]> Cypherpunks repositories - gostls13.git/log
gostls13.git
9 years agocmd/go: add -shared to darwin/arm{,64} default build mode
Elias Naur [Sat, 7 May 2016 05:24:39 +0000 (07:24 +0200)]
cmd/go: add -shared to darwin/arm{,64} default build mode

Buildmode c-archive now supports position independent code for
darwin/arm (in addition to darwin/arm64). Make PIC (-shared) the
default for both platforms in the default buildmode.

Without this change, gomobile will go install the standard library
into its separate package directory without PIC support.

Also add -shared to darwin/arm64 in buildmode c-archive, for
symmetry (darwin/arm64 always generates position independent code).

Fixes #15519

Change-Id: If27d2cbea8f40982e14df25da2703cbba572b5c6
Reviewed-on: https://go-review.googlesource.com/22920
Reviewed-by: David Crawshaw <crawshaw@golang.org>
Run-TryBot: David Crawshaw <crawshaw@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agocmd/compile: properly handle map assignments for OAS2DOTTYPE
Tal Shprecher [Thu, 5 May 2016 22:14:08 +0000 (15:14 -0700)]
cmd/compile: properly handle map assignments for OAS2DOTTYPE

The boolean destination in an OAS2DOTTYPE expression craps out during
compilation when trying to assign to a map entry because, unlike slice entries,
map entries are not directly addressable in memory. The solution is to
properly order the boolean destination node so that map entries are set
via autotmp variables.

Fixes #14678

Change-Id: If344e8f232b5bdac1b53c0f0d21eeb43ab17d3de
Reviewed-on: https://go-review.googlesource.com/22833
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agoruntime: use entire address space on 32 bit
Elias Naur [Wed, 9 Mar 2016 09:00:12 +0000 (10:00 +0100)]
runtime: use entire address space on 32 bit

In issue #13992, Russ mentioned that the heap bitmap footprint was
halved but that the bitmap size calculation hadn't been updated. This
presents the opportunity to either halve the bitmap size or double
the addressable virtual space. This CL doubles the addressable virtual
space. On 32 bit this can be tweaked further to allow the bitmap to
cover the entire 4GB virtual address space, removing a failure mode
if the kernel hands out memory with a too low address.

First, fix the calculation and double _MaxArena32 to cover 4GB virtual
memory space with the same bitmap size (256 MB).

Then, allow the fallback mode for the initial memory reservation
on 32 bit (or 64 bit with too little available virtual memory) to not
include space for the arena. mheap.sysAlloc will automatically reserve
additional space when the existing arena is full.

Finally, set arena_start to 0 in 32 bit mode, so that any address is
acceptable for subsequent (additional) reservations.

Before, the bitmap was always located just before arena_start, so
fix the two places relying on that assumption: Point the otherwise unused
mheap.bitmap to one byte after the end of the bitmap, and use it for
bitmap addressing instead of arena_start.

With arena_start set to 0 on 32 bit, the cgoInRange check is no longer a
sufficient check for Go pointers. Introduce and call inHeapOrStack to
check whether a pointer is to the Go heap or stack.

While we're here, remove sysReserveHigh which seems to be unused.

Fixes #13992

Change-Id: I592b513148a50b9d3967b5c5d94b86b3ec39acc2
Reviewed-on: https://go-review.googlesource.com/20471
Reviewed-by: Austin Clements <austin@google.com>
Run-TryBot: Austin Clements <austin@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agonet/url: remove RFC 3986 mention in package comment
Brad Fitzpatrick [Fri, 6 May 2016 18:33:39 +0000 (18:33 +0000)]
net/url: remove RFC 3986 mention in package comment

Change-Id: Ifd707a4bbfcb1721655b4fce2045f3b043e66818
Reviewed-on: https://go-review.googlesource.com/22859
Reviewed-by: Russ Cox <rsc@golang.org>
9 years agonet: skip more flaky net tests on flaky net builders
Brad Fitzpatrick [Fri, 6 May 2016 18:46:35 +0000 (18:46 +0000)]
net: skip more flaky net tests on flaky net builders

e.g. https://storage.googleapis.com/go-build-log/9b937dd8/linux-arm_df54a25a.log

Change-Id: Ic5864c7bd840b4f0c6341f919fcbcd5c708b14e7
Reviewed-on: https://go-review.googlesource.com/22881
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>

9 years agomisc/cgo/testcarchive: avoid possible pthread_create race
Ian Lance Taylor [Fri, 6 May 2016 18:32:18 +0000 (11:32 -0700)]
misc/cgo/testcarchive: avoid possible pthread_create race

The old code assumed that the thread ID set by pthread_create would be
available in the newly created thread.  While that is clearly true
eventually, it is not necessarily true immediately.  Rather than try to
pass down the thread ID, just call pthread_self in the created thread.

Fixes #15576 (I hope).

Change-Id: Ic07086b00e4fd5676c04719a299c583320da64a1
Reviewed-on: https://go-review.googlesource.com/22880
Run-TryBot: Ian Lance Taylor <iant@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
9 years agoRevert "net/url: validate ports in IPv4 addresses"
Russ Cox [Fri, 6 May 2016 15:34:25 +0000 (15:34 +0000)]
Revert "net/url: validate ports in IPv4 addresses"

This reverts commit 9f1ccd647fcdb1b703c1042c90434e15aff75013.

For #14860.

Change-Id: I63522a4dda8915dc8b972ae2e12495553ed65f09
Reviewed-on: https://go-review.googlesource.com/22861
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Run-TryBot: Russ Cox <rsc@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agonet/http: wait longer for subprocess to startup in test
Brad Fitzpatrick [Fri, 6 May 2016 18:21:22 +0000 (18:21 +0000)]
net/http: wait longer for subprocess to startup in test

Might deflake the occasional linux-amd64-race failures.

Change-Id: I273b0e32bb92236168eb99887b166e079799c1f1
Reviewed-on: https://go-review.googlesource.com/22858
Reviewed-by: Ian Lance Taylor <iant@golang.org>
9 years agonet/http: don't assume Response.Request is populated after redirect errors
Brad Fitzpatrick [Fri, 6 May 2016 18:11:38 +0000 (18:11 +0000)]
net/http: don't assume Response.Request is populated after redirect errors

Fixes #15577

Change-Id: I5f023790a393b17235db2e66c02c2483773ddc1a
Reviewed-on: https://go-review.googlesource.com/22857
Reviewed-by: Russ Cox <rsc@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agoRevert "testing/quick: generate more map and slice states"
Russ Cox [Fri, 6 May 2016 15:28:19 +0000 (15:28 +0000)]
Revert "testing/quick: generate more map and slice states"

This reverts commit 0ccabe2e0b42a2602e0f37ce28d5368aa811f530.

Change-Id: Ib1c230fb6801c0ee26f4a352b0c1130fa240a76a
Reviewed-on: https://go-review.googlesource.com/22860
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
9 years agoos: rename remaining four os1_*.go files to os_*.go
Brad Fitzpatrick [Fri, 6 May 2016 16:07:11 +0000 (16:07 +0000)]
os: rename remaining four os1_*.go files to os_*.go

Change-Id: Ice9c234960adc7857c8370b777a0b18e29d59281
Reviewed-on: https://go-review.googlesource.com/22853
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agonet: ignore network failures on some builders
Brad Fitzpatrick [Fri, 6 May 2016 16:24:57 +0000 (16:24 +0000)]
net: ignore network failures on some builders

We run the external network tests on builders, but some of our
builders have less-than-ideal DNS connectivity. This change continues
to run the tests on all builders, but marks certain builders as flaky
(network-wise), and only validates their DNS results if they got DNS
results.

Change-Id: I826dc2a6f6da55add89ae9c6db892b3b2f7b526b
Reviewed-on: https://go-review.googlesource.com/22852
Reviewed-by: Ian Lance Taylor <iant@golang.org>
9 years agoruntime: delete empty files
Brad Fitzpatrick [Fri, 6 May 2016 16:06:02 +0000 (16:06 +0000)]
runtime: delete empty files

I meant to delete these in CL 22850, actually.

