indirectvalue handling in fast evacuation routines

Maps with indirect values use the generic map routines.

Change-Id: Ib211e93f1dacefb988ba3d279f92a13065168079
Reviewed-on: https://go-review.googlesource.com/59135
Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
Reviewed-by: Martin Möhrmann <moehrmann@google.com>
7 years agoruntime: speed up key copying in specialized evacuate routines
Josh Bleecher Snyder [Sat, 19 Aug 2017 14:30:12 +0000 (07:30 -0700)]
runtime: speed up key copying in specialized evacuate routines

Similar to CL 59110.

Change-Id: Ia2858541c86a44b105eacbca9a46b1044632c5ca
Reviewed-on: https://go-review.googlesource.com/59134
Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
7 years agoruntime: remove handling of indirect key types in evacuate_fastX
Josh Bleecher Snyder [Sat, 19 Aug 2017 04:19:12 +0000 (21:19 -0700)]
runtime: remove handling of indirect key types in evacuate_fastX

None of the mapfast key types are indirect.

Change-Id: I1fb2682257567ee69504082a6cdad63c99916671
Reviewed-on: https://go-review.googlesource.com/59133
Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
Reviewed-by: Martin Möhrmann <moehrmann@google.com>
7 years agoruntime: remove handling of non-reflexive key types in evacuate_fastX
Josh Bleecher Snyder [Sat, 19 Aug 2017 04:17:40 +0000 (21:17 -0700)]
runtime: remove handling of non-reflexive key types in evacuate_fastX

All of the mapfast key types are reflexive.

Change-Id: I8595aed2a9d945cda1b5d08e2067dce0f1c0d585
Reviewed-on: https://go-review.googlesource.com/59132
Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
Reviewed-by: Martin Möhrmann <moehrmann@google.com>
7 years agoruntime: replace t.keysize with fixed key size in evacuate_fastX
Josh Bleecher Snyder [Sat, 19 Aug 2017 04:15:44 +0000 (21:15 -0700)]
runtime: replace t.keysize with fixed key size in evacuate_fastX

Change-Id: I89c3c3b21d7a4acbc49b14a52ac8d9a5861c0c39
Reviewed-on: https://go-review.googlesource.com/59131
Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
Reviewed-by: Martin Möhrmann <moehrmann@google.com>
7 years agoruntime: add specialized copies of growWork and evacuate
Josh Bleecher Snyder [Sat, 19 Aug 2017 04:13:09 +0000 (21:13 -0700)]
runtime: add specialized copies of growWork and evacuate

The newly added routines are exact copies of the generic routines,
except for the function names and that growWork_fastX calls evacuate_fastX.

Actual optimization will happen in subsequent CLs.
This is intended to ease reviewing.

Change-Id: I52ef7dd40b2bdfc9cba2496544c0604e6e71cf7f
Reviewed-on: https://go-review.googlesource.com/59130
Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
7 years agonet/http: Set a timeout on Request.Context when using TimeoutHandler
Michael Fraenkel [Thu, 22 Jun 2017 01:58:42 +0000 (21:58 -0400)]
net/http: Set a timeout on Request.Context when using TimeoutHandler

In TimeoutHandler, use a request whose context has been configured with
the handler's timeout

Fixes #20712

Change-Id: Ie670148f85fdad46841ff29232042309e15665ae
Reviewed-on: https://go-review.googlesource.com/46412
Run-TryBot: Tom Bergan <tombergan@google.com>
Reviewed-by: Tom Bergan <tombergan@google.com>
7 years agocmd/compile: use keyed struct for asm tests
Keith Randall [Mon, 28 Aug 2017 16:55:18 +0000 (09:55 -0700)]
cmd/compile: use keyed struct for asm tests

Just to make it clearer which regexps are positive and which
regexps are negative.

Change-Id: Ia190e89be28048fcae2491506f552afad90a5f85
Reviewed-on: https://go-review.googlesource.com/59490
Run-TryBot: Keith Randall <khr@golang.org>
Reviewed-by: Martin Möhrmann <moehrmann@google.com>
Reviewed-by: Ilya Tocar <ilya.tocar@intel.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

7 years agocmd/compile/internal/ssa: Mark ADD[Q|L]const as rematerializeable
Ilya Tocar [Wed, 9 Aug 2017 18:43:34 +0000 (13:43 -0500)]
cmd/compile/internal/ssa: Mark ADD[Q|L]const as rematerializeable

We can rematerialize only ops that have SP or SB as their only argument.
There are some ADDQconst(SP) that can be rematerialized, but are spilled/filled instead,
so mark addconst as rematerializeable. This shaves ~1kb from go tool.

Change-Id: Ib4cf4fe5f2ec9d3d7e5f0f77f1193eba66ca2f08
Reviewed-on: https://go-review.googlesource.com/54393
Run-TryBot: Ilya Tocar <ilya.tocar@intel.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
7 years agosort: add examples for IntsAreSorted, Float64s and Float64sAreSorted
Karsten Köhler [Mon, 21 Aug 2017 19:13:53 +0000 (21:13 +0200)]
sort: add examples for IntsAreSorted, Float64s and Float64sAreSorted

Change-Id: Ib4883470bd2271e546daea3156d4a48dd873aaa3
Reviewed-on: https://go-review.googlesource.com/57670
Reviewed-by: Joe Tsai <thebrokentoaster@gmail.com>
Run-TryBot: Joe Tsai <thebrokentoaster@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

7 years agomisc/ios,src/iostest.bash: support GOIOS_DEVICE_ID
Elias Naur [Wed, 23 Aug 2017 11:35:24 +0000 (13:35 +0200)]
misc/ios,src/iostest.bash: support GOIOS_DEVICE_ID

When running multiple iOS builds on the same host, GOIOS_DEVICE_ID
is used to distinguish the devices. To improve support,

- Only restart the particular device when invoking iostest.bash
with the -restart flag.
- Make the exec wrapper lock file per-device.

For the iOS builder.

Change-Id: Id6f222981f25036399a43c3202a393dba89d87cb
Reviewed-on: https://go-review.googlesource.com/57970
Run-TryBot: Elias Naur <elias.naur@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
7 years agocmd/compile: don't use MOVOstore for move on plan9/amd64
David du Colombier [Sat, 26 Aug 2017 21:05:36 +0000 (23:05 +0200)]
cmd/compile: don't use MOVOstore for move on plan9/amd64

The SSA compiler currently generates MOVOstore instructions
to optimize 16 bytes moves on AMD64 architecture.

However, we can't use the MOVOstore instruction on Plan 9,
because floating point operations are not allowed in the
note handler.

We rely on the useSSE flag to disable the use of the
MOVOstore instruction on Plan 9 and replace it by two
MOVQstore instructions.

Fixes #21625

Change-Id: Idfefcceadccafe1752b059b5fe113ce566c0e71c
Reviewed-on: https://go-review.googlesource.com/59171
Run-TryBot: David du Colombier <0intro@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ilya Tocar <ilya.tocar@intel.com>
7 years agonet/http: mention RegisterOnShutdown in Server.Shutdown docs
Artyom Pervukhin [Sun, 27 Aug 2017 08:30:19 +0000 (11:30 +0300)]
net/http: mention RegisterOnShutdown in Server.Shutdown docs

Closes #21637

Change-Id: Icc3528572ea2a25e62757cc8fbbb9c3fa96a78b2
Reviewed-on: https://go-review.googlesource.com/59350
Reviewed-by: Tom Bergan <tombergan@google.com>
7 years agocmd/compile: optimize ARM with more efficient MOVB/MOVBU/MOVH/MOVHU
Ben Shi [Thu, 24 Aug 2017 10:51:34 +0000 (10:51 +0000)]
cmd/compile: optimize ARM with more efficient MOVB/MOVBU/MOVH/MOVHU

Like the indexed MOVW (MOVWloadidx/MOVWstoreidx) used in current
ARM backend, the indexed MOVB/MOVBU/MOVH/MOVHU can also be used to
generate further optimized ARM code.

My patch implements this optimization. Here are some contrast test
results against the original go compiler.

