]> Cypherpunks repositories - gostls13.git/log
gostls13.git
5 years agoruntime: don't save/restore FP registers in softfloat mode on MIPS(64)
Cherry Zhang [Wed, 4 Mar 2020 16:14:53 +0000 (11:14 -0500)]
runtime: don't save/restore FP registers in softfloat mode on MIPS(64)

Fixes #37653.

Change-Id: I6188e44b4bc4aba7b56f29d9ce9de4618c70fd7b
Reviewed-on: https://go-review.googlesource.com/c/go/+/222057
Run-TryBot: Cherry Zhang <cherryyz@google.com>
Reviewed-by: Keith Randall <khr@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

5 years agobytes, strings: moves indexRabinKarp function to internal/bytealg
erifan01 [Thu, 21 Nov 2019 06:38:25 +0000 (14:38 +0800)]
bytes, strings: moves indexRabinKarp function to internal/bytealg

In order to facilitate optimization of IndexAny and LastIndexAny, this patch moves
three Rabin-Karp related functions indexRabinKarp, hashStr and hashStrRev in strings
package to initernal/bytealg. There are also three functions in the bytes package with
the same names and functions but different parameter types. To highlight this, this
patch also moves them to internal/bytealg and gives them slightly different names.

Related benchmark changes on amd64 and arm64:

name          old time/op    new time/op    delta
pkg:strings goos:linux goarch:amd64
Index-16        14.0ns ± 1%    14.1ns ± 2%    ~     (p=0.738 n=5+5)
LastIndex-16    15.5ns ± 1%    15.7ns ± 4%    ~     (p=0.897 n=5+5)
pkg:bytes goos:linux goarch:amd64
Index/10-16     26.5ns ± 1%    26.5ns ± 0%    ~     (p=0.873 n=5+5)
Index/32-16     26.2ns ± 0%    25.7ns ± 0%  -1.68%  (p=0.008 n=5+5)
Index/4K-16     5.12µs ± 4%    5.14µs ± 2%    ~     (p=0.841 n=5+5)
Index/4M-16     5.44ms ± 3%    5.34ms ± 2%    ~     (p=0.056 n=5+5)
Index/64M-16    85.8ms ± 3%    84.6ms ± 0%  -1.37%  (p=0.016 n=5+5)

name          old speed      new speed      delta
pkg:bytes goos:linux goarch:amd64
Index/10-16    377MB/s ± 1%   377MB/s ± 0%    ~     (p=1.000 n=5+5)
Index/32-16   1.22GB/s ± 1%  1.24GB/s ± 0%  +1.66%  (p=0.008 n=5+5)
Index/4K-16    800MB/s ± 4%   797MB/s ± 2%    ~     (p=0.841 n=5+5)
Index/4M-16    771MB/s ± 3%   786MB/s ± 2%    ~     (p=0.056 n=5+5)
Index/64M-16   783MB/s ± 3%   793MB/s ± 0%  +1.36%  (p=0.016 n=5+5)

name         old time/op   new time/op   delta
pkg:strings goos:linux goarch:arm64
Index-8       22.6ns ± 0%   22.5ns ± 0%    ~     (p=0.167 n=5+5)
LastIndex-8   17.5ns ± 0%   17.5ns ± 0%    ~     (all equal)
pkg:bytes goos:linux goarch:arm64
Index/10-8    25.0ns ± 0%   25.0ns ± 0%    ~     (all equal)
Index/32-8     160ns ± 0%    160ns ± 0%    ~     (all equal)
Index/4K-8    6.26µs ± 0%   6.26µs ± 0%    ~     (p=0.167 n=5+5)
Index/4M-8    6.30ms ± 0%   6.31ms ± 0%    ~     (p=1.000 n=5+5)
Index/64M-8    101ms ± 0%    101ms ± 0%    ~     (p=0.690 n=5+5)

name         old speed     new speed     delta
pkg:bytes goos:linux goarch:arm64
Index/10-8   399MB/s ± 0%  400MB/s ± 0%  +0.08%  (p=0.008 n=5+5)
Index/32-8   200MB/s ± 0%  200MB/s ± 0%    ~     (p=0.127 n=4+5)
Index/4K-8   654MB/s ± 0%  654MB/s ± 0%  +0.01%  (p=0.016 n=5+5)
Index/4M-8   665MB/s ± 0%  665MB/s ± 0%    ~     (p=0.833 n=5+5)
Index/64M-8  665MB/s ± 0%  665MB/s ± 0%    ~     (p=0.913 n=5+5)

Change-Id: Icce3bc162bb8613ac36dc963a46c51f8e82ab842
Reviewed-on: https://go-review.googlesource.com/c/go/+/208638
Run-TryBot: eric fang <eric.fang@arm.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
5 years agomisc/cgo/test: fix sigaltstack test on AIX
Clément Chigot [Tue, 3 Mar 2020 15:24:32 +0000 (16:24 +0100)]
misc/cgo/test: fix sigaltstack test on AIX

Increase the size of the signal stack as the value given by SIGSTKSZ
is too small for the Go signal handler.

Fixes #37609

Change-Id: I56f1006bc69a2a9fb43f9e0da00061964290a690
Reviewed-on: https://go-review.googlesource.com/c/go/+/221804
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Reviewed-by: Bryan C. Mills <bcmills@google.com>
5 years agocmd/compile: don't allow NaNs in floating-point constant ops
Keith Randall [Tue, 3 Mar 2020 17:56:20 +0000 (17:56 +0000)]
cmd/compile: don't allow NaNs in floating-point constant ops

Trying this CL again, with a fixed test that allows platforms
to disagree on the exact behavior of converting NaNs.