Change-Id: I0c286efd2b9f1caf0221aa88e3bcc03649c89517
Reviewed-on: https://go-review.googlesource.com/22851
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agoall: fix copy-and-paste errors in tests
Alberto Donizetti [Fri, 6 May 2016 16:16:52 +0000 (18:16 +0200)]
all: fix copy-and-paste errors in tests

Fixes #15570

Change-Id: I95d1ac26e342c3bbf36ad1f0209711ea96eaf487
Reviewed-on: https://go-review.googlesource.com/22870
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agocmd/vet: check sync.* types' copying
Aliaksandr Valialkin [Thu, 14 Apr 2016 21:33:28 +0000 (00:33 +0300)]
cmd/vet: check sync.* types' copying

Embed noLock struct into the following types, so `go vet -copylocks` catches
their copying additionally to types containing sync.Mutex:
  - sync.Cond
  - sync.WaitGroup
  - sync.Pool
  - atomic.Value

Fixes #14582

Change-Id: Icb543ef5ad10524ad239a15eec8a9b334b0e0660
Reviewed-on: https://go-review.googlesource.com/22015
Reviewed-by: Russ Cox <rsc@golang.org>
Run-TryBot: Russ Cox <rsc@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agosyscall,os,net: don't use ForkLock in plan9
Richard Miller [Fri, 6 May 2016 13:21:52 +0000 (14:21 +0100)]
syscall,os,net: don't use ForkLock in plan9

This is the follow-on to CL 22610: now that it's the child instead of
the parent which lists unwanted fds to close in syscall.StartProcess,
plan9 no longer needs the ForkLock to protect the list from changing.
The readdupdevice function is also now unused and can be removed.

Change-Id: I904c8bbf5dbaa7022b0f1a1de0862cd3064ca8c7
Reviewed-on: https://go-review.googlesource.com/22842
Reviewed-by: David du Colombier <0intro@gmail.com>
Run-TryBot: David du Colombier <0intro@gmail.com>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agoruntime: merge the last four os-vs-os1 files together
Brad Fitzpatrick [Fri, 6 May 2016 15:26:37 +0000 (08:26 -0700)]
runtime: merge the last four os-vs-os1 files together

Change-Id: Ib0ba691c4657fe18a4659753e70d97c623cb9c1d
Reviewed-on: https://go-review.googlesource.com/22850
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
9 years agocmd/compile/internal/mips64: fix large uint -> float conversion
Cherry Zhang [Thu, 5 May 2016 23:51:54 +0000 (16:51 -0700)]
cmd/compile/internal/mips64: fix large uint -> float conversion

Re-enable TestFP in cmd/compile/internal/gc on mips64.

Fixes #15552.

Change-Id: I5c3a5564b94d28c723358f0862468fb6da371991
Reviewed-on: https://go-review.googlesource.com/22835
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Reviewed-by: Minux Ma <minux@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agocrypto/sha1: Add AVX2 version for AMD64
Ilya Tocar [Fri, 29 Apr 2016 13:17:14 +0000 (16:17 +0300)]
crypto/sha1: Add AVX2 version for AMD64

name             old time/op    new time/op    delta
Hash8Bytes-48       271ns ± 8%     273ns ± 5%     ~     (p=0.313 n=19+19)
Hash320Bytes-48    1.04µs ± 7%    0.75µs ± 8%  -27.66%  (p=0.000 n=20+20)
Hash1K-48          2.72µs ± 6%    1.75µs ± 6%  -35.79%  (p=0.000 n=19+20)
Hash8K-48          19.9µs ± 7%    11.6µs ± 6%  -41.84%  (p=0.000 n=20+19)

name             old speed      new speed      delta
Hash8Bytes-48    29.5MB/s ± 8%  29.3MB/s ± 5%     ~     (p=0.314 n=19+19)
Hash320Bytes-48   307MB/s ± 7%   424MB/s ± 8%  +38.29%  (p=0.000 n=20+20)
Hash1K-48         377MB/s ± 6%   587MB/s ± 6%  +55.76%  (p=0.000 n=19+20)
Hash8K-48         413MB/s ± 7%   709MB/s ± 6%  +71.85%  (p=0.000 n=20+19)

Change-Id: I2963cf744eeb2e8191d4e4223fbf6f533a7fd405
Reviewed-on: https://go-review.googlesource.com/22607
Run-TryBot: Ilya Tocar <ilya.tocar@intel.com>
Reviewed-by: Russ Cox <rsc@golang.org>
Run-TryBot: Russ Cox <rsc@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agocrypto/sha256: Use AVX2 if possible
Ilya Tocar [Fri, 29 Apr 2016 13:19:17 +0000 (16:19 +0300)]
crypto/sha256: Use AVX2 if possible

name          old time/op    new time/op    delta
Hash8Bytes-4     376ns ± 0%     246ns ± 0%  -34.57%  (p=0.000 n=20+20)
Hash1K-4        5.21µs ± 0%    2.82µs ± 0%  -45.83%  (p=0.000 n=20+20)
Hash8K-4        38.6µs ± 0%    20.8µs ± 0%  -46.05%  (p=0.000 n=20+20)

name          old speed      new speed      delta
Hash8Bytes-4  21.2MB/s ± 0%  32.4MB/s ± 0%  +52.70%  (p=0.000 n=15+19)
Hash1K-4       197MB/s ± 0%   363MB/s ± 0%  +84.60%  (p=0.000 n=20+20)
Hash8K-4       212MB/s ± 0%   393MB/s ± 0%  +85.36%  (p=0.000 n=20+20)

Change-Id: Ib50119c591074ff486d11d3566e24b691bcc6598
Reviewed-on: https://go-review.googlesource.com/22608
Run-TryBot: Russ Cox <rsc@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Russ Cox <rsc@golang.org>
9 years agomisc/cgo/testcarchive: don't force -no_pie on Darwin
Elias Naur [Fri, 6 May 2016 13:09:57 +0000 (15:09 +0200)]
misc/cgo/testcarchive: don't force -no_pie on Darwin

Now that darwin/arm supports position independent code, allow the
binaries generated by the c-archive tests be position independent
(PIE) as well.

Change-Id: If0517f06e92349ada29a4e3e0a951f08b0fcc710
Reviewed-on: https://go-review.googlesource.com/22841
Reviewed-by: David Crawshaw <crawshaw@golang.org>
Run-TryBot: Elias Naur <elias.naur@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agocmd/internal/obj/x86: add AVX2 instrutions needed for sha1/sha512/sha256 acceleration
Ilya Tocar [Fri, 29 Apr 2016 13:14:57 +0000 (16:14 +0300)]
cmd/internal/obj/x86: add AVX2 instrutions needed for sha1/sha512/sha256 acceleration

This means: VPSHUFB, VPSHUFD, VPERM2F128, VPALIGNR, VPADDQ, VPADDD, VPSRLDQ,
VPSLLDQ, VPSRLQ, VPSLLQ, VPSRLD, VPSLLD, VPOR, VPBLENDD, VINSERTI128,
VPERM2I128, RORXL, RORXQ.

Change-Id: Ief27190ee6acfa86b109262af5d999bc101e923d
Reviewed-on: https://go-review.googlesource.com/22606
Run-TryBot: Ilya Tocar <ilya.tocar@intel.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Russ Cox <rsc@golang.org>
9 years agoruntime: get randomness from AT_RANDOM AUXV on linux/mips64x
Shenghou Ma [Fri, 6 May 2016 04:53:42 +0000 (00:53 -0400)]
runtime: get randomness from AT_RANDOM AUXV on linux/mips64x

Fixes #15148.

Change-Id: If3b628f30521adeec1625689dbc98aaf4a9ec858
Reviewed-on: https://go-review.googlesource.com/22811
Reviewed-by: Keith Randall <khr@golang.org>
Run-TryBot: Minux Ma <minux@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agocmd/go, cmd/cgo: pass "-mabi=64" to gcc on mips64
Cherry Zhang [Thu, 5 May 2016 16:10:49 +0000 (09:10 -0700)]
cmd/go, cmd/cgo: pass "-mabi=64" to gcc on mips64

Change-Id: I9ac2ae57a00cee23d6255db02419b0a0f087d4f3
Reviewed-on: https://go-review.googlesource.com/22801
Reviewed-by: Minux Ma <minux@golang.org>
Run-TryBot: Minux Ma <minux@golang.org>

9 years agotest: update test for issue 15548
Robert Griesemer [Fri, 6 May 2016 00:46:58 +0000 (17:46 -0700)]
test: update test for issue 15548

Accidentally checked in the version of file c.go that doesn't
exhibit the bug - hence the test was not testing the bug fix.
Double-checked that this version exposes the bug w/o the fix.