1. The total size of all .a files in pkg/ shrinks by 0.03%.

2. The compilecmp benchmark shows a little decline.
name        old time/op       new time/op       delta
Template          2.35s ± 1%        2.37s ± 3%  +0.94%  (p=0.006 n=19+19)
Unicode           1.33s ± 3%        1.33s ± 2%    ~     (p=0.158 n=20+18)
GoTypes           7.86s ± 2%        7.84s ± 1%    ~     (p=0.284 n=19+18)
Compiler          37.5s ± 1%        37.7s ± 2%    ~     (p=0.101 n=20+19)
SSA               83.4s ± 2%        83.6s ± 2%    ~     (p=0.231 n=20+20)
Flate             1.46s ± 2%        1.45s ± 1%    ~     (p=0.097 n=20+17)
GoParser          1.86s ± 2%        1.86s ± 4%    ~     (p=0.738 n=20+20)
Reflect           5.10s ± 1%        5.11s ± 1%    ~     (p=0.290 n=20+18)
Tar               1.78s ± 2%        1.77s ± 2%    ~     (p=0.166 n=19+20)
XML               2.61s ± 2%        2.61s ± 2%    ~     (p=0.665 n=19+19)
[Geo mean]        4.67s             4.68s       +0.16%

name        old user-time/op  new user-time/op  delta
Template          2.79s ± 3%        2.80s ± 2%    ~     (p=0.662 n=20+20)
Unicode           1.62s ± 3%        1.64s ± 4%    ~     (p=0.252 n=20+20)
GoTypes           9.58s ± 2%        9.62s ± 2%    ~     (p=0.250 n=20+20)
Compiler          46.2s ± 1%        46.2s ± 1%    ~     (p=0.602 n=20+19)
SSA                108s ± 1%         108s ± 2%    ~     (p=0.242 n=18+20)
Flate             1.69s ± 3%        1.69s ± 4%    ~     (p=0.470 n=20+20)
GoParser          2.16s ± 3%        2.20s ± 4%  +1.70%  (p=0.005 n=19+20)
Reflect           6.02s ± 2%        6.02s ± 2%    ~     (p=0.700 n=20+17)
Tar               2.11s ± 2%        2.11s ± 3%    ~     (p=0.480 n=18+20)
XML               3.07s ± 2%        3.11s ± 4%  +1.50%  (p=0.043 n=20+20)
[Geo mean]        5.61s             5.64s       +0.55%

name        old text-bytes    new text-bytes    delta
HelloSize         586kB ± 0%        586kB ± 0%    ~     (all equal)

name        old data-bytes    new data-bytes    delta
HelloSize        5.46kB ± 0%       5.46kB ± 0%    ~     (all equal)

name        old bss-bytes     new bss-bytes     delta
HelloSize        72.9kB ± 0%       72.9kB ± 0%    ~     (all equal)

name        old exe-bytes     new exe-bytes     delta
HelloSize        1.03MB ± 0%       1.03MB ± 0%    ~     (all equal)

3. The go1 benchmark shows improvement totally, and even more than 10%
improvement in the test case Revcomp.
name                     old time/op    new time/op    delta
BinaryTree17-4              42.0s ± 1%     41.5s ± 1%   -1.32%  (p=0.000 n=39+40)
Fannkuch11-4                24.1s ± 1%     23.6s ± 0%   -2.38%  (p=0.000 n=40+40)
FmtFprintfEmpty-4           843ns ± 0%     839ns ± 1%   -0.46%  (p=0.000 n=33+40)
FmtFprintfString-4         1.44µs ± 1%    1.37µs ± 1%   -5.48%  (p=0.000 n=40+35)
FmtFprintfInt-4            1.44µs ± 1%    1.41µs ± 2%   -1.50%  (p=0.000 n=40+40)
FmtFprintfIntInt-4         2.07µs ± 1%    2.06µs ± 0%   -0.78%  (p=0.000 n=40+40)
FmtFprintfPrefixedInt-4    2.50µs ± 1%    2.33µs ± 1%   -6.85%  (p=0.000 n=40+40)
FmtFprintfFloat-4          4.36µs ± 1%    4.34µs ± 0%   -0.39%  (p=0.017 n=40+40)
FmtManyArgs-4              8.11µs ± 0%    8.00µs ± 0%   -1.37%  (p=0.000 n=40+40)
GobDecode-4                 105ms ± 2%     103ms ± 2%   -2.17%  (p=0.000 n=39+39)
GobEncode-4                90.1ms ± 2%    88.6ms ± 1%   -1.67%  (p=0.000 n=40+39)
Gzip-4                      4.18s ± 1%     4.09s ± 1%   -2.03%  (p=0.000 n=40+40)
Gunzip-4                    608ms ± 1%     603ms ± 1%   -0.86%  (p=0.000 n=40+34)
HTTPClientServer-4          674µs ± 3%     661µs ± 2%   -1.82%  (p=0.000 n=40+39)
JSONEncode-4                256ms ± 1%     243ms ± 0%   -5.11%  (p=0.000 n=39+31)
JSONDecode-4                915ms ± 1%     904ms ± 1%   -1.18%  (p=0.000 n=40+36)
Mandelbrot200-4            49.2ms ± 0%    49.3ms ± 0%     ~     (p=0.254 n=34+40)
GoParse-4                  46.9ms ± 2%    46.9ms ± 1%     ~     (p=0.737 n=40+39)
RegexpMatchEasy0_32-4      1.28µs ± 1%    1.27µs ± 1%   -0.71%  (p=0.000 n=40+40)
RegexpMatchEasy0_1K-4      7.86µs ± 4%    7.67µs ± 4%   -2.46%  (p=0.000 n=38+40)
RegexpMatchEasy1_32-4      1.28µs ± 1%    1.28µs ± 1%   -0.54%  (p=0.000 n=40+40)
RegexpMatchEasy1_1K-4      10.4µs ± 2%    10.3µs ± 2%   -0.88%  (p=0.003 n=40+39)
RegexpMatchMedium_32-4     2.05µs ± 0%    2.04µs ± 0%   -0.34%  (p=0.000 n=40+33)
RegexpMatchMedium_1K-4      541µs ± 1%     535µs ± 1%   -1.02%  (p=0.000 n=40+38)
RegexpMatchHard_32-4       29.3µs ± 1%    29.1µs ± 1%   -0.51%  (p=0.000 n=40+40)
RegexpMatchHard_1K-4        881µs ± 1%     871µs ± 1%   -1.15%  (p=0.000 n=40+40)
Revcomp-4                  81.7ms ± 2%    67.5ms ± 2%  -17.37%  (p=0.000 n=39+39)
Template-4                  1.05s ± 1%     1.08s ± 2%   +3.67%  (p=0.000 n=40+40)
TimeParse-4                7.24µs ± 1%    7.09µs ± 1%   -2.13%  (p=0.000 n=40+40)
TimeFormat-4               13.2µs ± 1%    13.1µs ± 0%   -0.31%  (p=0.007 n=40+31)
[Geo mean]                  733µs          718µs        -2.03%