We store 32-bit floating point constants in a 64-bit field, by
converting that 32-bit float to 64-bit float to store it, and convert
it back to use it.

That works for *almost* all floating-point constants. The exception is
signaling NaNs. The round trip described above means we can't represent
a 32-bit signaling NaN, because conversions strip the signaling bit.

To fix this issue, just forbid NaNs as floating-point constants in SSA
form. This shouldn't affect any real-world code, as people seldom
constant-propagate NaNs (except in test code).

Additionally, NaNs are somewhat underspecified (which of the many NaNs
do you get when dividing 0/0?), so when cross-compiling there's a
danger of using the compiler machine's NaN regime for some math, and
the target machine's NaN regime for other math. Better to use the
target machine's NaN regime always.

Update #36400

Change-Id: Idf203b688a15abceabbd66ba290d4e9f63619ecb
Reviewed-on: https://go-review.googlesource.com/c/go/+/221790
Run-TryBot: Keith Randall <khr@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
5 years agocmd/compile: remove walkinrange optimization
Michael Munday [Tue, 3 Mar 2020 12:44:19 +0000 (12:44 +0000)]
cmd/compile: remove walkinrange optimization

The walkinrange optimization has been superseded by CL 165998.

Has a small positive impact on binary sizes:

compilecmp master -> HEAD
master (e37cc29863): cmd/compile: optimize integer-in-range checks
HEAD (1a70680a34): cmd/compile: remove walkinrange optimization
platform: linux/amd64

file      before    after     Δ       %
addr2line 4329144   4325048   -4096   -0.095%
api       6060970   6056874   -4096   -0.068%
asm       5196905   5192809   -4096   -0.079%
cgo       4898769   4890577   -8192   -0.167%
compile   20222193  20209713  -12480  -0.062%
cover     5331580   5323388   -8192   -0.154%
dist      3732778   3728682   -4096   -0.110%
doc       4748488   4740296   -8192   -0.173%
link      6707380   6695092   -12288  -0.183%
nm        4278685   4274589   -4096   -0.096%
pack      2305038   2300942   -4096   -0.178%
pprof     14874834  14870738  -4096   -0.028%
test2json 2849221   2845125   -4096   -0.144%
vet       8393173   8384981   -8192   -0.098%
go        15205572  15193284  -12288  -0.081%
total     131812292 131709700 -102592 -0.078%

Updates #30645.

Change-Id: I42d74481652c90fef1a9bc58c70836e42c9b1c4b
Reviewed-on: https://go-review.googlesource.com/c/go/+/221802
Run-TryBot: Michael Munday <mike.munday@ibm.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
5 years agocmd/compile: simplify converted SSA form for 'if false'
Josh Bleecher Snyder [Mon, 2 Mar 2020 21:51:20 +0000 (13:51 -0800)]
cmd/compile: simplify converted SSA form for 'if false'

The goal here is to make it easier for a human to
examine the SSA when a function contains lots of dead code.

No significant compiler metric or generated code differences.

Change-Id: I81915fa4639bc8820cc9a5e45e526687d0d1f57a
Reviewed-on: https://go-review.googlesource.com/c/go/+/221791
Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
5 years agoall: fix two minor typos in comments
Josh Bleecher Snyder [Tue, 9 May 2017 19:48:23 +0000 (12:48 -0700)]
all: fix two minor typos in comments

Change-Id: Iec6cd81c9787d3419850aa97e75052956ad139bc
Reviewed-on: https://go-review.googlesource.com/c/go/+/221789
Reviewed-by: Emmanuel Odeke <emm.odeke@gmail.com>
5 years agonet/http: fix handling of HTTP/2 upgrade failures
Russ Cox [Tue, 25 Feb 2020 18:01:59 +0000 (13:01 -0500)]
net/http: fix handling of HTTP/2 upgrade failures

If an error occurs during the HTTP/2 upgrade phase, originally this
resulted in a pconn with pconn.alt set to an http2erringRoundTripper,
which always fails. This is not wanted - we want to retry in this case.

CL 202078 added a check for the http2erringRoundTripper to treat it
as a failed pconn, but the handling of the failure was wrong in the case
where the pconn is not in the idle list at all (common in HTTP/2).
This made the added test TestDontCacheBrokenHTTP2Conn flaky.

CL 218097 (unsubmitted) proposed to expand the handling of the
http2erringRoundTripper after the new check, to dispose of the pconn
more thoroughly. Bryan Mills pointed out in that review that we probably
shouldn't make the never-going-to-work pconn in the first place.

This CL changes the upgrade phase look for the http2erringRoundTripper
and return the underlying error instead of claiming to have a working
connection. Having done that, the CL undoes the change in CL 202078
and with it the need for CL 218097, but it keeps the new test added
by CL 202078.

On my laptop, before this commit, TestDontCacheBrokenHTTP2Conn
failed 66 times out of 20,000. With this commit, I see 0 out of 20,000.

Fixes #34978.
Fixes #35113.