Change-Id: Ie4dc455229d1ac802a80164b5d549c2ad4d971f5
Reviewed-on: https://go-review.googlesource.com/22837
Run-TryBot: Robert Griesemer <gri@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Minux Ma <minux@golang.org>
9 years agocmd/go: fail with nice error message on bad GOOS/GOARCH pair
Brad Fitzpatrick [Fri, 6 May 2016 00:52:37 +0000 (17:52 -0700)]
cmd/go: fail with nice error message on bad GOOS/GOARCH pair

Fixes #12272

Change-Id: I2115ec62ed4061084c482eb385a583a1c1909888
Reviewed-on: https://go-review.googlesource.com/22838
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Reviewed-by: Minux Ma <minux@golang.org>
9 years agoarchive/tar: centralize all information about tar header format
Joe Tsai [Thu, 17 Sep 2015 07:22:56 +0000 (00:22 -0700)]
archive/tar: centralize all information about tar header format

The Reader and Writer have hard-coded constants regarding the
offsets and lengths of certain fields in the tar format sprinkled
all over. This makes it harder to verify that the offsets are
correct since a reviewer would need to search for them throughout
the code. Instead, all information about the layout of header
fields should be centralized in one single file. This has the
advantage of being both centralized, and also acting as a form
of documentation about the header struct format.

This method was chosen over using "encoding/binary" since that
method would cause an allocation of a header struct every time
binary.Read was called. This method causes zero allocations and
its logic is no longer than if structs were declared.

Updates #12594

Change-Id: Ic7a0565d2a2cd95d955547ace3b6dea2b57fab34
Reviewed-on: https://go-review.googlesource.com/14669
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
9 years agoruntime: stop traceback at foreign function
Russ Cox [Wed, 6 Jan 2016 17:45:23 +0000 (12:45 -0500)]
runtime: stop traceback at foreign function

This can only happen when profiling and there is foreign code
at the top of the g0 stack but we're not in cgo.
That in turn only happens with the race detector.

Fixes #13568.

Change-Id: I23775132c9c1a3a3aaae191b318539f368adf25e
Reviewed-on: https://go-review.googlesource.com/18322
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agoRevert "net: add support for Zone of IPNet"
Mikio Hara [Thu, 5 May 2016 23:51:16 +0000 (08:51 +0900)]
Revert "net: add support for Zone of IPNet"

Updates #14518.

This reverts commit 3e9264c9ae781a2cd28127deaed6ae26f84b4b15.

Change-Id: I2531b04efc735b5b51ef675541172f2f5ae747d9
Reviewed-on: https://go-review.googlesource.com/22836
Reviewed-by: Mikio Hara <mikioh.mikioh@gmail.com>
Run-TryBot: Mikio Hara <mikioh.mikioh@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
9 years agoall: use SeekStart, SeekCurrent, SeekEnd
Joe Tsai [Tue, 5 Apr 2016 18:22:53 +0000 (11:22 -0700)]
all: use SeekStart, SeekCurrent, SeekEnd

CL/19862 (f79b50b8d5bc159561c1dcf7c17e2a0db96a9a11) recently introduced the constants
SeekStart, SeekCurrent, and SeekEnd to the io package. We should use these constants
consistently throughout the code base.

Updates #15269

Change-Id: If7fcaca7676e4a51f588528f5ced28220d9639a2
Reviewed-on: https://go-review.googlesource.com/22097
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Run-TryBot: Joe Tsai <joetsai@digital-static.net>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agocmd/compile: repair MININT conversion bug in arm softfloat
David Chase [Thu, 5 May 2016 20:35:10 +0000 (13:35 -0700)]
cmd/compile: repair MININT conversion bug in arm softfloat

Negative-case conversion code was wrong for minimum int32,
used negate-then-widen instead of widen-then-negate.

Test already exists; this fixes the failure.

Fixes #15563.

Change-Id: I4b0b3ae8f2c9714bdcc405d4d0b1502ccfba2b40
Reviewed-on: https://go-review.googlesource.com/22830
Run-TryBot: David Chase <drchase@google.com>
Reviewed-by: Keith Randall <khr@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agogo/token: document postcondition of SetLines
Alan Donovan [Thu, 5 May 2016 21:56:58 +0000 (14:56 -0700)]
go/token: document postcondition of SetLines

Change-Id: Ie163deade396b3e298a93845b9ca4d52333ea82a
Reviewed-on: https://go-review.googlesource.com/22831
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
9 years agoos/exec: re-enable TestExtraFiles for plan9
Richard Miller [Fri, 29 Apr 2016 20:02:59 +0000 (21:02 +0100)]
os/exec: re-enable TestExtraFiles for plan9

This test should now succeed after CL 22610 which fixes issue #7118

Change-Id: Ie785a84d77b27c832a1ddd81699bf25dab24b97d
Reviewed-on: https://go-review.googlesource.com/22640
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Reviewed-by: David du Colombier <0intro@gmail.com>
Run-TryBot: David du Colombier <0intro@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agosyscall: simplify closing of extra fds in plan9 StartProcess
Richard Miller [Fri, 29 Apr 2016 16:39:33 +0000 (17:39 +0100)]
syscall: simplify closing of extra fds in plan9 StartProcess

Reviving earlier work by @ality in https://golang.org/cl/57890043
to make the closing of extra file descriptors in syscall.StartProcess
less race-prone. Instead of making a list of open fds in the parent
before forking, the child can read through the list of open fds and
close the ones not explicitly requested.  Also eliminate the
complication of keeping open any extra fds which were inherited by
the parent when it started.

This CL will be followed by one to eliminate the ForkLock in plan9,
which is now redundant.

Fixes #5605

Change-Id: I6b4b942001baa54248b656c52dced3b62021c486
Reviewed-on: https://go-review.googlesource.com/22610
Run-TryBot: David du Colombier <0intro@gmail.com>
Reviewed-by: David du Colombier <0intro@gmail.com>
9 years agodoc: update broken links in release notes
Andrew Gerrand [Thu, 5 May 2016 20:19:07 +0000 (13:19 -0700)]
doc: update broken links in release notes

Fixes #15559

Change-Id: Ie58650f35e32c1f49669134b62876357abcdc583
Reviewed-on: https://go-review.googlesource.com/22823
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
9 years agocmd/compile: verify imported types after they are fully imported
Robert Griesemer [Thu, 5 May 2016 16:39:50 +0000 (09:39 -0700)]
cmd/compile: verify imported types after they are fully imported

Fixes #15548.

Change-Id: I1dfa9c8739a4b6d5e4c737c1a1e09e80e045b7aa
Reviewed-on: https://go-review.googlesource.com/22803
Reviewed-by: Alan Donovan <adonovan@google.com>
9 years agoruntime: print signal name in panic, if name is known
Emmanuel Odeke [Wed, 4 May 2016 07:42:13 +0000 (01:42 -0600)]
runtime: print signal name in panic, if name is known

Adds a small function signame that infers a signal name
from the signal table, otherwise will fallback to using
hex(sig) as previously. No signal table is present for
Windows hence it will always print the hex value.

Sample code and new result:
```go
package main

import (
  "fmt"
  "time"
)

func main() {
  defer func() {
    if err := recover(); err != nil {
      fmt.Printf("err=%v\n", err)
    }
  }()

  ticker := time.Tick(1e9)
  for {
    <-ticker
  }
}
```

```shell
$ go run main.go &
$ kill -11 <pid>
fatal error: unexpected signal during runtime execution
[signal SIGSEGV: segmentation violation code=0x1 addr=0xb01dfacedebac1e
pc=0xc71db]
...
```

Fixes #13969

Change-Id: Ie6be312eb766661f1cea9afec352b73270f27f9d
Reviewed-on: https://go-review.googlesource.com/22753
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agonet: fix hostLookupOrder("")
Russ Cox [Wed, 6 Jan 2016 19:37:45 +0000 (14:37 -0500)]
net: fix hostLookupOrder("")

Fixes #13623.

Change-Id: I1bd96aa7b6b715e4dbdcf0c37c2d29228df6565c
Reviewed-on: https://go-review.googlesource.com/18329
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agosync/atomic, runtime/internal/atomic: improve ppc64x atomics
Lynn Boger [Thu, 28 Apr 2016 12:16:08 +0000 (07:16 -0500)]
sync/atomic, runtime/internal/atomic: improve ppc64x atomics

The following performance improvements have been made to the
low-level atomic functions for ppc64le & ppc64:

- For those cases containing a lwarx and stwcx (or other sizes):
sync, lwarx, maybe something, stwcx, loop to sync, sync, isync
The sync is moved before (outside) the lwarx/stwcx loop, and the
 sync after is removed, so it becomes:
sync, lwarx, maybe something, stwcx, loop to lwarx, isync

- For the Or8 and And8, the shifting and manipulation of the
address to the word aligned version were removed and the
instructions were changed to use lbarx, stbcx instead of
register shifting, xor, then lwarx, stwcx.