name                     old speed      new speed      delta
GobDecode-4              7.28MB/s ± 2%  7.44MB/s ± 2%   +2.23%  (p=0.000 n=39+39)
GobEncode-4              8.52MB/s ± 2%  8.67MB/s ± 1%   +1.70%  (p=0.000 n=40+39)
Gzip-4                   4.65MB/s ± 1%  4.74MB/s ± 1%   +1.94%  (p=0.000 n=37+40)
Gunzip-4                 31.9MB/s ± 1%  32.2MB/s ± 1%   +0.90%  (p=0.000 n=40+36)
JSONEncode-4             7.57MB/s ± 1%  7.98MB/s ± 0%   +5.41%  (p=0.000 n=40+31)
JSONDecode-4             2.12MB/s ± 1%  2.15MB/s ± 1%   +1.23%  (p=0.000 n=40+40)
GoParse-4                1.23MB/s ± 1%  1.23MB/s ± 1%     ~     (p=0.769 n=39+40)
RegexpMatchEasy0_32-4    25.0MB/s ± 1%  25.2MB/s ± 1%   +0.71%  (p=0.000 n=40+40)
RegexpMatchEasy0_1K-4     130MB/s ± 5%   134MB/s ± 4%   +2.53%  (p=0.000 n=38+40)
RegexpMatchEasy1_32-4    24.9MB/s ± 1%  25.1MB/s ± 1%   +0.55%  (p=0.000 n=40+40)
RegexpMatchEasy1_1K-4    98.5MB/s ± 2%  99.4MB/s ± 2%   +0.88%  (p=0.003 n=40+39)
RegexpMatchMedium_32-4    490kB/s ± 0%   490kB/s ± 0%     ~     (all equal)
RegexpMatchMedium_1K-4   1.89MB/s ± 1%  1.91MB/s ± 1%   +1.02%  (p=0.000 n=40+38)
RegexpMatchHard_32-4     1.10MB/s ± 1%  1.10MB/s ± 0%   +0.41%  (p=0.000 n=40+33)
RegexpMatchHard_1K-4     1.16MB/s ± 1%  1.17MB/s ± 1%   +1.21%  (p=0.000 n=40+40)
Revcomp-4                31.1MB/s ± 2%  37.6MB/s ± 2%  +21.03%  (p=0.000 n=39+39)
Template-4               1.86MB/s ± 1%  1.79MB/s ± 1%   -3.51%  (p=0.000 n=40+38)
[Geo mean]               6.66MB/s       6.80MB/s        +2.13%

fixes #21492

Change-Id: Ia26e7ca393f0a5f31de240e8ff9a220453ca7e0d
Reviewed-on: https://go-review.googlesource.com/58450
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

7 years agoregexp: examples for Regexp.Expand and Regexp.ExpandString functions
Matej Baćo [Mon, 28 Aug 2017 10:41:27 +0000 (12:41 +0200)]
regexp: examples for Regexp.Expand and Regexp.ExpandString functions

Current documentation lacks simple examples for functions Regexp.Expand
and Regexp.ExpandString whose usage is unclear from description alone.
This commit adds examples that demonstrate usage in practical way.

Fixes #21649

Change-Id: I7b2c06c8ab747f69a6578f0595bf0f3c742ac479
Reviewed-on: https://go-review.googlesource.com/59470
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

7 years agocmd/internal/objfile: add arm64 disassembler support
Wei Xiao [Thu, 24 Aug 2017 10:24:36 +0000 (18:24 +0800)]
cmd/internal/objfile: add arm64 disassembler support

Fixes #19157

Change-Id: Ieea286e8dc03929c3645f3113c33df569f8e26f3
Reviewed-on: https://go-review.googlesource.com/58930
Reviewed-by: Cherry Zhang <cherryyz@google.com>
7 years agoall: remove some unused result params
Daniel Martí [Tue, 15 Aug 2017 20:15:26 +0000 (21:15 +0100)]
all: remove some unused result params

Most of these are return values that were part of a receiving parameter,
so they're still accessible.

A few others are not, but those have never had a use.

Found with github.com/mvdan/unparam, after Kevin Burke's suggestion that
the tool should also warn about unused result parameters.

Change-Id: Id8b5ed89912a99db22027703a88bd94d0b292b8b
Reviewed-on: https://go-review.googlesource.com/55910
Run-TryBot: Daniel Martí <mvdan@mvdan.cc>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
7 years agocmd/link: refactor *reloc
Hiroshi Ioka [Sun, 27 Aug 2017 13:00:00 +0000 (22:00 +0900)]
cmd/link: refactor *reloc

* use bool instead of int if it's adequate.
* remove blank lines.

Change-Id: Ic4a5644a33ed9fc7ce388ef8ba15f1732446fcfc
Reviewed-on: https://go-review.googlesource.com/59375
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
7 years agocmd/compile: use raceX instead of raceXrange for types without subcomponents
Josh Bleecher Snyder [Tue, 2 May 2017 18:15:41 +0000 (11:15 -0700)]
cmd/compile: use raceX instead of raceXrange for types without subcomponents

Change-Id: I9882488e69565dc9da6814fefbdba3621daf74fe
Reviewed-on: https://go-review.googlesource.com/59332
Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Dmitry Vyukov <dvyukov@google.com>
Reviewed-by: Marvin Stenger <marvin.stenger94@gmail.com>
7 years agocrypto/rsa: fix URL for the PKCS #1 v2.2 document in pss.go
Ulrich Kunitz [Sun, 27 Aug 2017 09:27:45 +0000 (11:27 +0200)]
crypto/rsa: fix URL for the PKCS #1 v2.2 document in pss.go

The PKCS #1 v2.2 document has been moved to an EMC
website with a new URL. This CL updates the reference to the document to
the new URL.

The new URL is referenced under https://en.wikipedia.org/wiki/PKCS_1

Fixes #21642

Change-Id: Ib8738b0c4c3bb9ec427bebea20c4aacd607ba0db
Reviewed-on: https://go-review.googlesource.com/59351
Reviewed-by: Emmanuel Odeke <emm.odeke@gmail.com>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
7 years agofmt: document verbs %b %d %o %x %X for printing pointers
Rob Pike [Mon, 28 Aug 2017 04:35:33 +0000 (14:35 +1000)]
fmt: document verbs %b %d %o %x %X for printing pointers

This has been supported since Go 1 and there's even a test for it.
The documentation was missing.

Fixes #21409.

Change-Id: I5813488f6a98c1b4506c239e968d43344b91be12
Reviewed-on: https://go-review.googlesource.com/59412
Reviewed-by: Ian Lance Taylor <iant@golang.org>
7 years agocmd/go/internal/load: fix IsMetaPackage function name in doc
Dmitri Shuralyov [Mon, 28 Aug 2017 03:17:12 +0000 (23:17 -0400)]
cmd/go/internal/load: fix IsMetaPackage function name in doc

The IsMetaPackage function was made exported when it was moved from
cmd/go to cmd/go/internal/load in CL 36196. Its documentation wasn't
updated accordingly. This change fixes that, resolving a golint issue.

Updates #18653.

Change-Id: Icf89461000754d0f09e6617b11c838e4c050d5a6
Reviewed-on: https://go-review.googlesource.com/59430
Reviewed-by: Dave Cheney <dave@cheney.net>
7 years agocmd/link: avoid leaking file in ldshlibsyms
Hiroshi Ioka [Tue, 22 Aug 2017 10:50:43 +0000 (19:50 +0900)]
cmd/link: avoid leaking file in ldshlibsyms

Change-Id: If27f7fbf94ede6f9a57f2520aaf75e6506e0b3f8
Reviewed-on: https://go-review.googlesource.com/59374
Run-TryBot: Dave Cheney <dave@cheney.net>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Dave Cheney <dave@cheney.net>
7 years agocmd/compile: don't generate race calls for zero-sized values
Josh Bleecher Snyder [Tue, 2 May 2017 18:02:53 +0000 (11:02 -0700)]
cmd/compile: don't generate race calls for zero-sized values

One example of a heavily-used zero-size value is encoding/binary.BigEndian.

Change-Id: I8e873c447e154ab2ca61b7315df774693891270c
Reviewed-on: https://go-review.googlesource.com/59330
Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Dmitry Vyukov <dvyukov@google.com>
7 years agoall: remove some double spaces from comments
Daniel Martí [Sat, 19 Aug 2017 20:33:51 +0000 (22:33 +0200)]
all: remove some double spaces from comments

Went mainly for the ones that make no sense, such as the ones
mid-sentence or after commas.

Change-Id: Ie245d2c19cc7428a06295635cf6a9482ade25ff0
Reviewed-on: https://go-review.googlesource.com/57293
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Daniel Martí <mvdan@mvdan.cc>
TryBot-Result: Gobot Gobot <gobot@golang.org>

7 years agoarchive/tar: improve package documentation
Joe Tsai [Fri, 25 Aug 2017 22:34:35 +0000 (15:34 -0700)]
archive/tar: improve package documentation

Many aspects of the package is woefully undocumented.
With the recent flurry of improvements, the package is now at feature
parity with the GNU and TAR tools. Thoroughly all of the public API
and perform some minor stylistic cleanup in some code segments.