Change-Id: Ibd908b63c2ae96e159e8e604213d8373afb350e3
Reviewed-on: https://go-review.googlesource.com/c/go/+/220905
Reviewed-by: Bryan C. Mills <bcmills@google.com>
Run-TryBot: Russ Cox <rsc@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

5 years agoencoding/hex: remove unused variable from BenchmarkDump
Tim Cooper [Tue, 3 Mar 2020 13:08:06 +0000 (07:08 -0600)]
encoding/hex: remove unused variable from BenchmarkDump

Change-Id: I1fd47e5eab27346cec488098d4f6102a0749bd28
Reviewed-on: https://go-review.googlesource.com/c/go/+/221788
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
5 years agocmd/compile: optimize integer-in-range checks
Michael Munday [Mon, 20 May 2019 18:55:56 +0000 (11:55 -0700)]
cmd/compile: optimize integer-in-range checks

This CL incorporates code from CL 201206 by Josh Bleecher Snyder
(thanks Josh).

This CL restores the integer-in-range optimizations in the SSA
backend. The fuse pass is enhanced to detect inequalities that
could be merged and fuse their associated blocks while the generic
rules optimize them into a single unsigned comparison.

For example, the inequality `x >= 0 && x < 10` will now be optimized
to `unsigned(x) < 10`.

Overall has a fairly positive impact on binary sizes.

name                      old time/op       new time/op       delta
Template                        192ms ± 1%        192ms ± 1%    ~     (p=0.757 n=17+18)
Unicode                        76.6ms ± 2%       76.5ms ± 2%    ~     (p=0.603 n=19+19)
GoTypes                         694ms ± 1%        693ms ± 1%    ~     (p=0.569 n=19+20)
Compiler                        3.26s ± 0%        3.27s ± 0%  +0.25%  (p=0.000 n=20+20)
SSA                             7.41s ± 0%        7.49s ± 0%  +1.10%  (p=0.000 n=17+19)
Flate                           120ms ± 1%        120ms ± 1%  +0.38%  (p=0.003 n=19+19)
GoParser                        152ms ± 1%        152ms ± 1%    ~     (p=0.061 n=17+19)
Reflect                         422ms ± 1%        425ms ± 2%  +0.76%  (p=0.001 n=18+20)
Tar                             167ms ± 1%        167ms ± 0%    ~     (p=0.730 n=18+19)
XML                             233ms ± 4%        231ms ± 1%    ~     (p=0.752 n=20+17)
LinkCompiler                    927ms ± 8%        928ms ± 8%    ~     (p=0.857 n=19+20)
ExternalLinkCompiler            1.81s ± 2%        1.81s ± 2%    ~     (p=0.513 n=19+20)
LinkWithoutDebugCompiler        556ms ±10%        583ms ±13%  +4.95%  (p=0.007 n=20+20)
[Geo mean]                      478ms             481ms       +0.52%

name                      old user-time/op  new user-time/op  delta
Template                        270ms ± 5%        269ms ± 7%    ~     (p=0.925 n=20+20)
Unicode                         134ms ± 7%        131ms ±14%    ~     (p=0.593 n=18+20)
GoTypes                         981ms ± 3%        987ms ± 2%  +0.63%  (p=0.049 n=19+18)
Compiler                        4.50s ± 2%        4.50s ± 1%    ~     (p=0.588 n=19+20)
SSA                             10.6s ± 2%        10.6s ± 1%    ~     (p=0.141 n=20+19)
Flate                           164ms ± 8%        165ms ±10%    ~     (p=0.738 n=20+20)
GoParser                        202ms ± 5%        203ms ± 6%    ~     (p=0.820 n=20+20)
Reflect                         587ms ± 6%        597ms ± 3%    ~     (p=0.087 n=20+18)
Tar                             230ms ± 6%        228ms ± 8%    ~     (p=0.569 n=19+20)
XML                             311ms ± 6%        314ms ± 5%    ~     (p=0.369 n=20+20)
LinkCompiler                    878ms ± 8%        887ms ± 7%    ~     (p=0.289 n=20+20)
ExternalLinkCompiler            1.60s ± 7%        1.60s ± 7%    ~     (p=0.820 n=20+20)
LinkWithoutDebugCompiler        498ms ±12%        489ms ±11%    ~     (p=0.398 n=20+20)
[Geo mean]                      611ms             611ms       +0.05%

name                      old alloc/op      new alloc/op      delta
Template                       36.1MB ± 0%       36.0MB ± 0%  -0.32%  (p=0.000 n=20+20)
Unicode                        28.3MB ± 0%       28.3MB ± 0%  -0.03%  (p=0.000 n=19+20)
GoTypes                         121MB ± 0%        121MB ± 0%    ~     (p=0.226 n=16+20)
Compiler                        563MB ± 0%        563MB ± 0%    ~     (p=0.166 n=20+19)
SSA                            1.32GB ± 0%       1.33GB ± 0%  +0.88%  (p=0.000 n=20+19)
Flate                          22.7MB ± 0%       22.7MB ± 0%  -0.02%  (p=0.033 n=19+20)
GoParser                       27.9MB ± 0%       27.9MB ± 0%  -0.02%  (p=0.001 n=20+20)
Reflect                        78.3MB ± 0%       78.2MB ± 0%  -0.01%  (p=0.019 n=20+20)
Tar                            34.0MB ± 0%       34.0MB ± 0%  -0.04%  (p=0.000 n=20+20)
XML                            43.9MB ± 0%       43.9MB ± 0%  -0.07%  (p=0.000 n=20+19)
LinkCompiler                    205MB ± 0%        205MB ± 0%  +0.44%  (p=0.000 n=20+18)
ExternalLinkCompiler            223MB ± 0%        223MB ± 0%  +0.03%  (p=0.000 n=20+20)
LinkWithoutDebugCompiler        139MB ± 0%        142MB ± 0%  +1.75%  (p=0.000 n=20+20)
[Geo mean]                     93.7MB            93.9MB       +0.20%