- New instructions LWSYNC, LBAR, STBCC were tested and added.
runtime/atomic_ppc64x.s was changed to use the LWSYNC opcode
instead of the WORD encoding.

Fixes #15469

Ran some of the benchmarks in the runtime and sync directories.
Some results varied from run to run but the trend was improvement
based on best times for base and new:

runtime.test:
BenchmarkChanNonblocking-128         0.88          0.89          +1.14%
BenchmarkChanUncontended-128         569           511           -10.19%
BenchmarkChanContended-128           63110         53231         -15.65%
BenchmarkChanSync-128                691           598           -13.46%
BenchmarkChanSyncWork-128            11355         11649         +2.59%
BenchmarkChanProdCons0-128           2402          2090          -12.99%
BenchmarkChanProdCons10-128          1348          1363          +1.11%
BenchmarkChanProdCons100-128         1002          746           -25.55%
BenchmarkChanProdConsWork0-128       2554          2720          +6.50%
BenchmarkChanProdConsWork10-128      1909          1804          -5.50%
BenchmarkChanProdConsWork100-128     1624          1580          -2.71%
BenchmarkChanCreation-128            237           212           -10.55%
BenchmarkChanSem-128                 705           667           -5.39%
BenchmarkChanPopular-128             5081190       4497566       -11.49%

BenchmarkCreateGoroutines-128             532           473           -11.09%
BenchmarkCreateGoroutinesParallel-128     35.0          34.7          -0.86%
BenchmarkCreateGoroutinesCapture-128      4923          4200          -14.69%

sync.test:
BenchmarkUncontendedSemaphore-128      112           94.2          -15.89%
BenchmarkContendedSemaphore-128        133           128           -3.76%
BenchmarkMutexUncontended-128          1.90          1.67          -12.11%
BenchmarkMutex-128                     353           310           -12.18%
BenchmarkMutexSlack-128                304           283           -6.91%
BenchmarkMutexWork-128                 554           541           -2.35%
BenchmarkMutexWorkSlack-128            567           556           -1.94%
BenchmarkMutexNoSpin-128               275           242           -12.00%
BenchmarkMutexSpin-128                 1129          1030          -8.77%
BenchmarkOnce-128                      1.08          0.96          -11.11%
BenchmarkPool-128                      29.8          27.4          -8.05%
BenchmarkPoolOverflow-128              40564         36583         -9.81%
BenchmarkSemaUncontended-128           3.14          2.63          -16.24%
BenchmarkSemaSyntNonblock-128          1087          1069          -1.66%
BenchmarkSemaSyntBlock-128             897           893           -0.45%
BenchmarkSemaWorkNonblock-128          1034          1028          -0.58%
BenchmarkSemaWorkBlock-128             949           886           -6.64%

Change-Id: I4403fb29d3cd5254b7b1ce87a216bd11b391079e
Reviewed-on: https://go-review.googlesource.com/22549
Reviewed-by: Michael Munday <munday@ca.ibm.com>
Reviewed-by: Minux Ma <minux@golang.org>
9 years agocontext: use https in docs
Shenghou Ma [Thu, 5 May 2016 18:22:34 +0000 (14:22 -0400)]
context: use https in docs

Change-Id: I9354712768702e3b083c77f30165a34cb414d686
Reviewed-on: https://go-review.googlesource.com/22810
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
9 years agocmd/compile: optimize lookupVarOutgoing
Josh Bleecher Snyder [Wed, 4 May 2016 21:04:03 +0000 (14:04 -0700)]
cmd/compile: optimize lookupVarOutgoing

If b has exactly one predecessor, as happens
frequently with static calls, we can make
lookupVarOutgoing generate less garbage.

Instead of generating a value that is just
going to be an OpCopy and then get eliminated,
loop. This can lead to lots of looping.
However, this loop is way cheaper than generating
lots of ssa.Values and then eliminating them.

For a subset of the code in #15537:

Before:

       28.31 real        36.17 user         1.68 sys
2282450944  maximum resident set size

After:

        9.63 real        11.66 user         0.51 sys
 638144512  maximum resident set size

Updates #15537.

Excitingly, it appears that this also helps
regular code:

name       old time/op      new time/op      delta
Template        288ms ± 6%       276ms ± 7%   -4.13%        (p=0.000 n=21+24)
Unicode         143ms ± 8%       141ms ±10%     ~           (p=0.287 n=24+25)
GoTypes         932ms ± 4%       874ms ± 4%   -6.20%        (p=0.000 n=23+22)
Compiler        4.89s ± 4%       4.58s ± 4%   -6.46%        (p=0.000 n=22+23)
MakeBash        40.2s ±13%       39.8s ± 9%     ~           (p=0.648 n=23+23)

name       old user-ns/op   new user-ns/op   delta
Template   388user-ms ±10%  373user-ms ± 5%   -3.80%        (p=0.000 n=24+25)
Unicode    203user-ms ± 6%  202user-ms ± 7%     ~           (p=0.492 n=22+24)
GoTypes    1.29user-s ± 4%  1.17user-s ± 4%   -9.67%        (p=0.000 n=25+23)
Compiler   6.86user-s ± 5%  6.28user-s ± 4%   -8.49%        (p=0.000 n=25+25)

name       old alloc/op     new alloc/op     delta
Template       51.5MB ± 0%      47.6MB ± 0%   -7.47%        (p=0.000 n=22+25)
Unicode        37.2MB ± 0%      37.1MB ± 0%   -0.21%        (p=0.000 n=25+25)
GoTypes         166MB ± 0%       138MB ± 0%  -16.83%        (p=0.000 n=25+25)
Compiler        756MB ± 0%       628MB ± 0%  -16.96%        (p=0.000 n=25+23)

name       old allocs/op    new allocs/op    delta
Template         450k ± 0%        445k ± 0%   -1.02%        (p=0.000 n=25+25)
Unicode          356k ± 0%        356k ± 0%     ~           (p=0.374 n=24+25)
GoTypes         1.31M ± 0%       1.25M ± 0%   -4.18%        (p=0.000 n=25+25)
Compiler        5.29M ± 0%       5.02M ± 0%   -5.15%        (p=0.000 n=25+23)

It also seems to help in other cases in which
phi insertion is a pain point (#14774, #14934).

Change-Id: Ibd05ed7b99d262117ece7bb250dfa8c3d1cc5dd2
Reviewed-on: https://go-review.googlesource.com/22790
Reviewed-by: Keith Randall <khr@golang.org>
Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com>

9 years agocmd/compile/internal/gc: skip TestFP on mips64x
Cherry Zhang [Thu, 5 May 2016 06:47:37 +0000 (23:47 -0700)]
cmd/compile/internal/gc: skip TestFP on mips64x

The legacy mips64 backend doesn't handle large uint->float conversion
correctly. See #15552.

Change-Id: I84ceeaa95cc4e85f09cc46dfb30ab5d151f6b205
Reviewed-on: https://go-review.googlesource.com/22800
Reviewed-by: Minux Ma <minux@golang.org>
9 years agocmd/compile/internal/gc: remove duplicated TestFP
Cherry Zhang [Thu, 5 May 2016 06:42:10 +0000 (23:42 -0700)]
cmd/compile/internal/gc: remove duplicated TestFP

TestFp and TestFP are same, remove one.

Change-Id: Iffdece634cd4572421974496298925e7c6ac13a9
Reviewed-on: https://go-review.googlesource.com/22799
Reviewed-by: Minux Ma <minux@golang.org>
Run-TryBot: Minux Ma <minux@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agocmd/compile: enable constant-time CFG editing
Keith Randall [Thu, 28 Apr 2016 23:52:47 +0000 (16:52 -0700)]
cmd/compile: enable constant-time CFG editing

Provide indexes along with block pointers for Preds
and Succs arrays.  This allows us to splice edges in
and out of those arrays in constant time.