Change-Id: Ic892fd72c587f30dfe91d1b25b88c9c8048cc389
Reviewed-on: https://go-review.googlesource.com/59210
Run-TryBot: Joe Tsai <thebrokentoaster@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
7 years agoarchive/tar: add raw support for global PAX records
Joe Tsai [Fri, 25 Aug 2017 22:15:41 +0000 (15:15 -0700)]
archive/tar: add raw support for global PAX records

The PAX specification says the following:
<<<
'g' represents global extended header records for the following files in the archive.
The format of these extended header records shall be as described in pax Extended Header.
Each value shall affect all subsequent files that do not override that value
in their own extended header record and until another global extended header record
is reached that provides another value for the same field.
>>>

This CL adds support for parsing and composing global PAX records,
but intentionally does not provide support for automatically
persisting the global state across files.

Changes made:
* When Reader encounters a TypeXGlobalRecord header, it parses the
PAX records and returns them to the user ad-verbatim. Reader does not
store them in its state, ensuring it has no effect on future Next calls.
* When Writer receives a TypeXGlobalRecord header, it writes the
PAX records to the archive ad-verbatim. It does not store them in
its state, ensuring it has no effect on future WriteHeader calls.
* The restriction regarding empty record values is lifted since this
value is used to represent deletion in global headers.

Why provide raw support only:
* Some archives in the wild have a global header section (often empty)
and it is the user's responsibility to manually read and discard it's body.
The logic added here allows users to more easily skip over these sections.
* For users that do care about global headers, having access to the raw
records allows them to implement the functionality of global headers themselves
and manually persist the global state across files.
* We can still upgrade to a full implementation in the future.

Why we don't provide full support:
* Even though the PAX specification describes their operation in detail,
both the GNU and BSD tar tools (which are the most common implementations)
do not have a consistent interpretation of many details.
* Global headers were a controversial feature in PAX, by admission of the
specification itself:
  <<<
  The concept of a global extended header (typeflag g) was controversial.

  The typeflag g global headers should not be used with interchange media that
  could suffer partial data loss in transporting the archive.
  >>>
* Having state persist from entry-to-entry complicates the implementation
for a feature that is not widely used and not well supported.

Change-Id: I1d904cacc2623ddcaa91525a5470b7dbe226c7e8
Reviewed-on: https://go-review.googlesource.com/59190
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Joe Tsai <thebrokentoaster@gmail.com>

7 years agocmd/link: fix debug message
Hiroshi Ioka [Sat, 22 Jul 2017 00:11:35 +0000 (09:11 +0900)]
cmd/link: fix debug message

Change-Id: I6cb0ed9b726da34106ba239b57e2da732a8e1b71
Reviewed-on: https://go-review.googlesource.com/50730
Reviewed-by: Michael Hudson-Doyle <michael.hudson@canonical.com>
Reviewed-by: Avelino <t@avelino.xxx>
7 years agotesting: error if -parallel is given N<1
Daniel Martí [Wed, 9 Aug 2017 07:56:08 +0000 (16:56 +0900)]
testing: error if -parallel is given N<1

Otherwise, if there are any parallel tests, it will hang and panic with
"all goroutines are asleep - deadlock!".

Do not use flag.Uint to handle the error for us because we also want to
error on N==0, and because it would make setting the default to
GOMAXPROCS(0) more difficult, since it's an int.

Check for it right after flag.Parse, and mimic flag errors by printing
the usage and returning exit code 2.

Fixes #20542.

Change-Id: I0c9d4587f83d406a8f5e42ed74e40be46d639ffb
Reviewed-on: https://go-review.googlesource.com/54150
Run-TryBot: Daniel Martí <mvdan@mvdan.cc>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
7 years agotesting: ensure profiles are written upon -timeout panic
Meir Fischer [Fri, 14 Jul 2017 02:51:08 +0000 (22:51 -0400)]
testing: ensure profiles are written upon -timeout panic

This addresses the case of a -timeout panic, but not the more
general case of a signal arriving. See CL 48370 and CL 44352
for recent difficulties in that area.

"-timeout" here means flag usage to distinguish from the
default timeout termination which uses signals.

Fixes #19394

Change-Id: I5452d5422c0c080e940cbcc8c6606049975268c6
Reviewed-on: https://go-review.googlesource.com/48491
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

7 years agocrypto/tls: fix docstring of Config.ClientSessionCache
Artyom Pervukhin [Fri, 18 Aug 2017 10:26:41 +0000 (13:26 +0300)]
crypto/tls: fix docstring of Config.ClientSessionCache

Closes #21519

Change-Id: I1247e9435de93aae7e4db2b6e8e5be1b010c296b
Reviewed-on: https://go-review.googlesource.com/56832
Reviewed-by: Avelino <t@avelino.xxx>
Reviewed-by: Adam Langley <agl@golang.org>
7 years agoarchive/tar: support arbitrary PAX records
Joe Tsai [Thu, 24 Aug 2017 06:44:33 +0000 (23:44 -0700)]
archive/tar: support arbitrary PAX records

This CL adds the following new publicly visible API:
type Header struct { ...; PAXRecords map[string]string }

The new Header.PAXRecords field is a map of all PAX extended header records.

We suggest (but do not enforce) that users use VENDOR-prefixed keys
according to the following in the PAX specification:
<<<
The standard developers have reserved keyword name space for vendor extensions.
It is suggested that the format to be used is:
VENDOR.keyword
where VENDOR is the name of the vendor or organization in all uppercase letters.
>>>

When reading, the Header.PAXRecords is populated with all PAX records
encountered so far, including basic ones (e.g., "path", "mtime", etc).
When writing, the fields of Header will be merged into PAXRecords,
overwriting any records that may conflict.

Since PAXRecords is a more expressive feature than Xattrs and
is entirely a superset of Xattrs, we mark Xattrs as deprecated,
and steer users towards the new PAXRecords API.

The issue has a discussion about adding a Header.SetPAXRecord method
to help validate records and keep the Header fields in sync.
However, we do not include that in this CL since that helper
method can always be added in the future.

There is no support for global records.

Fixes #14472

Change-Id: If285a52749acc733476cf75a2c7ad15bc1542071
Reviewed-on: https://go-review.googlesource.com/58390
Run-TryBot: Joe Tsai <thebrokentoaster@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
7 years agogo/build: add go1.10 build tag
Joe Tsai [Fri, 25 Aug 2017 20:43:56 +0000 (13:43 -0700)]
go/build: add go1.10 build tag

Add this early in the cycle so that we can start regression testing
of the master toolchain.

Change-Id: Ida3ccad6e9642648f489babd12877fc8a5eca07a
Reviewed-on: https://go-review.googlesource.com/59151
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Joe Tsai <thebrokentoaster@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

7 years agoruntime: unify sigTabT type across Unix systems
Ian Lance Taylor [Fri, 25 Aug 2017 20:18:30 +0000 (13:18 -0700)]
runtime: unify sigTabT type across Unix systems

Change-Id: I8e8a3a118b1216f191c9076b70a88f6f3f19f79f
Reviewed-on: https://go-review.googlesource.com/59150
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
7 years agobytes: Add missing examples to functions
Borja Clemente [Wed, 23 Aug 2017 13:51:03 +0000 (15:51 +0200)]
bytes: Add missing examples to functions

Fixes #21570

Change-Id: Ia0734929a04fbce8fdd5fbcb1b7baff9a8bbe39e
Reviewed-on: https://go-review.googlesource.com/58030
Reviewed-by: Robert Griesemer <gri@golang.org>
Reviewed-by: Joe Tsai <thebrokentoaster@gmail.com>
Run-TryBot: Robert Griesemer <gri@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

7 years agosort: fix mix-up between "!less" and "greater" in examples
Tom Levy [Tue, 22 Aug 2017 05:10:46 +0000 (17:10 +1200)]
sort: fix mix-up between "!less" and "greater" in examples

If Less(a, b) returns true when a is less than b, the correct way to
check if a is greater than b is to use Less(b, a). It is wrong to use
!Less(a, b) because that checks if a is greater than *or equal to* b.