name                      old allocs/op     new allocs/op     delta
Template                         363k ± 0%         361k ± 0%  -0.58%  (p=0.000 n=20+19)
Unicode                          329k ± 0%         329k ± 0%  -0.06%  (p=0.000 n=19+20)
GoTypes                         1.28M ± 0%        1.28M ± 0%  -0.01%  (p=0.000 n=20+20)
Compiler                        5.40M ± 0%        5.40M ± 0%  -0.01%  (p=0.000 n=20+20)
SSA                             12.7M ± 0%        12.8M ± 0%  +0.80%  (p=0.000 n=20+20)
Flate                            228k ± 0%         228k ± 0%    ~     (p=0.194 n=20+20)
GoParser                         295k ± 0%         295k ± 0%  -0.04%  (p=0.000 n=20+20)
Reflect                          949k ± 0%         949k ± 0%  -0.01%  (p=0.000 n=20+20)
Tar                              337k ± 0%         337k ± 0%  -0.06%  (p=0.000 n=20+20)
XML                              418k ± 0%         417k ± 0%  -0.17%  (p=0.000 n=20+20)
LinkCompiler                     553k ± 0%         554k ± 0%  +0.22%  (p=0.000 n=20+19)
ExternalLinkCompiler            1.52M ± 0%        1.52M ± 0%  +0.27%  (p=0.000 n=20+20)
LinkWithoutDebugCompiler         186k ± 0%         186k ± 0%  +0.06%  (p=0.000 n=20+20)
[Geo mean]                       723k              723k       +0.03%

name                      old text-bytes    new text-bytes    delta
HelloSize                       828kB ± 0%        828kB ± 0%  -0.01%  (p=0.000 n=20+20)

name                      old data-bytes    new data-bytes    delta
HelloSize                      13.4kB ± 0%       13.4kB ± 0%    ~     (all equal)

name                      old bss-bytes     new bss-bytes     delta
HelloSize                       180kB ± 0%        180kB ± 0%    ~     (all equal)

name                      old exe-bytes     new exe-bytes     delta
HelloSize                      1.23MB ± 0%       1.23MB ± 0%  -0.33%  (p=0.000 n=20+20)

file      before    after     Δ       %
addr2line 4320075   4311883   -8192   -0.190%
asm       5191932   5187836   -4096   -0.079%
buildid   2835338   2831242   -4096   -0.144%
compile   20531717  20569099  +37382  +0.182%
cover     5322511   5318415   -4096   -0.077%
dist      3723749   3719653   -4096   -0.110%
doc       4743515   4739419   -4096   -0.086%
fix       3413960   3409864   -4096   -0.120%
link      6690119   6686023   -4096   -0.061%
nm        4269616   4265520   -4096   -0.096%
pprof     14942189  14929901  -12288  -0.082%
trace     11807164  11790780  -16384  -0.139%
vet       8384104   8388200   +4096   +0.049%
go        15339076  15334980  -4096   -0.027%
total     132258257 132226007 -32250  -0.024%

Fixes #30645.

Change-Id: If551ac5996097f3685870d083151b5843170aab0
Reviewed-on: https://go-review.googlesource.com/c/go/+/165998
Run-TryBot: Michael Munday <mike.munday@ibm.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
5 years agocmd/compile: absorb SNEZ into branch on riscv64
Joel Sing [Sun, 1 Mar 2020 17:25:54 +0000 (04:25 +1100)]
cmd/compile: absorb SNEZ into branch on riscv64

Change-Id: I55fd93843a7fb574a7dd66ebb87fdd96e944d555
Reviewed-on: https://go-review.googlesource.com/c/go/+/221682
Reviewed-by: Cherry Zhang <cherryyz@google.com>
5 years agocmd/compile: optimize subtraction of zero on riscv64
Joel Sing [Sun, 1 Mar 2020 17:24:35 +0000 (04:24 +1100)]
cmd/compile: optimize subtraction of zero on riscv64

Change-Id: I9a994b01e9fecb13077c30df4b7677d40d179cce
Reviewed-on: https://go-review.googlesource.com/c/go/+/221681
Reviewed-by: Cherry Zhang <cherryyz@google.com>
5 years agocmd/compile: improve subtraction of constants on riscv64
Joel Sing [Sun, 1 Mar 2020 17:23:12 +0000 (04:23 +1100)]
cmd/compile: improve subtraction of constants on riscv64

Convert subtraction of a constant into an ADDI with a negative immediate,
where possible.

Change-Id: Ie8d54b7538f0012e5f898abea233b2957fe31899
Reviewed-on: https://go-review.googlesource.com/c/go/+/221679
Reviewed-by: Cherry Zhang <cherryyz@google.com>
5 years agoruntime: use CBZ/CBNZ in linux/arm64 assembly code
Xiangdong Ji [Thu, 21 Nov 2019 07:46:37 +0000 (07:46 +0000)]
runtime: use CBZ/CBNZ in linux/arm64 assembly code

Replace compare and branch on zero/non-zero instructions in linux/arm64
assembly files with CBZ/CBNZ.