Fixes worst-case O(n^2) behavior in deadcode and fuse.

benchmark                     old ns/op      new ns/op     delta
BenchmarkFuse1-8              2065           2057          -0.39%
BenchmarkFuse10-8             9408           9073          -3.56%
BenchmarkFuse100-8            105238         76277         -27.52%
BenchmarkFuse1000-8           3982562        1026750       -74.22%
BenchmarkFuse10000-8          301220329      12824005      -95.74%
BenchmarkDeadCode1-8          1588           1566          -1.39%
BenchmarkDeadCode10-8         4333           4250          -1.92%
BenchmarkDeadCode100-8        32031          32574         +1.70%
BenchmarkDeadCode1000-8       590407         468275        -20.69%
BenchmarkDeadCode10000-8      17822890       5000818       -71.94%
BenchmarkDeadCode100000-8     1388706640     78021127      -94.38%
BenchmarkDeadCode200000-8     5372518479     168598762     -96.86%

Change-Id: Iccabdbb9343fd1c921ba07bbf673330a1c36ee17
Reviewed-on: https://go-review.googlesource.com/22589
Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
Run-TryBot: Keith Randall <khr@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agoruntime: skip TestCgoCallbackGC on linux/mips64x
Cherry Zhang [Thu, 5 May 2016 05:07:50 +0000 (22:07 -0700)]
runtime: skip TestCgoCallbackGC on linux/mips64x

Builder is too slow. This test passed on builder machines but took
15+ min.

Change-Id: Ief9d67ea47671a57e954e402751043bc1ce09451
Reviewed-on: https://go-review.googlesource.com/22798
Reviewed-by: Minux Ma <minux@golang.org>
Run-TryBot: Minux Ma <minux@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agocmd/link: fix external linker argument for mips64
Cherry Zhang [Thu, 5 May 2016 04:38:45 +0000 (21:38 -0700)]
cmd/link: fix external linker argument for mips64

I overlooked it when rebasing CL 19803.

Change-Id: Ife9d6bcc6a772715d137af903c64bafac0cdb216
Reviewed-on: https://go-review.googlesource.com/22797
Reviewed-by: Minux Ma <minux@golang.org>
9 years agoruntime: put tracebackctxt C functions in .c file
Ian Lance Taylor [Wed, 4 May 2016 20:27:27 +0000 (13:27 -0700)]
runtime: put tracebackctxt C functions in .c file

Since tracebackctxt.go uses //export functions, the C functions can't be
externally visible in the C comment. The code was using attributes to
work around that, but that failed on Windows.

Change-Id: If4449fd8209a8998b4f6855ea89e5db1471b2981
Reviewed-on: https://go-review.googlesource.com/22786
Reviewed-by: Minux Ma <minux@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agodebug/pe: unexport newly introduced identifiers
Alex Brainman [Tue, 3 May 2016 06:37:33 +0000 (16:37 +1000)]
debug/pe: unexport newly introduced identifiers

CLs 22181, 22332 and 22336 intorduced new functionality to be used
in cmd/link (see issue #15345 for details). But we didn't have chance
to use new functionality yet. Unexport newly introduced identifiers,
so we don't have to commit to the API until we actually tried it.

Rename File.COFFSymbols into File._COFFSymbols,
COFFSymbol.FullName into COFFSymbol._FullName,
Section.Relocs into Section._Relocs,
Reloc into _Relocs,
File.StringTable into File._StringTable and
StringTable into _StringTable.

Updates #15345

Change-Id: I770eeb61f855de85e0c175225d5d1c006869b9ec
Reviewed-on: https://go-review.googlesource.com/22720
Reviewed-by: David Crawshaw <crawshaw@golang.org>
Run-TryBot: Alex Brainman <alex.brainman@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agocompress/flate: distinguish between base and min match length.
Nigel Tao [Wed, 4 May 2016 20:43:13 +0000 (06:43 +1000)]
compress/flate: distinguish between base and min match length.

Change-Id: I93db5cd86e3fb568e4444cad95268ba4a02ce8a0
Reviewed-on: https://go-review.googlesource.com/22787
Reviewed-by: Nigel Tao <nigeltao@golang.org>
9 years agocmd/compile: fix uint64 to float casts on ppc64
Michael Munday [Wed, 4 May 2016 18:26:46 +0000 (14:26 -0400)]
cmd/compile: fix uint64 to float casts on ppc64

Adds the FCFIDU instruction and uses it instead of the FCFID
instruction for unsigned integer to float casts. This change means
that unsigned integers do not have to be cast to signed integers
before being cast to a floating point value. Therefore it is no
longer necessary to insert instructions to detect and fix
values that overflow int64.

The previous code generating the uint64 to int64 cast handled
overflow by truncating the uint64 value. This truncation can
change the result of the rounding performed by the integer to
float cast.

The FCFIDU instruction was added in Power ISA 2.06B.

Fixes #15539.

Change-Id: Ia37a9631293eff91032d4cd9a9bec759d2142437
Reviewed-on: https://go-review.googlesource.com/22772
Reviewed-by: David Chase <drchase@google.com>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agocmd/compile: fix debugFormat for new export format
Robert Griesemer [Wed, 4 May 2016 22:13:36 +0000 (15:13 -0700)]
cmd/compile: fix debugFormat for new export format

Change-Id: Ic3415f3ee643636eab4ff7d2351b8ad0dae62895
Reviewed-on: https://go-review.googlesource.com/22792
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
9 years agoruntime: clean up profiling data files produced by TestCgoPprof
Mohit Agarwal [Wed, 4 May 2016 18:34:54 +0000 (00:04 +0530)]
runtime: clean up profiling data files produced by TestCgoPprof

Fixes #15541

Change-Id: I9b6835157db0eb86de13591e785f971ffe754baa
Reviewed-on: https://go-review.googlesource.com/22783
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agonet/http: correct RFC for MethodPatch
Shenghou Ma [Wed, 4 May 2016 21:06:50 +0000 (17:06 -0400)]
net/http: correct RFC for MethodPatch

Fixes #15546.

Change-Id: I39c29ea6999812dd5f1c45f67bddad28f20b6c6b
Reviewed-on: https://go-review.googlesource.com/22773
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
9 years agocmd/dist: remove the use of debug/elf package
Shenghou Ma [Thu, 7 Apr 2016 20:03:53 +0000 (16:03 -0400)]
cmd/dist: remove the use of debug/elf package

debug/elf is only needed to determine the endianness of the host
machine, which is easy to do without debug/elf.

Fixes #15180.

Change-Id: I21035ed3884871270765a1ca3b812a5d4890a7ee
Reviewed-on: https://go-review.googlesource.com/21662
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Run-TryBot: Minux Ma <minux@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agocmd/compile: check that SSA memory args are in the right place
Josh Bleecher Snyder [Wed, 4 May 2016 18:43:42 +0000 (11:43 -0700)]
cmd/compile: check that SSA memory args are in the right place

Fixes #15510

Change-Id: I2e0568778ef90cf29712753b8c42109ef84a0256
Reviewed-on: https://go-review.googlesource.com/22784
Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
9 years agocmd/dist: add flag to build tests only, but not run them
Brad Fitzpatrick [Wed, 4 May 2016 16:08:27 +0000 (09:08 -0700)]
cmd/dist: add flag to build tests only, but not run them

To speed up the ssacheck check builder and make it on by default as a
trybot.

Change-Id: I91a3347491507c84f4878dff744ca426ba3e2e9f
Reviewed-on: https://go-review.googlesource.com/22755
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Andrew Gerrand <adg@golang.org>
9 years agoruntime/internal/atomic: fix vet warnings
Shenghou Ma [Wed, 27 Apr 2016 17:47:41 +0000 (13:47 -0400)]
runtime/internal/atomic: fix vet warnings

Change-Id: Ib29cf7abbbdaed81e918e5e41bca4e9b8da24621
Reviewed-on: https://go-review.googlesource.com/22503
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
9 years agocmd/vet: add mips64x assembly check support
Shenghou Ma [Wed, 27 Apr 2016 17:47:19 +0000 (13:47 -0400)]
cmd/vet: add mips64x assembly check support

Change-Id: I0a6a92604dbfa4b0f9c5ae483b574331f246dcad
Reviewed-on: https://go-review.googlesource.com/22502
Run-TryBot: Minux Ma <minux@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Rob Pike <r@golang.org>
9 years agodoc: update go1.7.txt
Shenghou Ma [Wed, 4 May 2016 16:48:20 +0000 (12:48 -0400)]
doc: update go1.7.txt

Change-Id: I78efbfe0d7e9825126109a568c99f548a099b0fb
Reviewed-on: https://go-review.googlesource.com/22771
Reviewed-by: Minux Ma <minux@golang.org>
9 years agotest: enable fixedbugs/issue10607.go test on linux/mips64x
Cherry Zhang [Thu, 28 Apr 2016 02:18:58 +0000 (22:18 -0400)]
test: enable fixedbugs/issue10607.go test on linux/mips64x

external linking is now supported.