1. The decreasingDistance function in Example_sortKeys makes this
   mistake. Fix it.

2. The documentation of multiSorter.Less says it loops along the less
   functions until it finds a comparison "that is either Less or
   !Less". This is nonsense, because (Less(a, b) or !Less(a, b)) is
   always true. Fix the documentation to say that it finds a
   comparison "that discriminates between the two items (one is less
   than the other)". The implementation already does this correctly.

Change-Id: If52b79f68e4fdb0d1095edf29bdecdf154a61b8d
Reviewed-on: https://go-review.googlesource.com/57752
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

7 years agocmd/internal/obj/arm64: fix assemble fcsels/fcseld bug
fanzha02 [Wed, 14 Jun 2017 06:44:43 +0000 (06:44 +0000)]
cmd/internal/obj/arm64: fix assemble fcsels/fcseld bug

The current code treats the type of SIMD&FP register as C_REG incorrectly.

The fix code converts C_REG type into C_FREG type.

Uncomment fcsels/fcseld test cases.

Fixes #21582
Change-Id: I754c51f72a0418bd352cbc0f7740f14cc599c72d
Reviewed-on: https://go-review.googlesource.com/58350
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

7 years agocmd/compile: bug fixes for DWARF location lists
Heschi Kreinick [Fri, 25 Aug 2017 16:27:15 +0000 (12:27 -0400)]
cmd/compile: bug fixes for DWARF location lists

Fix two small but serious bugs in the DWARF location list code that
should have been caught by the automated tests I didn't write.

After emitting debug information for a user variable, mark it as done
so that it doesn't get emitted again. Otherwise it would be written once
per slot it was decomposed into.

Correct a merge error in CL 44350: the location list abbreviations need
to have DW_AT_decl_line too, otherwise the resulting DWARF is gibberish.

Change-Id: I6ab4b8b32b7870981dac80eadf0ebfc4015ccb01
Reviewed-on: https://go-review.googlesource.com/59070
Run-TryBot: Heschi Kreinick <heschi@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: David Chase <drchase@google.com>
7 years agomath: add examples for trig functions
jaredculp [Thu, 10 Aug 2017 21:00:59 +0000 (17:00 -0400)]
math: add examples for trig functions

Change-Id: Ic3ce2f3c055f2636ec8fc9cec8592e596b18dc05
Reviewed-on: https://go-review.googlesource.com/54771
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
7 years agocmd/compile: memory clearing optimization for arm64
Wei Xiao [Thu, 27 Jul 2017 01:55:03 +0000 (01:55 +0000)]
cmd/compile: memory clearing optimization for arm64

Use "STP (ZR, ZR), O(R)" instead of "MOVD ZR, O(R)" to implement memory clearing.
Also improve assembler supports to STP/LDP.
Results (A57@2GHzx8):

benchmark                   old ns/op     new ns/op     delta
BenchmarkClearFat8-8        1.00          1.00          +0.00%
BenchmarkClearFat12-8       1.01          1.01          +0.00%
BenchmarkClearFat16-8       1.01          1.01          +0.00%
BenchmarkClearFat24-8       1.52          1.52          +0.00%
BenchmarkClearFat32-8       3.00          2.02          -32.67%
BenchmarkClearFat40-8       3.50          2.52          -28.00%
BenchmarkClearFat48-8       3.50          3.03          -13.43%
BenchmarkClearFat56-8       4.00          3.50          -12.50%
BenchmarkClearFat64-8       4.25          4.00          -5.88%
BenchmarkClearFat128-8      8.01          8.01          +0.00%
BenchmarkClearFat256-8      16.1          16.0          -0.62%
BenchmarkClearFat512-8      32.1          32.0          -0.31%
BenchmarkClearFat1024-8     64.1          64.1          +0.00%

Change-Id: Ie5f5eac271ff685884775005825f206167a5c146
Reviewed-on: https://go-review.googlesource.com/55610
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
7 years agocmd/compile/internal/ssa: combine consecutive loads and stores on amd64
Ilya Tocar [Fri, 18 Aug 2017 19:03:33 +0000 (14:03 -0500)]
cmd/compile/internal/ssa: combine consecutive loads and stores on amd64

Sometimes (often for calls) we generate code like this:

MOVQ  (addr),AX
MOVQ  8(addr),BX
MOVQ  AX,(otheraddr)
MOVQ  BX,8(otheraddr)

Replace it with

MOVUPS (addr),X0
MOVUPS X0,(otheraddr)

For completeness do the same for 8,16,32-bit loads/stores too.
Shaves 1% from code sections of go tool.

/localdisk/itocar/golang/bin/go 10293917
go_old 10334877 [40960 bytes]

read-only data = 682 bytes (0.040769%)
global text (code) = 38961 bytes (1.036503%)
Total difference 39643 bytes (0.674628%)

Updates #6853

Change-Id: I1f0d2f60273a63a079b58927cd1c4e3429d2e7ae
Reviewed-on: https://go-review.googlesource.com/57130
Run-TryBot: Ilya Tocar <ilya.tocar@intel.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
7 years agospec: explicitly define notion of "representability" (clarification)
griesemer [Mon, 21 Aug 2017 13:47:51 +0000 (15:47 +0200)]
spec: explicitly define notion of "representability" (clarification)

Throughout the spec we use the notion of a constant x being
representable by a value of type T. While intuitively clear,
at least for floating-point and complex constants types, the
concept was not well-defined. In the section on Conversions
there was an extra rule for floating-point types only and it
missed the case of floating-point values overflowing to an
infinity after rounding.

Since the concept is important to Go, and a compiler most
certainly will have a function to test "representability",
it seems warranted to define the term explicitly in the spec.

This change introduces a new entry "Representability" under
the section on "Properties of types and values", and defines
the term explicitly, together with examples.

The phrase used is "representable by" rather than "representable as"
because the former use is prevalent in the spec.

Additionally, it clarifies that a floating-point constant
that overflows to an infinity after rounding is never
representable by a value of a floating-point type, even though
infinities are valid values of IEEE floating point types.
This is required because there are not infinite value constants
in the language (like there is also no -0.0) and representability
also matters for constant conversions. This is not a language
change, and type-checkers have been following this rule before.

The change also introduces links throughout the spec to the new
section as appropriate and removes duplicate text and examples
elsewhere (Constants and Conversions sections), leading to
simplifications in the relevant paragraphs.

Fixes #15389.

Change-Id: I8be0e071552df0f18998ef4c5ef521f64ffe8c44
Reviewed-on: https://go-review.googlesource.com/57530
Reviewed-by: Rob Pike <r@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
7 years agoruntime: optimize storing new keys in mapassign_fastNN
Josh Bleecher Snyder [Thu, 24 Aug 2017 21:50:35 +0000 (14:50 -0700)]
runtime: optimize storing new keys in mapassign_fastNN

Prior to this change, we use typedmemmove to write the key
value to its new location in mapassign_fast32 and mapassign_fast64.
(The use of typedmemmove was a last-minute fix in the 1.9 cycle;
see #21297 and CL 53414.)

This is significantly less inefficient than direct assignment or
calling writebarrierptr directly.

Fortunately, there aren't many cases to consider.

On systems with 32 bit pointers:

* A 32 bit AMEM value either is a single pointer or has no pointers.
* A 64 bit AMEM value may contain a pointer at the beginning,
  a pointer at 32 bits, or two pointers.

On systems with 64 bit pointers:

* A 32 bit AMEM value contains no pointers.
* A 64 bit AMEM value either is a single pointer or has no pointers.

All combinations except the 32 bit pointers / 64 bit AMEM value are
cheap and easy to handle, and the problematic case is likely rare.
The most popular map keys appear to be ints and pointers.

So we handle them exhaustively. The sys.PtrSize checks are constant branches
and are eliminated by the compiler.

An alternative fix would be to return a pointer to the key,
and have the calling code do the assignment, at which point the compiler
would have full type information.

Initial tests suggest that the performance difference between these
strategies is negligible, and this fix is considerably simpler,
and has much less impact on binary size.