Change-Id: I4dbf56678f85827e83b5863804368bc28a4603b5
Reviewed-on: https://go-review.googlesource.com/c/go/+/209617
Run-TryBot: Tobias Klauser <tobias.klauser@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Tobias Klauser <tobias.klauser@gmail.com>
5 years agodoc/mem: remove unnecessary pre tags within same snippet
Jingwei [Sat, 14 Dec 2019 07:52:17 +0000 (15:52 +0800)]
doc/mem: remove unnecessary pre tags within same snippet

currently the snippet is segmented but should be one code snippet.

Change-Id: Ic747faf9bb1b52f9d1786eca70616a05b71ee801
Reviewed-on: https://go-review.googlesource.com/c/go/+/211198
Reviewed-by: Emmanuel Odeke <emm.odeke@gmail.com>
5 years agocmd/compile/internal/riscv64: correct ssa.BlockRetJmp
Joel Sing [Sun, 1 Mar 2020 17:26:21 +0000 (04:26 +1100)]
cmd/compile/internal/riscv64: correct ssa.BlockRetJmp

The obj.Prog needs to be an obj.ARET rather than an obj.AJMP, otherwise the
epilogue does not get correctly produced.

Change-Id: Ie1262f2028d3b51720eeb0364a627fbde8b14df9
Reviewed-on: https://go-review.googlesource.com/c/go/+/221683
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

5 years agoflag: changed flag variable name in package doc, for clarity
yuz [Sun, 1 Mar 2020 12:35:56 +0000 (21:35 +0900)]
flag: changed flag variable name in package doc, for clarity

Changed the flag variable name to nFlag instead of flagname,
because flagname was confusing.

Change-Id: I20dd4c4b4f605395d427a125ba4fd14580e5d766
Reviewed-on: https://go-review.googlesource.com/c/go/+/221678
Reviewed-by: Rob Pike <r@golang.org>
5 years agoruntime: during panic, print value instead of address, if kind is printable
Emmanuel T Odeke [Mon, 2 Mar 2020 01:41:44 +0000 (17:41 -0800)]
runtime: during panic, print value instead of address, if kind is printable

Make panics more useful by printing values, if their
underlying kind is printable, instead of just their memory address.

Thus now given any custom type derived from any of:
    float*, int*, string, uint*

if we have panic with such a result, its value will be printed.

Thus given any of:
    type MyComplex128 complex128
    type MyFloat64 float64
    type MyString string
    type MyUintptr uintptr

    panic(MyComplex128(32.1 + 10i))
    panic(MyFloat64(-93.7))
    panic(MyString("This one"))
    panic(MyUintptr(93))

They will now print in the panic:

    panic: main.MyComplex64(+1.100000e-001+3.000000e+000i)
    panic: main.MyFloat64(-9.370000e+001)
    panic: main.MyString("This one")
    panic: main.MyUintptr(93)

instead of:

    panic: (main.MyComplex128) (0xe0100,0x138cc0)
    panic: (main.MyFloat64) (0xe0100,0x138068)
    panic: (main.MyString) (0x48aa00,0x4c0840)
    panic: (main.MyUintptr) (0xe0100,0x137e58)

and anything else will be printed as in the past with:

    panic: (main.MyStruct) (0xe4ee0,0x40a0e0)

Also while here, updated the Go1.15 release notes.

Fixes #37531

Change-Id: Ia486424344a386014f2869ab3483e42a9ef48ac4
Reviewed-on: https://go-review.googlesource.com/c/go/+/221779
Run-TryBot: Emmanuel Odeke <emm.odeke@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
5 years agocmd/compile/internal/syntax: add -skip flag to exclude files from TestStdLib
Robert Griesemer [Sat, 29 Feb 2020 06:25:39 +0000 (22:25 -0800)]
cmd/compile/internal/syntax: add -skip flag to exclude files from TestStdLib

TestStdLib reports parsed lines and lines/s information. To make
it easier to compare apples to apples when making changes in the
std lib, a regular expression provided via the -skip flag filters
files we don't want to process.

Change-Id: I27d9c32032eac4e78581205892e4f26947c91bd9
Reviewed-on: https://go-review.googlesource.com/c/go/+/221600
Reviewed-by: Emmanuel Odeke <emm.odeke@gmail.com>
5 years agoruntime: prevent allocation when converting small ints to interfaces
Josh Bleecher Snyder [Mon, 27 Jan 2020 19:28:05 +0000 (11:28 -0800)]
runtime: prevent allocation when converting small ints to interfaces

Prior to this change, we avoid allocation when
converting 0 to an interface.