Change-Id: I13e90c39dad86e60781adecdbe8e6bc9e522f740
Reviewed-on: https://go-review.googlesource.com/19811
Reviewed-by: Minux Ma <minux@golang.org>
9 years agocmd/go: enable TestNodeReading on linux/mips64x
Cherry Zhang [Thu, 28 Apr 2016 02:18:48 +0000 (22:18 -0400)]
cmd/go: enable TestNodeReading on linux/mips64x

external linking is now supported.

Change-Id: I3f552f5f09391205fced509fe8a5a38297ea8153
Reviewed-on: https://go-review.googlesource.com/19810
Reviewed-by: Minux Ma <minux@golang.org>
9 years agocmd/dist: enable cgo and external linking on linux/mips64x
Cherry Zhang [Thu, 28 Apr 2016 02:18:44 +0000 (22:18 -0400)]
cmd/dist: enable cgo and external linking on linux/mips64x

Fixes #14126

Change-Id: I21c8e06c01d3ef02ee09dc185d4443e2da8fd52b
Reviewed-on: https://go-review.googlesource.com/19809
Reviewed-by: Minux Ma <minux@golang.org>
9 years agomisc/cgo/test: add mips64x test case for issue9400
Cherry Zhang [Thu, 28 Apr 2016 02:18:38 +0000 (22:18 -0400)]
misc/cgo/test: add mips64x test case for issue9400

Change-Id: If2b4abb6ff322c20e35de025298c8e5ab53edd42
Reviewed-on: https://go-review.googlesource.com/19808
Reviewed-by: Minux Ma <minux@golang.org>
9 years agocmd/link/internal/ld: force external linking for mips64x with cgo
Cherry Zhang [Thu, 28 Apr 2016 02:18:33 +0000 (22:18 -0400)]
cmd/link/internal/ld: force external linking for mips64x with cgo

cgo internal linking is not supported yet (issue #14449).

Change-Id: Ic968916383d77b7f449db8f230c928a1e81939e0
Reviewed-on: https://go-review.googlesource.com/19807
Reviewed-by: Minux Ma <minux@golang.org>
9 years agoruntime: add linux/mips64x cgo support
Cherry Zhang [Thu, 28 Apr 2016 02:18:24 +0000 (22:18 -0400)]
runtime: add linux/mips64x cgo support

Change-Id: Id40dd05b7b264f3b779fdf9ccc2421ba4bc70589
Reviewed-on: https://go-review.googlesource.com/19806
Reviewed-by: Minux Ma <minux@golang.org>
9 years agoruntime/cgo: add context argument to crosscall2 on mips64
Cherry Zhang [Tue, 3 May 2016 20:49:54 +0000 (13:49 -0700)]
runtime/cgo: add context argument to crosscall2 on mips64

Change-Id: Id018516075842afd8af12fbf207763a851d5a851
Reviewed-on: https://go-review.googlesource.com/22754
Reviewed-by: Ian Lance Taylor <iant@golang.org>
9 years agomisc/cgo/fortran: fix gfortran compile test
Martin Möhrmann [Tue, 3 May 2016 09:10:26 +0000 (11:10 +0200)]
misc/cgo/fortran: fix gfortran compile test

Fixes #14544

Change-Id: I58b0b164ebbfeafe4ab32039a063df53e3018a6d
Reviewed-on: https://go-review.googlesource.com/22730
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Reviewed-by: Sean Lake <odysseus9672@gmail.com>
9 years agoarchive/zip: pool flate readers
Niko Dziemba [Wed, 10 Feb 2016 14:57:16 +0000 (15:57 +0100)]
archive/zip: pool flate readers

Similar to the flate Writer pools already used,
this adds pooling for flate Readers.

compress/flate allows re-using of Readers, see
https://codereview.appspot.com/97140043/

In a real-world scenario when reading ~ 500 small files from a ZIP
archive this gives a speedup of 1.5x-2x.

Fixes #14289

Change-Id: I2d98ad983e95ab7d97e06fd0145f619b4f47caa4
Reviewed-on: https://go-review.googlesource.com/19416
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
9 years agocmd/link: always read type data for dynimport symbols
Michael Hudson-Doyle [Tue, 3 May 2016 23:23:24 +0000 (11:23 +1200)]
cmd/link: always read type data for dynimport symbols

Consider three shared libraries:

 libBase.so -- defines a type T
 lib2.so    -- references type T
 lib3.so    -- also references type T, and something from lib2

lib2.so will contain a type symbol for T in its symbol table, but no
definition. If, when linking lib3.so the linker reads the symbols from lib2.so
before libBase.so, the linker didn't read the type data and later crashed.

The fix is trivial but the test change is a bit messy because the order the
linker reads the shared libraries in ends up depending on the order of the
import statements in the file so I had to rename one of the test packages so
that gofmt doesn't fix the test by accident...

Fixes #15516

Change-Id: I124b058f782c900a3a54c15ed66a0d91d0cde5ce
Reviewed-on: https://go-review.googlesource.com/22744
Run-TryBot: Michael Hudson-Doyle <michael.hudson@canonical.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
9 years agocmd/api: remove debug/pe replated changes from next.txt
Alex Brainman [Wed, 4 May 2016 02:20:14 +0000 (12:20 +1000)]
cmd/api: remove debug/pe replated changes from next.txt

See CL 22720 for details.

Updates #15345

Change-Id: If93ddbb8137d57da9846b671160b4cebe1992570
Reviewed-on: https://go-review.googlesource.com/22752
Reviewed-by: David Crawshaw <crawshaw@golang.org>
9 years agocmd/compile: use SSA tests on legacy compiler
Keith Randall [Wed, 4 May 2016 00:21:36 +0000 (17:21 -0700)]
cmd/compile: use SSA tests on legacy compiler

Why not?  Because the 386 backend can't handle one of them.
But other than that, it should work.

Change-Id: Iaeb9735f8c3c281136a0734376dec5ddba21be3b
Reviewed-on: https://go-review.googlesource.com/22748
Run-TryBot: Keith Randall <khr@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: David Chase <drchase@google.com>
9 years agonet: allow netgo to use lookup from nsswitch.conf
Vishvananda Ishaya [Wed, 17 Feb 2016 01:58:11 +0000 (17:58 -0800)]
net: allow netgo to use lookup from nsswitch.conf

Change https://golang.org/cl/8945 allowed Go to use its own DNS resolver
instead of libc in a number of cases. The code parses nsswitch.conf and
attempts to resolve things in the same order. Unfortunately, builds with
netgo completely ignore this parsing and always search via
hostLookupFilesDNS.

This commit modifies the logic to allow binaries built with netgo to
parse nsswitch.conf and attempt to resolve using the order specified
there. If the parsing results in hostLookupCGo, it falls back to the
original hostLookupFilesDNS. Tests are also added to ensure that both
the parsing and the fallback work properly.

Fixes #14354

Change-Id: Ib079ad03d7036a4ec57f18352a15ba55d933f261
Reviewed-on: https://go-review.googlesource.com/19523
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
9 years agoruntime: use cgo traceback for SIGPROF
Ian Lance Taylor [Fri, 29 Apr 2016 22:20:27 +0000 (15:20 -0700)]
runtime: use cgo traceback for SIGPROF

If we collected a cgo traceback when entering the SIGPROF signal
handler, record it as part of the profiling stack trace.

This serves as the promised test for https://golang.org/cl/21055 .

Change-Id: I5f60cd6cea1d9b7c3932211483a6bfab60ed21d2
Reviewed-on: https://go-review.googlesource.com/22650
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Austin Clements <austin@google.com>
9 years agocmd/vet: fix test's dependence on perl
Brad Fitzpatrick [Tue, 3 May 2016 23:33:32 +0000 (23:33 +0000)]
cmd/vet: fix test's dependence on perl

Change-Id: I774dbd4f90ef271a0969c3c8e65d145669312e3e
Reviewed-on: https://go-review.googlesource.com/22745
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
Reviewed-by: Ross Light <light@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agocmd/compile: never CSE two memories
Keith Randall [Tue, 3 May 2016 20:58:28 +0000 (13:58 -0700)]
cmd/compile: never CSE two memories

It never makes sense to CSE two ops that generate memory.
We might as well start those ops off in their own partition.

Fixes #15520

Change-Id: I0091ed51640f2c10cd0117f290b034dde7a86721
Reviewed-on: https://go-review.googlesource.com/22741
Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
Run-TryBot: Keith Randall <khr@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agostrings, bytes: fix Reader 0 byte read at EOF
Brad Fitzpatrick [Tue, 3 May 2016 19:54:49 +0000 (19:54 +0000)]
strings, bytes: fix Reader 0 byte read at EOF

0 byte reads at EOF weren't returning EOF.