Fixes #21321

Change-Id: Ib03200e89e2324dd3c76d041131447df66f22bfe
Reviewed-on: https://go-review.googlesource.com/59110
Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com>
Reviewed-by: Austin Clements <austin@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

7 years agocmd/compile: enforce that MOVXconvert is a no-op on 386 and amd64
Josh Bleecher Snyder [Thu, 24 Aug 2017 18:31:58 +0000 (11:31 -0700)]
cmd/compile: enforce that MOVXconvert is a no-op on 386 and amd64

Follow-up to CL 58371.

Change-Id: I3d2aaec84ee6db3ef1bd4fcfcaf46cc297c7176b
Reviewed-on: https://go-review.googlesource.com/58610
Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
7 years agocmd/compile,math: improve code generation for math.Abs
Keith Randall [Thu, 24 Aug 2017 20:19:40 +0000 (13:19 -0700)]
cmd/compile,math: improve code generation for math.Abs

Implement int reg <-> fp reg moves on amd64.
If we see a load to int reg followed by an int->fp move, then we can just
load to the fp reg instead.  Same for stores.

math.Abs is now:

MOVQ "".x+8(SP), AX
SHLQ $1, AX
SHRQ $1, AX
MOVQ AX, "".~r1+16(SP)

math.Copysign is now:

MOVQ "".x+8(SP), AX
SHLQ $1, AX
SHRQ $1, AX
MOVQ "".y+16(SP), CX
SHRQ $63, CX
SHLQ $63, CX
ORQ CX, AX
MOVQ AX, "".~r2+24(SP)

math.Float64bits is now:

MOVSD "".x+8(SP), X0
MOVSD X0, "".~r1+16(SP)
(it would be nicer to use a non-SSE reg for this, nothing is perfect)

And due to the fix for #21440, the inlined version of these improve as well.

name      old time/op  new time/op  delta
Abs       1.38ns ± 5%  0.89ns ±10%  -35.54%  (p=0.000 n=10+10)
Copysign  1.56ns ± 7%  1.35ns ± 6%  -13.77%  (p=0.000 n=9+10)

Fixes #13095

Change-Id: Ibd7f2792412a6668608780b0688a77062e1f1499
Reviewed-on: https://go-review.googlesource.com/58732
Run-TryBot: Keith Randall <khr@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Reviewed-by: Ilya Tocar <ilya.tocar@intel.com>
7 years agopath/filepath: add example for Ext
Kevin Burke [Thu, 27 Jul 2017 19:51:17 +0000 (12:51 -0700)]
path/filepath: add example for Ext

Make it dead simple to see visually what the function outputs in
various scenarios.

Change-Id: I8f6fcd72fa1515361481f0510412cde221e1d4e3
Reviewed-on: https://go-review.googlesource.com/51630
Run-TryBot: Kevin Burke <kev@inburke.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
Reviewed-by: Hiroshi Ioka <hirochachacha@gmail.com>
7 years agoruntime: capture runtimeInitTime after nanotime is initialized
Austin Clements [Thu, 24 Aug 2017 19:06:26 +0000 (15:06 -0400)]
runtime: capture runtimeInitTime after nanotime is initialized

CL 36428 changed the way nanotime works so on Darwin and Windows it
now depends on runtime.startNano, which is computed at runtime.init
time. Unfortunately, the `runtimeInitTime = nanotime()` initialization
happened *before* runtime.init, so on these platforms runtimeInitTime
is set incorrectly. The one (and only) consequence of this is that the
start time printed in gctrace lines is bogus:

gc 1 18446653480.186s 0%: 0.092+0.47+0.038 ms clock, 0.37+0.15/0.81/1.8+0.15 ms cpu, 4->4->1 MB, 5 MB goal, 8 P

To fix this, this commit moves the runtimeInitTime initialization to
shortly after runtime.init, at which point nanotime is safe to use.

This also requires changing the condition in newproc1 that currently
uses runtimeInitTime != 0 simply to detect whether or not the main M
has started. Since runtimeInitTime could genuinely be 0 now, this
introduces a separate flag to newproc1.

Fixes #21554.

Change-Id: Id874a4b912d3fa3d22f58d01b31ffb3548266d3b
Reviewed-on: https://go-review.googlesource.com/58690
Run-TryBot: Austin Clements <austin@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Rick Hudson <rlh@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
7 years agomisc/trace: update trace-viewer
Hana Kim [Thu, 24 Aug 2017 20:32:54 +0000 (16:32 -0400)]
misc/trace: update trace-viewer

Generated with
 github.com/catapult/tracing/bin/vulcanize_trace_viewer
catapult @ ab4d571fa

Renamed trace_viewer_lean.html to trace_viewer_full.html
to make it clear we are using the full version of trace viewer
(waiting for https://github.com/catapult-project/catapult/issues/2247
to be fixed).

Update #15302

Change-Id: Ice808bb27ab79a1dec9fc863e0c5a761027ebfbe
Reviewed-on: https://go-review.googlesource.com/58750
Reviewed-by: Dmitry Vyukov <dvyukov@google.com>
7 years agocmd/vendor/golang.org/x/arch: pull latest updates from x repo (commit edaf650)
Wei Xiao [Thu, 17 Aug 2017 05:56:44 +0000 (13:56 +0800)]
cmd/vendor/golang.org/x/arch: pull latest updates from x repo (commit edaf650)

Updates #21486

Change-Id: I78ca76490d8e9b52e055c1f0b8d10bdb227e3a80
Reviewed-on: https://go-review.googlesource.com/56331
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

7 years agoencoding/gob: fix Debug to properly print uint
Danny Rosseau [Thu, 10 Aug 2017 19:08:23 +0000 (13:08 -0600)]
encoding/gob: fix Debug to properly print uint

Fix debugger printing of uint that mistakenly
invoked .int64() instead of .uint64()

Fixes #21392

Change-Id: I107a7e87e0efbb06303c1e627dee76c369f75d1e
Reviewed-on: https://go-review.googlesource.com/54750
Reviewed-by: Emmanuel Odeke <emm.odeke@gmail.com>
Reviewed-by: Rob Pike <r@golang.org>
Run-TryBot: Emmanuel Odeke <emm.odeke@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

7 years agostrconv: optimize Atoi for common case
Aliaksandr Valialkin [Fri, 2 Jun 2017 16:16:19 +0000 (19:16 +0300)]
strconv: optimize Atoi for common case

Benchmark results on GOOS=linux:

GOARCH=amd64

name              old time/op  new time/op  delta
Atoi/Pos/7bit-4   20.1ns ± 2%   8.6ns ± 1%  -57.34%  (p=0.000 n=10+10)
Atoi/Pos/26bit-4  25.8ns ± 7%  11.9ns ± 0%  -53.91%  (p=0.000 n=10+8)
Atoi/Pos/31bit-4  27.3ns ± 2%  13.2ns ± 1%  -51.56%  (p=0.000 n=10+10)
Atoi/Pos/56bit-4  37.2ns ± 5%  18.2ns ± 1%  -51.26%  (p=0.000 n=10+10)
Atoi/Pos/63bit-4  38.7ns ± 1%  38.6ns ± 1%     ~     (p=0.297 n=9+10)
Atoi/Neg/7bit-4   17.6ns ± 1%   7.2ns ± 0%  -59.22%  (p=0.000 n=10+10)
Atoi/Neg/26bit-4  24.4ns ± 1%  12.4ns ± 1%  -49.28%  (p=0.000 n=10+10)
Atoi/Neg/31bit-4  26.9ns ± 0%  14.0ns ± 1%  -47.88%  (p=0.000 n=7+10)
Atoi/Neg/56bit-4  36.2ns ± 1%  19.5ns ± 0%  -46.24%  (p=0.000 n=10+9)
Atoi/Neg/63bit-4  38.9ns ± 1%  38.8ns ± 1%     ~     (p=0.385 n=9+10)