This change extends that optimization to larger value types
whose values happens to be in the range 0 to 255.
This is marginally more expensive in the case of a 0 value,
in that the address is computed rather than fixed.

name                         old time/op  new time/op  delta
ConvT2ESmall-8               2.36ns ± 4%  2.65ns ± 4%  +12.23%  (p=0.000 n=87+91)
ConvT2EUintptr-8             2.36ns ± 4%  2.84ns ± 6%  +20.05%  (p=0.000 n=96+99)
ConvT2ELarge-8               23.8ns ± 2%  23.1ns ± 3%   -2.94%  (p=0.000 n=93+95)
ConvT2ISmall-8               2.67ns ± 5%  2.74ns ±27%     ~     (p=0.214 n=99+100)
ConvT2IUintptr-8             2.65ns ± 5%  2.46ns ± 5%   -7.19%  (p=0.000 n=98+98)
ConvT2ILarge-8               24.2ns ± 2%  23.5ns ± 4%   -3.16%  (p=0.000 n=91+97)
ConvT2Ezero/zero/16-8        2.79ns ± 6%  2.99ns ± 4%   +7.52%  (p=0.000 n=94+88)
ConvT2Ezero/zero/32-8        2.34ns ± 3%  2.65ns ± 3%  +13.06%  (p=0.000 n=92+98)
ConvT2Ezero/zero/64-8        2.35ns ± 4%  2.65ns ± 6%  +12.86%  (p=0.000 n=99+94)
ConvT2Ezero/zero/str-8       2.55ns ± 4%  2.54ns ± 4%     ~     (p=0.063 n=97+99)
ConvT2Ezero/zero/slice-8     2.82ns ± 4%  2.85ns ± 5%   +1.00%  (p=0.000 n=99+95)
ConvT2Ezero/zero/big-8       94.3ns ± 5%  93.4ns ± 4%   -0.94%  (p=0.000 n=88+90)
ConvT2Ezero/nonzero/str-8    29.6ns ± 3%  27.7ns ± 3%   -6.69%  (p=0.000 n=98+97)
ConvT2Ezero/nonzero/slice-8  36.6ns ± 2%  37.1ns ± 2%   +1.31%  (p=0.000 n=94+90)
ConvT2Ezero/nonzero/big-8    93.4ns ± 3%  92.7ns ± 3%   -0.74%  (p=0.000 n=88+84)
ConvT2Ezero/smallint/16-8    13.3ns ± 4%   2.7ns ± 6%  -79.82%  (p=0.000 n=100+97)
ConvT2Ezero/smallint/32-8    12.5ns ± 1%   2.9ns ± 5%  -77.17%  (p=0.000 n=85+96)
ConvT2Ezero/smallint/64-8    14.7ns ± 3%   2.6ns ± 3%  -82.05%  (p=0.000 n=94+94)
ConvT2Ezero/largeint/16-8    14.0ns ± 4%  13.2ns ± 7%   -5.44%  (p=0.000 n=95+99)
ConvT2Ezero/largeint/32-8    12.8ns ± 4%  12.9ns ± 3%     ~     (p=0.096 n=99+87)
ConvT2Ezero/largeint/64-8    15.5ns ± 2%  15.0ns ± 2%   -3.46%  (p=0.000 n=95+96)

An example of a program for which this makes a perceptible difference
is running the compiler with the -S flag:

name        old time/op       new time/op       delta
Template          349ms ± 2%        344ms ± 2%   -1.48%  (p=0.000 n=23+25)
Unicode           138ms ± 4%        136ms ± 3%   -1.67%  (p=0.003 n=25+25)
GoTypes           1.25s ± 2%        1.24s ± 2%   -1.11%  (p=0.001 n=24+25)
Compiler          5.73s ± 2%        5.67s ± 2%   -1.09%  (p=0.002 n=25+24)
SSA               20.2s ± 2%        19.9s ± 2%   -1.45%  (p=0.000 n=25+23)
Flate             216ms ± 4%        210ms ± 2%   -2.77%  (p=0.000 n=25+24)
GoParser          283ms ± 2%        278ms ± 3%   -1.58%  (p=0.000 n=23+23)
Reflect           757ms ± 2%        745ms ± 2%   -1.58%  (p=0.000 n=25+25)
Tar               303ms ± 4%        296ms ± 2%   -2.20%  (p=0.000 n=22+23)
XML               415ms ± 2%        411ms ± 3%   -0.94%  (p=0.002 n=25+22)
[Geo mean]        726ms             715ms        -1.59%

name        old user-time/op  new user-time/op  delta
Template          434ms ± 3%        427ms ± 2%   -1.66%  (p=0.000 n=23+24)
Unicode           204ms ±12%        198ms ±12%   -2.83%  (p=0.032 n=25+25)
GoTypes           1.59s ± 2%        1.56s ± 2%   -1.64%  (p=0.000 n=22+25)
Compiler          7.50s ± 1%        7.40s ± 2%   -1.32%  (p=0.000 n=25+25)
SSA               27.2s ± 2%        26.8s ± 2%   -1.50%  (p=0.000 n=24+23)
Flate             266ms ± 6%        254ms ± 3%   -4.38%  (p=0.000 n=25+25)
GoParser          357ms ± 2%        351ms ± 2%   -1.90%  (p=0.000 n=24+23)
Reflect           966ms ± 2%        947ms ± 2%   -1.94%  (p=0.000 n=24+25)
Tar               387ms ± 2%        380ms ± 3%   -1.83%  (p=0.000 n=22+24)
XML               538ms ± 1%        532ms ± 1%   -1.15%  (p=0.000 n=24+20)
[Geo mean]        942ms             923ms        -2.02%