Change-Id: I19b5fd5a72e83d49566a230ce4067be03f00d14b
Reviewed-on: https://go-review.googlesource.com/22740
Reviewed-by: Bryan Mills <bcmills@google.com>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agocmd/compile: use correct packages when exporting/importing _ (blank) names
Robert Griesemer [Tue, 3 May 2016 00:03:36 +0000 (17:03 -0700)]
cmd/compile: use correct packages when exporting/importing _ (blank) names

1) Blank parameters cannot be accessed so the package doesn't matter.
   Do not export it, and consistently use localpkg when importing a
   blank parameter.

2) More accurately replicate fmt.go and parser.go logic when importing
   a blank struct field. Blank struct fields get exported without
   package qualification.
   (This is actually incorrect, even with the old textual export format,
   but we will fix that in a separate change. See also issue 15514.)

Fixes #15491.

Change-Id: I7978e8de163eb9965964942aee27f13bf94a7c3c
Reviewed-on: https://go-review.googlesource.com/22714
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
9 years agocmd/trace: make binary argument optional
Dmitry Vyukov [Sun, 24 Apr 2016 11:33:33 +0000 (13:33 +0200)]
cmd/trace: make binary argument optional

1.7 traces embed symbol info and we now generate symbolized pprof profiles,
so we don't need the binary. Make binary argument optional as 1.5 traces
still need it.

Change-Id: I65eb13e3d20ec765acf85c42d42a8d7aae09854c
Reviewed-on: https://go-review.googlesource.com/22410
Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
Reviewed-by: Austin Clements <austin@google.com>
9 years agoruntime: per-P contexts for race detector
Dmitry Vyukov [Fri, 26 Feb 2016 20:57:16 +0000 (21:57 +0100)]
runtime: per-P contexts for race detector

Race runtime also needs local malloc caches and currently uses
a mix of per-OS-thread and per-goroutine caches. This leads to
increased memory consumption. But more importantly cache of
synchronization objects is per-goroutine and we don't always
have goroutine context when feeing memory in GC. As the result
synchronization object descriptors leak (more precisely, they
can be reused if another synchronization object is recreated
at the same address, but it does not always help). For example,
the added BenchmarkSyncLeak has effectively runaway memory
consumption (based on a real long running server).

This change updates race runtime with support for per-P contexts.
BenchmarkSyncLeak now stabilizes at ~1GB memory consumption.

Long term, this will allow us to remove race runtime dependency
on glibc (as malloc is the main cornerstone).

I've also implemented a different scheme to pass P context to
race runtime: scheduler notified race runtime about association
between G and P by calling procwire(g, p)/procunwire(g, p).
But it turned out to be very messy as we have lots of places
where the association changes (e.g. syscalls). So I dropped it
in favor of the current scheme: race runtime asks scheduler
about the current P.

Fixes #14533

Change-Id: Iad10d2f816a44affae1b9fed446b3580eafd8c69
Reviewed-on: https://go-review.googlesource.com/19970
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Dmitry Vyukov <dvyukov@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agoruntime: fix CPU underutilization
Dmitry Vyukov [Fri, 18 Mar 2016 15:34:11 +0000 (16:34 +0100)]
runtime: fix CPU underutilization

Runqempty is a critical predicate for scheduler. If runqempty spuriously
returns true, then scheduler can fail to schedule arbitrary number of
runnable goroutines on idle Ps for arbitrary long time. With the addition
of runnext runqempty predicate become broken (can spuriously return true).
Consider that runnext is not nil and the main array is empty. Runqempty
observes that the array is empty, then it is descheduled for some time.
Then queue owner pushes another element to the queue evicting runnext
into the array. Then queue owner pops runnext. Then runqempty resumes
and observes runnext is nil and returns true. But there were no point
in time when the queue was empty.

Fix runqempty predicate to not return true spuriously.

Change-Id: Ifb7d75a699101f3ff753c4ce7c983cf08befd31e
Reviewed-on: https://go-review.googlesource.com/20858
Reviewed-by: Austin Clements <austin@google.com>
Run-TryBot: Dmitry Vyukov <dvyukov@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agocmd/cgo: an approach to tsan that works with gcc
Michael Hudson-Doyle [Mon, 2 May 2016 02:46:40 +0000 (14:46 +1200)]
cmd/cgo: an approach to tsan that works with gcc

GCC, unlike clang, does not provide any way for code being compiled to tell if
-fsanitize-thread was passed. But cgo can look to see if that flag is being
passed and generate different code in that case.

Fixes #14602

Change-Id: I86cb5318c2e35501ae399618c05af461d1252d2d
Reviewed-on: https://go-review.googlesource.com/22688
Run-TryBot: Michael Hudson-Doyle <michael.hudson@canonical.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
9 years agonet/http: keep idle conns sorted by usage
Brad Fitzpatrick [Mon, 2 May 2016 22:22:25 +0000 (22:22 +0000)]
net/http: keep idle conns sorted by usage

Addressing feedback from Alan Su in https://golang.org/cl/22655

Change-Id: Ie0724efea2b4da67503c074e265ec7f8d7de7791
Reviewed-on: https://go-review.googlesource.com/22709
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
9 years agocmd/link: bump object file version number
David Crawshaw [Mon, 2 May 2016 21:32:14 +0000 (17:32 -0400)]
cmd/link: bump object file version number

The format has been tweaked several times in the latest cycle, so
replace go13ld with go17ld.

Change-Id: I343c49b02b7516fd781bc96ad46640579da68c59
Reviewed-on: https://go-review.googlesource.com/22708
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agonet/http: fix typo in comment in main_test
Emmanuel Odeke [Sun, 1 May 2016 23:18:18 +0000 (17:18 -0600)]
net/http: fix typo in comment in main_test

Change-Id: I22d4b5a0d5c146a65d4ef77a32e23f7780ba1d95
Reviewed-on: https://go-review.googlesource.com/22684
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
9 years agonet/http: remove some TODOs
Brad Fitzpatrick [Mon, 2 May 2016 16:07:18 +0000 (16:07 +0000)]
net/http: remove some TODOs

Change-Id: Iaf200ba9a308bc8f511eec4a70dbeb014bf5fdc3
Reviewed-on: https://go-review.googlesource.com/22690
Reviewed-by: Andrew Gerrand <adg@golang.org>
9 years agoall: make copyright headers consistent with one space after period
Emmanuel Odeke [Sun, 10 Apr 2016 21:32:26 +0000 (14:32 -0700)]
all: make copyright headers consistent with one space after period

Follows suit with https://go-review.googlesource.com/#/c/20111.

Generated by running
$ grep -R 'Go Authors.  All' * | cut -d":" -f1 | while read F;do perl -pi -e 's/Go
Authors.  All/Go Authors. All/g' $F;done

The code in cmd/internal/unvendor wasn't changed.

Fixes #15213

Change-Id: I4f235cee0a62ec435f9e8540a1ec08ae03b1a75f
Reviewed-on: https://go-review.googlesource.com/21819
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agocmd/cgo, misc/cgo/test: make -Wdeclaration-after-statement clean
Ian Lance Taylor [Mon, 2 May 2016 00:03:46 +0000 (17:03 -0700)]
cmd/cgo, misc/cgo/test: make -Wdeclaration-after-statement clean

I got a complaint that cgo output triggers warnings with
-Wdeclaration-after-statement.  I don't think it's worth testing for
this--C has permitted declarations after statements since C99--but it is
easy enough to fix.  It may break again; so it goes.

This CL also fixes errno handling to avoid getting confused if the tsan
functions happen to change the global errno variable.

Change-Id: I0ec7c63a6be5653ef44799d134c8d27cb5efa441
Reviewed-on: https://go-review.googlesource.com/22686
Run-TryBot: Ian Lance Taylor <iant@golang.org>
Reviewed-by: Minux Ma <minux@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agocmd/link: pass -Wl,-z,relro to host linker in -buildmode=PIE
Michael Hudson-Doyle [Mon, 2 May 2016 01:58:48 +0000 (13:58 +1200)]
cmd/link: pass -Wl,-z,relro to host linker in -buildmode=PIE

Fixes #15485

Change-Id: I8e9314be91db89873130b232b589a284822e6643
Reviewed-on: https://go-review.googlesource.com/22687
Run-TryBot: Michael Hudson-Doyle <michael.hudson@canonical.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
9 years agocmd/compile: don't SSA any variables when -N
Keith Randall [Fri, 29 Apr 2016 19:09:32 +0000 (12:09 -0700)]
cmd/compile: don't SSA any variables when -N

Turn SSAing of variables off when compiling with optimizations off.
This helps keep variable names around that would otherwise be
optimized away.