GOARCH=386

name              old time/op  new time/op  delta
Atoi/Pos/7bit-4   89.6ns ± 1%   8.2ns ± 1%  -90.84%  (p=0.000 n=9+10)
Atoi/Pos/26bit-4   187ns ± 2%    12ns ± 1%  -93.71%  (p=0.000 n=10+9)
Atoi/Pos/31bit-4   225ns ± 1%   225ns ± 1%     ~     (p=0.995 n=10+10)
Atoi/Neg/7bit-4   86.2ns ± 1%   8.5ns ± 1%  -90.14%  (p=0.000 n=10+10)
Atoi/Neg/26bit-4   183ns ± 1%    13ns ± 1%  -92.77%  (p=0.000 n=9+10)
Atoi/Neg/31bit-4   223ns ± 0%   223ns ± 0%     ~     (p=0.247 n=8+9)

Fixes #20557

Change-Id: Ib6245d88cffd4b037419e2bf8e4a71b86c6d773f
Reviewed-on: https://go-review.googlesource.com/44692
Reviewed-by: Robert Griesemer <gri@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
7 years agocmd/compile: simplify noding for struct embedding
Matthew Dempsky [Thu, 24 Aug 2017 22:36:47 +0000 (15:36 -0700)]
cmd/compile: simplify noding for struct embedding

Since golang.org/cl/31670, we've stopped using the 'embedded' function
for handling struct embeddings within package export data. Now the
only remaining use is for Go source files, which allows for some
substantial simplifications:

1. CenterDot never appears within Go source files, so that logic can
simply be removed.

2. The field name will always be declared in the local package.

Passes toolstash-check.

Change-Id: I59505f62824206dd5de0782918f98fbef6e93224
Reviewed-on: https://go-review.googlesource.com/58790
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Robert Griesemer <gri@golang.org>
7 years agospec: clarify zero value for complex types
griesemer [Thu, 24 Aug 2017 13:20:18 +0000 (15:20 +0200)]
spec: clarify zero value for complex types

The enumeration of numeric types missed the complex types.
Clarify by removing the explicit enumeration and referring
to numeric types instead.

Fixes #21579.

Change-Id: If36c2421f8501eeec82a07f442ac2e16a35927ba
Reviewed-on: https://go-review.googlesource.com/58491
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Reviewed-by: Rob Pike <r@golang.org>
7 years agocmd/compile: fix node position for imported constants
Matthew Dempsky [Wed, 23 Aug 2017 20:50:47 +0000 (13:50 -0700)]
cmd/compile: fix node position for imported constants

Discovered while debugging CL 53644.

No test case because these are purely internal conversions that should
never end up resulting in compiler warnings or even generated code.

Updates #19683.

Change-Id: I0d9333ef2c963fa22eb9b5335bb022bcc9b25708
Reviewed-on: https://go-review.googlesource.com/58190
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Robert Griesemer <gri@golang.org>
7 years agospec: clarify nil case in type switches
griesemer [Thu, 24 Aug 2017 13:05:53 +0000 (15:05 +0200)]
spec: clarify nil case in type switches

The old wording seemed to imply that nil is a kind of type.
Slightly reworded for clarity.

Fixes #21580.

Change-Id: I29898bf0125a10cb8dbb5c7e63ec5399ebc590ca
Reviewed-on: https://go-review.googlesource.com/58490
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
Reviewed-by: Rob Pike <r@golang.org>
7 years agocmd/compile: free value earlier in nilcheck
Keith Randall [Thu, 24 Aug 2017 23:11:44 +0000 (16:11 -0700)]
cmd/compile: free value earlier in nilcheck

When we remove a nil check, add it back to the free Value pool immediately.

Fixes #18732

Change-Id: I8d644faabbfb52157d3f2d071150ff0342ac28dc
Reviewed-on: https://go-review.googlesource.com/58810
Run-TryBot: Keith Randall <khr@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
7 years agosort: fix TestAdversary
Tom Levy [Thu, 24 Aug 2017 01:44:04 +0000 (13:44 +1200)]
sort: fix TestAdversary

There are some major problems with TestAdversary (based on "A Killer
Adversary for Quicksort"[1] by M. D. McIlroy). See #21581 for details.

Rewrite the test to closely match the version in the paper so it can
be verified as correct by virtue of similarity.

The only major difference between this new version and the version in
the paper is that this version swaps the values directly instead of
permuting an array of indices because we don't need to recover the
original permutation.

This new version also counts the number of calls to Less() and fails
the test if there are too many.

Fixes #21581.

[1]: http://www.cs.dartmouth.edu/~doug/mdmspe.pdf

Change-Id: Ia94b5b6d288b8fa3805a5fa27661cebbc5bad9a7
Reviewed-on: https://go-review.googlesource.com/58330
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
7 years agosync/atomic: remove references to old atomic pointer hammer tests
Tom Levy [Fri, 25 Aug 2017 05:13:34 +0000 (17:13 +1200)]
sync/atomic: remove references to old atomic pointer hammer tests

The tests were removed in https://golang.org/cl/2311 but some
references to them were missed.

Change-Id: I163e554a0cc99401a012deead8fda813ad74dbfe
Reviewed-on: https://go-review.googlesource.com/58870
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

7 years agocmd/compile: don't use MOVOstore instruction on plan9/amd64
David du Colombier [Thu, 24 Aug 2017 23:56:50 +0000 (01:56 +0200)]
cmd/compile: don't use MOVOstore instruction on plan9/amd64

CL 54410 and CL 56250 recently added use of the MOVOstore
instruction to improve performance.

However, we can't use the MOVOstore instruction on Plan 9,
because floating point operations are not allowed in the
note handler.

This change adds a configuration flag useSSE to enable the
use of SSE instructions for non-floating point operations.
This flag is enabled by default and disabled on Plan 9.
When this flag is disabled, the MOVOstore instruction is
not used and the MOVQstoreconst instruction is used instead.

Fixes #21599

Change-Id: Ie609e5d9b82ec0092ae874bab4ce01caa5bc8fb8
Reviewed-on: https://go-review.googlesource.com/58850
Reviewed-by: Keith Randall <khr@golang.org>
7 years agobuild: add `go env GOROOT` as default GOROOT_BOOTSTRAP value
Wei Congrui [Tue, 22 Aug 2017 05:36:19 +0000 (13:36 +0800)]
build: add `go env GOROOT` as default GOROOT_BOOTSTRAP value

This change also added the same check in make.bash to make.rc,
which makes sure $GOROOT_BOOTSTRAP != $GOROOT.

Fixes #14339

Change-Id: I2758f4a845bae42ace02492fc6a911f6d6247d26
Reviewed-on: https://go-review.googlesource.com/57753
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
7 years agoarchive/tar: return better WriteHeader errors
Joe Tsai [Thu, 24 Aug 2017 01:36:46 +0000 (18:36 -0700)]
archive/tar: return better WriteHeader errors

WriteHeader may fail to encode a header for any number of reasons,
which can be frustrating for the user when trying to create a tar archive.
As we validate the Header, we generate an informative error message
intended for human consumption and return that if and only if no
format can be selected.

This allows WriteHeader to return informative errors like:
    tar: cannot encode header: invalid PAX record: "linkpath = \x00hello"
    tar: cannot encode header: invalid PAX record: "SCHILY.xattr.foo=bar = baz"
    tar: cannot encode header: Format specifies GNU; and only PAX supports Xattrs
    tar: cannot encode header: Format specifies GNU; and GNU cannot encode ModTime=1969-12-31 15:59:59.0000005 -0800 PST
    tar: cannot encode header: Format specifies GNU; and GNU supports sparse files only with TypeGNUSparse
    tar: cannot encode header: Format specifies USTAR; and USTAR cannot encode ModTime=292277026596-12-04 07:30:07 -0800 PST
    tar: cannot encode header: Format specifies USTAR; and USTAR does not support sparse files
    tar: cannot encode header: Format specifies PAX; and only GNU supports TypeGNUSparse

Updates #18710

Change-Id: I82a498d6f29d02c4e73bce47b768eb578da8499c
Reviewed-on: https://go-review.googlesource.com/58310
Run-TryBot: Joe Tsai <thebrokentoaster@gmail.com>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
7 years agocmd/compile: remove more nil ptr checks after newobject
Keith Randall [Thu, 24 Aug 2017 22:23:27 +0000 (15:23 -0700)]
cmd/compile: remove more nil ptr checks after newobject

For code like the following (where x escapes):

   x := []int{1}

We're currently generating a nil check.  The line above is really 3 operations:

t := new([1]int)
t[0] = 1
x := t[:]

We remove the nil check for t[0] = 1, but not for t[:].