name        old alloc/op      new alloc/op      delta
Template         54.1MB ± 0%       52.9MB ± 0%   -2.26%  (p=0.000 n=25+25)
Unicode          33.5MB ± 0%       33.1MB ± 0%   -1.03%  (p=0.000 n=25+24)
GoTypes           189MB ± 0%        185MB ± 0%   -2.27%  (p=0.000 n=25+25)
Compiler          875MB ± 0%        858MB ± 0%   -1.99%  (p=0.000 n=23+25)
SSA              3.19GB ± 0%       3.13GB ± 0%   -1.95%  (p=0.000 n=25+25)
Flate            32.9MB ± 0%       32.2MB ± 0%   -2.26%  (p=0.000 n=25+25)
GoParser         44.0MB ± 0%       42.9MB ± 0%   -2.33%  (p=0.000 n=25+25)
Reflect           117MB ± 0%        114MB ± 0%   -2.60%  (p=0.000 n=25+25)
Tar              48.6MB ± 0%       47.5MB ± 0%   -2.18%  (p=0.000 n=25+24)
XML              65.7MB ± 0%       64.4MB ± 0%   -1.96%  (p=0.000 n=23+25)
[Geo mean]        118MB             115MB        -2.08%

name        old allocs/op     new allocs/op     delta
Template          1.07M ± 0%        0.92M ± 0%  -14.29%  (p=0.000 n=25+25)
Unicode            539k ± 0%         494k ± 0%   -8.27%  (p=0.000 n=25+25)
GoTypes           3.97M ± 0%        3.43M ± 0%  -13.71%  (p=0.000 n=24+25)
Compiler          17.6M ± 0%        15.4M ± 0%  -12.69%  (p=0.000 n=25+24)
SSA               66.1M ± 0%        58.1M ± 0%  -12.17%  (p=0.000 n=25+25)
Flate              629k ± 0%         536k ± 0%  -14.73%  (p=0.000 n=24+24)
GoParser           929k ± 0%         799k ± 0%  -13.96%  (p=0.000 n=25+25)
Reflect           2.49M ± 0%        2.11M ± 0%  -15.28%  (p=0.000 n=25+25)
Tar                919k ± 0%         788k ± 0%  -14.30%  (p=0.000 n=25+25)
XML               1.28M ± 0%        1.11M ± 0%  -12.85%  (p=0.000 n=24+25)
[Geo mean]        2.32M             2.01M       -13.24%

There is a slight increase in binary size from this change:

file      before    after     Δ       %
addr2line 4307728   4307760   +32     +0.001%
api       5972680   5972728   +48     +0.001%
asm       5114200   5114232   +32     +0.001%
buildid   2843720   2847848   +4128   +0.145%
cgo       4823736   4827864   +4128   +0.086%
compile   24912056  24912104  +48     +0.000%
cover     5259800   5259832   +32     +0.001%
dist      3665080   3665128   +48     +0.001%
doc       4672712   4672744   +32     +0.001%
fix       3376952   3376984   +32     +0.001%
link      6618008   6622152   +4144   +0.063%
nm        4253280   4257424   +4144   +0.097%
objdump   4655376   4659504   +4128   +0.089%
pack      2294280   2294328   +48     +0.002%
pprof     14747476  14751620  +4144   +0.028%
test2json 2819320   2823448   +4128   +0.146%
trace     11665068  11669212  +4144   +0.036%
vet       8342360   8342408   +48     +0.001%

Change-Id: I38ef70244e23069bfd14334061d43ae22a294519
Reviewed-on: https://go-review.googlesource.com/c/go/+/216401
Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
5 years agocmd/compile: regenerate rules
Josh Bleecher Snyder [Mon, 2 Mar 2020 21:57:17 +0000 (13:57 -0800)]
cmd/compile: regenerate rules

CL 210897 went in concurrently with some rulegen.go changes.
Regenerate.

Change-Id: I39ffa8bdffdfcc7f60cc8158d188fb1a3e70fcb6
Reviewed-on: https://go-review.googlesource.com/c/go/+/221787
Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
5 years agogo/types: simplify method set computation
Robert Griesemer [Sat, 8 Feb 2020 00:01:01 +0000 (16:01 -0800)]
go/types: simplify method set computation

After fixing #37081 we don't need to explicitly keep track of
field collisions in the method set computation anymore; we only
need to know which field (names) exists at each embedding level.
Simplify the code by removing the dedicated fieldSet data type
in favor of a simple string set.

Follow-up on https://golang.org/cl/218617; separate CL to make it
easier to identify a problem with these two changes, should there
be one.

Updates #37081.

Change-Id: I5c259c63c75a148a42d5c3e1e4860e1ffe5631bd
Reviewed-on: https://go-review.googlesource.com/c/go/+/218618
Reviewed-by: Alan Donovan <adonovan@google.com>
5 years agogo/types: fix method set computation
Robert Griesemer [Fri, 7 Feb 2020 23:26:19 +0000 (15:26 -0800)]
go/types: fix method set computation

When computing method sets, any struct field that "shadows" a
method at a lower embedding level eliminates that method from
the method set. Treat any field at a given level as a "collision"
for any methods at lower embedding level.

Method sets are not directly used by go/types (except for self-
verification in debug mode); they are a functionality provided
by go/types. Thus, the method sets that go/types is using were
not affected by this bug.

Fixes #37081.

Change-Id: Ic1937e01891b3614a6f7965d4384aeb485f3fe3e
Reviewed-on: https://go-review.googlesource.com/c/go/+/218617
Reviewed-by: Alan Donovan <adonovan@google.com>
5 years agocmd/compile: add -d=ssa/check/seed=SEED
Josh Bleecher Snyder [Fri, 24 Jan 2020 21:52:41 +0000 (13:52 -0800)]
cmd/compile: add -d=ssa/check/seed=SEED

This change adds the option to run the ssa checker with a random seed.
The current system uses a completely fixed seed,
which is good for reproducibility but bad for exploring the state space.