Fixes #14744

Change-Id: I31db8cf269c068c7c5851808f13e5955a09810ca
Reviewed-on: https://go-review.googlesource.com/22681
Run-TryBot: Keith Randall <khr@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: David Chase <drchase@google.com>
9 years agonet/http: gofmt -w -s
Mikio Hara [Sun, 1 May 2016 19:44:46 +0000 (04:44 +0900)]
net/http: gofmt -w -s

Change-Id: I7e07888e90c7449f119e74b97995efcd7feef76e
Reviewed-on: https://go-review.googlesource.com/22682
Reviewed-by: Ian Lance Taylor <iant@golang.org>
9 years agocmd/compile: fix reslice
Keith Randall [Sun, 1 May 2016 05:28:37 +0000 (22:28 -0700)]
cmd/compile: fix reslice

:= is the wrong thing here.  The new variable masks the old
variable so we allocate the slice afresh each time around the loop.

Change-Id: I759c30e1bfa88f40decca6dd7d1e051e14ca0844
Reviewed-on: https://go-review.googlesource.com/22679
Run-TryBot: Keith Randall <khr@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Minux Ma <minux@golang.org>
9 years agonet/http: fix typo in comment
Brad Fitzpatrick [Sun, 1 May 2016 15:18:13 +0000 (15:18 +0000)]
net/http: fix typo in comment

Change-Id: I753e62879a56582a9511e3f34fdeac929202efbf
Reviewed-on: https://go-review.googlesource.com/22680
Reviewed-by: Ralph Corderoy <ralph@inputplus.co.uk>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
9 years agonet/http: add Response.Uncompressed bool
Brad Fitzpatrick [Sat, 16 Apr 2016 16:35:32 +0000 (09:35 -0700)]
net/http: add Response.Uncompressed bool

The Transport's automatic gzip uncompression lost information in the
process (the compressed Content-Length, if known). Normally that's
okay, but it's not okay for reverse proxies which have to be able to
generate a valid HTTP response from the Transport's provided
*Response.

Reverse proxies should normally be disabling compression anyway and
just piping the compressed pipes though and not wasting CPU cycles
decompressing them. So also document that on the new Uncompressed
field.

Then, using the new field, fix Response.Write to not inject a bogus
"Connection: close" header when it doesn't see a transfer encoding or
content-length.

Updates #15366 (the http2 side remains, once this is submitted)

Change-Id: I476f40aa14cfa7aa7b3bf99021bebba4639f9640
Reviewed-on: https://go-review.googlesource.com/22671
Reviewed-by: Andrew Gerrand <adg@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agonet/http: provide access to the listener address an HTTP request arrived on
Brad Fitzpatrick [Sun, 1 May 2016 04:11:26 +0000 (21:11 -0700)]
net/http: provide access to the listener address an HTTP request arrived on

This adds a context key named LocalAddrContextKey (for now, see #15229) to
let users access the net.Addr of the net.Listener that accepted the connection
that sent an HTTP request. This is similar to ServerContextKey which provides
access to the *Server. (A Server may have multiple Listeners)

Fixes #6732

Change-Id: I74296307b68aaaab8df7ad4a143e11b5227b5e62
Reviewed-on: https://go-review.googlesource.com/22672
Reviewed-by: Andrew Gerrand <adg@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agonet/http: add Transport.IdleConnTimeout
Brad Fitzpatrick [Sun, 1 May 2016 02:11:42 +0000 (21:11 -0500)]
net/http: add Transport.IdleConnTimeout

Don't keep idle HTTP client connections open forever. Add a new knob,
Transport.IdleConnTimeout, and make the default be 90 seconds. I
figure 90 seconds is more than a minute, and less than infinite, and I
figure enough code has things waking up once a minute polling APIs.

This also removes the Transport's idleCount field which was unused and
redundant with the size of the idleLRU map (which was actually used).

Change-Id: Ibb698a9a9a26f28e00a20fe7ed23f4afb20c2322
Reviewed-on: https://go-review.googlesource.com/22670
Reviewed-by: Andrew Gerrand <adg@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agonet/http: fix a few crashes with a ClientTrace with nil funcs
Brad Fitzpatrick [Sun, 1 May 2016 02:27:04 +0000 (21:27 -0500)]
net/http: fix a few crashes with a ClientTrace with nil funcs

And add a test.

Updates #12580

Change-Id: Ia7eaba09b8e7fd0eddbcaefb948d01ab10af876e
Reviewed-on: https://go-review.googlesource.com/22659
Reviewed-by: Andrew Gerrand <adg@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agonet/http: document some errors more, mark ErrWriteAfterFlush as unused
Brad Fitzpatrick [Sun, 1 May 2016 03:11:34 +0000 (20:11 -0700)]
net/http: document some errors more, mark ErrWriteAfterFlush as unused

Fixes #15150

Change-Id: I1a892d5b0516a37dac050d3bb448e0a2571db16e
Reviewed-on: https://go-review.googlesource.com/22658
Reviewed-by: Andrew Gerrand <adg@golang.org>
9 years agoarchive/zip: improve BenchmarkCompressedZipGarbage
Josh Bleecher Snyder [Thu, 28 Apr 2016 22:01:41 +0000 (15:01 -0700)]
archive/zip: improve BenchmarkCompressedZipGarbage

Before this CL:

$ go test -bench=CompressedZipGarbage -count=5 -run=NONE archive/zip
BenchmarkCompressedZipGarbage-8        50  20677087 ns/op   42973 B/op      47 allocs/op
BenchmarkCompressedZipGarbage-8       100  20584764 ns/op   24294 B/op      47 allocs/op
BenchmarkCompressedZipGarbage-8        50  20859221 ns/op   42973 B/op      47 allocs/op
BenchmarkCompressedZipGarbage-8       100  20901176 ns/op   24294 B/op      47 allocs/op
BenchmarkCompressedZipGarbage-8        50  21282409 ns/op   42973 B/op      47 allocs/op

The B/op number is effectively meaningless. There
is a surprisingly large one-time cost that gets
divided by the number of iterations that your
machine can get through in a second.

This CL discards the first run, which helps.
It is not a panacea. Running with -benchtime=10s
will allow the sync.Pool to be emptied,
which brings the problem back.
However, since there are more iterations to divide
the cost through, it’s not quite as bad,
and running with a high benchtime is rare.

This CL changes the meaning of the B/op number,
which is unfortunate, since it won’t have the
same order of magnitude as previous Go versions.
But it wasn’t really comparable before anyway,
since it didn’t have any reliable meaning at all.

After this CL:

$ go test -bench=CompressedZipGarbage -count=5 -run=NONE archive/zip
BenchmarkCompressedZipGarbage-8         100   20881890 ns/op     5616 B/op       47 allocs/op
BenchmarkCompressedZipGarbage-8          50   20622757 ns/op     5616 B/op       47 allocs/op
BenchmarkCompressedZipGarbage-8          50   20628193 ns/op     5616 B/op       47 allocs/op
BenchmarkCompressedZipGarbage-8         100   20756612 ns/op     5616 B/op       47 allocs/op
BenchmarkCompressedZipGarbage-8         100   20639774 ns/op     5616 B/op       47 allocs/op

Change-Id: Iedee04f39328974c7fa272a6113d423e7ffce50f
Reviewed-on: https://go-review.googlesource.com/22585
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
9 years agodoc: update go1.7.txt
Brad Fitzpatrick [Thu, 28 Apr 2016 18:51:12 +0000 (13:51 -0500)]
doc: update go1.7.txt

Change-Id: I53dd5affc3a1e1f741fe44c7ce691bb2cd432764
Reviewed-on: https://go-review.googlesource.com/22657
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
9 years agocmd/internal/obj/mips, cmd/link: add support TLS relocation for mips64x
Cherry Zhang [Thu, 28 Apr 2016 02:18:14 +0000 (22:18 -0400)]
cmd/internal/obj/mips, cmd/link: add support TLS relocation for mips64x

a new relocation R_ADDRMIPSTLS is added, which resolves to 16-bit offset
of a TLS address on mips64x.

Change-Id: Ic60d0e1ba49ff1c433cead242f5884677ab227a5
Reviewed-on: https://go-review.googlesource.com/19804
Reviewed-by: Minux Ma <minux@golang.org>