Our current nil check removal rule is too strict about the possible
memory arguments of the nil check. Unlike zeroing or storing to the
result of runtime.newobject, the nilness of runtime.newobject is
always false, even after other stores have happened in the meantime.

Change-Id: I95fad4e3a59c27effdb37c43ea215e18f30b1e5f
Reviewed-on: https://go-review.googlesource.com/58711
Run-TryBot: Keith Randall <khr@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
7 years agocmd/vet: check only for ASCII spaces (0x20) in struct tags
Francesc Campoy Flores [Thu, 24 Aug 2017 17:07:15 +0000 (10:07 -0700)]
cmd/vet: check only for ASCII spaces (0x20) in struct tags

Change-Id: I6e9b5caeca842b6bf72afefb31f5140608b86d20
Reviewed-on: https://go-review.googlesource.com/58530
Run-TryBot: Francesc Campoy Flores <campoy@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Rob Pike <r@golang.org>
7 years agocmd/link: don't create go.info symbols for non-Go functions
Heschi Kreinick [Thu, 24 Aug 2017 18:29:13 +0000 (14:29 -0400)]
cmd/link: don't create go.info symbols for non-Go functions

In writelines the linker uses various auxiliary information about a
function to create its line table entries. (It also does some unrelated
stuff, but never mind.) There's no reason to do this for non-Go
functions, so it bails out if the symbol has no FuncInfo.

However, it does so *after* it looks up (and implicitly creates!) the
go.info symbol for the function, which doesn't make sense and risks
creating duplicate symbols for static C functions. Move the check up so
that it doesn't do that.

Since non-Go functions can't reference Go types, there shouldn't be any
relocations to type info DIEs that need to be built, so there should be
no harm not doing that.

I wanted to change the Lookup to an ROLookup but that broke the
shared-mode tests with an inscrutable error.

No test. It seems too specific to worry about, but if someone disagrees
I can figure something out.

Fixes #21566

Change-Id: I61f03b7c504a3bf1c4245a8811795b6303469e91
Reviewed-on: https://go-review.googlesource.com/58630
Reviewed-by: Ian Lance Taylor <iant@golang.org>
7 years agodoc: add go1.9 to golang.org/project
Chris Broadfoot [Tue, 1 Aug 2017 00:28:35 +0000 (17:28 -0700)]
doc: add go1.9 to golang.org/project

Pre-emptive. Go 1.9 is expected to be released in August.

Change-Id: I0f58c012c4110bf490022dc2c1d69c0988d73bfa
Reviewed-on: https://go-review.googlesource.com/52351
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
7 years agodoc: document go1.9
Chris Broadfoot [Thu, 24 Aug 2017 18:52:35 +0000 (11:52 -0700)]
doc: document go1.9

Change-Id: I97075f24319a4b96cbeb9e3ff2e7b2056ff59e32
Reviewed-on: https://go-review.googlesource.com/58651
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
7 years agoruntime: refactor walking of bucket overflows
Josh Bleecher Snyder [Wed, 23 Aug 2017 14:49:25 +0000 (07:49 -0700)]
runtime: refactor walking of bucket overflows

This eliminates a nil check of b while evaluating b.tophash,
which is in the inner loop of many hot map functions.
It also makes the code a bit clearer.

Also remove some gotos in favor of labeled breaks.

On non-x86 architectures, this change introduces a pointless reg-reg move,
although the cause is well-understood (#21572).

Change-Id: Ib7ee58b59ea5463b92e1590c8b8f5c0ef87d410a
Reviewed-on: https://go-review.googlesource.com/58372
Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Martin Möhrmann <moehrmann@google.com>
Reviewed-by: Keith Randall <khr@golang.org>
7 years agocmd/link: set correct alignment of ELF note section
Ian Lance Taylor [Thu, 24 Aug 2017 00:44:51 +0000 (17:44 -0700)]
cmd/link: set correct alignment of ELF note section

Otherwise the default computation in symalign kicked in, setting the
alignment to be too high. This didn't matter with GNU ld, which put
each loadable note into a separate PT_NOTE segment, but it did matter
with gold which accumulated them all into a single PT_NOTE segment,
respecting the requested alignment. In the single PT_NOTE segment
generated by gold, the incorrect section alignment made the notes
unreadable.

Fixes #21564

Change-Id: I15eb408bb04a2566c9fdfb6828e14188d9ef2280
Reviewed-on: https://go-review.googlesource.com/58290
Reviewed-by: Russ Cox <rsc@golang.org>
7 years agocmd/compile: eliminate stores to unread auto variables
Michael Munday [Wed, 29 Mar 2017 18:01:41 +0000 (14:01 -0400)]
cmd/compile: eliminate stores to unread auto variables

This is a crude compiler pass to eliminate stores to auto variables
that are only ever written to.

Eliminates an unnecessary store to x from the following code:

func f() int {
var x := 1
return *(&x)
}

Fixes #19765.

Change-Id: If2c63a8ae67b8c590b6e0cc98a9610939a3eeffa
Reviewed-on: https://go-review.googlesource.com/38746
Run-TryBot: Michael Munday <mike.munday@ibm.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
7 years agocmd/compile: mark MOVQconvert as resultInArg0 on x86 architectures
Josh Bleecher Snyder [Wed, 23 Aug 2017 23:46:00 +0000 (16:46 -0700)]
cmd/compile: mark MOVQconvert as resultInArg0 on x86 architectures

This prevents unnecessary reg-reg moves during pointer arithmetic.

This change reduces the size of the full hello world binary by 0.4%.

Updates #21572

Change-Id: Ia0427021e5c94545a0dbd83a6801815806e5b12d
Reviewed-on: https://go-review.googlesource.com/58371
Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com>
Run-TryBot: Martin Möhrmann <moehrmann@google.com>
Reviewed-by: Keith Randall <khr@golang.org>
7 years agoruntime: convert more unsafe.Pointer arithmetic to add
Josh Bleecher Snyder [Wed, 23 Aug 2017 14:48:04 +0000 (07:48 -0700)]
runtime: convert more unsafe.Pointer arithmetic to add

Change-Id: Icfe24d5660666093f3e645f82d30b7687c8077be
Reviewed-on: https://go-review.googlesource.com/58370
Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Martin Möhrmann <moehrmann@google.com>
7 years agoruntime: avoid infinite loop in growslice
Martin Möhrmann [Wed, 23 Aug 2017 07:05:29 +0000 (09:05 +0200)]
runtime: avoid infinite loop in growslice

On 386 the below code triggered an infinite loop in growslice:
x = make([]byte, 1<<30-1, 1<<30-1)
x = append(x, x...)

Check for overflow when calculating the new slice capacity
and set the new capacity to the requested capacity when an overflow
is detected to avoid an infinite loop.

No automatic test added due to requiring to allocate 1GB of memory
on a 32bit plaform before use of append is able to trigger the
overflow check.

Fixes #21441

Change-Id: Ia871cc9f88479dacf2c7044531b233f83d2fcedf
Reviewed-on: https://go-review.googlesource.com/57950
Run-TryBot: Martin Möhrmann <moehrmann@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Marvin Stenger <marvin.stenger94@gmail.com>
Reviewed-by: Keith Randall <khr@golang.org>
7 years agotesting: parallelize tests over count
Dmitry Vyukov [Mon, 5 Jun 2017 08:37:37 +0000 (10:37 +0200)]
testing: parallelize tests over count

Currently all package tests are executed once
with Parallel tests executed in parallel.
Then this process is repeated count*cpu times.
Tests are not parallelized over count*cpu.
Parallelizing over cpu is not possible as
GOMAXPROCS is a global setting. But it is
possible for count.

Parallelize over count.

Brings down testing of my package with -count=100
form 10s to 0.3s.

Change-Id: I76d8322adeb8c5c6e70b99af690291fd69d6402a
Reviewed-on: https://go-review.googlesource.com/44830
Run-TryBot: Dmitry Vyukov <dvyukov@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>