Preserve what we have, but also provide a way for the caller
to provide a seed. The caller can report the seed
alongside any failures.

Change-Id: I2676a8112d8260e6cac86d95d2e8db4d3221aeeb
Reviewed-on: https://go-review.googlesource.com/c/go/+/216418
Reviewed-by: Keith Randall <khr@golang.org>
5 years agocmd/compile: remove duplicate ppc64 rules
Josh Bleecher Snyder [Thu, 6 Feb 2020 18:36:29 +0000 (10:36 -0800)]
cmd/compile: remove duplicate ppc64 rules

Const64 gets lowered to MOVDconst.
Change rules using interior Const64 to use MOVDconst instead,
to be less dependent on rule application order.

As a result of doing this, some of the rules end up being
exact duplicates; remove those.

We had those exact duplicates because of the order dependency;
ppc64 had no way to optimize away shifts by a constant
if the initial lowering didn't catch it.

Add those optimizations as well.
The outcome is the same, but this makes the overall rules more robust.

Change-Id: Iadd97a9fe73d52358d571d022ace145e506d160b
Reviewed-on: https://go-review.googlesource.com/c/go/+/220877
Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Lynn Boger <laboger@linux.vnet.ibm.com>
5 years agoruntime: print instruction bytes when reporting a SIGILL
Keith Randall [Fri, 28 Feb 2020 20:59:38 +0000 (12:59 -0800)]
runtime: print instruction bytes when reporting a SIGILL

Print the bytes of the instruction that generated a SIGILL.
This should help us respond to bug reports without having to
go back-and-forth with the reporter to get the instruction involved.
Might also help with SIGILL problems that are difficult to reproduce.

Update #37513

Change-Id: I33059b1dbfc97bce16142a843f32a88a6547e280
Reviewed-on: https://go-review.googlesource.com/c/go/+/221431
Run-TryBot: Keith Randall <khr@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
5 years agocmd/compile: make pre-elimination of rulegen bounds checks more precise
Josh Bleecher Snyder [Mon, 2 Mar 2020 00:44:06 +0000 (16:44 -0800)]
cmd/compile: make pre-elimination of rulegen bounds checks more precise

In cases in which we had a named value whose args were all _,
like this rule from ARM.rules:

(MOVBUreg x:(MOVBUload _ _)) -> (MOVWreg x)

We previously inserted

_ = x.Args[1]

even though it is unnecessary.
This change eliminates this pointless bounds check.
And in other cases, we now check bounds just as far as strictly necessary.

No significant movement on any compiler metrics.
Just nicer (and less) code.

Passes toolstash-check -all.

Change-Id: I075dfe9f926cc561cdc705e9ddaab563164bed3a
Reviewed-on: https://go-review.googlesource.com/c/go/+/221781
Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
5 years agocmd/compile: add streamlined Block Reset+AddControl routines
Josh Bleecher Snyder [Sun, 1 Mar 2020 21:09:09 +0000 (13:09 -0800)]
cmd/compile: add streamlined Block Reset+AddControl routines

For use in rewrite rules. Shrinks cmd/compile:

compile 20082104  19967416  -114688 -0.571%

Passes toolstash-check -all.

Change-Id: Ic856508b27ec5b7fb9b6ca63e955a7139ae7dc30
Reviewed-on: https://go-review.googlesource.com/c/go/+/221780
Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
5 years agoruntime: use quiet NaNs in softfloat implementation
Keith Randall [Fri, 28 Feb 2020 23:42:03 +0000 (15:42 -0800)]
runtime: use quiet NaNs in softfloat implementation

Update #37455

Change-Id: Ieac0823aa398d73187c009037be15ba34c84f3d9
Reviewed-on: https://go-review.googlesource.com/c/go/+/221433
Run-TryBot: Emmanuel Odeke <emm.odeke@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
5 years agohash/maphash: add more tests for seed generation
Keith Randall [Mon, 24 Feb 2020 03:23:15 +0000 (19:23 -0800)]
hash/maphash: add more tests for seed generation

Test all the paths by which a Hash picks its seed.
Make sure they all behave identically to a preset seed.

Change-Id: I2f7950857697f2f07226b96655574c36931b2aae
Reviewed-on: https://go-review.googlesource.com/c/go/+/220686
Run-TryBot: Keith Randall <khr@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Vladimir Evgrafov <evgrafov.vladimir@gmail.com>
Reviewed-by: Alan Donovan <adonovan@google.com>
5 years agocmd/compile: add specialized Value reset for OpCopy
Josh Bleecher Snyder [Wed, 30 Oct 2019 17:29:47 +0000 (10:29 -0700)]
cmd/compile: add specialized Value reset for OpCopy

This:

* Simplifies and shortens the generated code for rewrite rules.
* Shrinks cmd/compile by 86k (0.4%) and makes it easier to compile.
* Removes the stmt boundary code wrangling from Value.reset,
  in favor of doing it in the one place where it actually does some work,
  namely the writebarrier pass. (This was ascertained by inspecting the
  code for cases in which notStmtBoundary values were generated.)

Passes toolstash-check -all.

Change-Id: I25671d4c4bbd772f235195d11da090878ea2cc07
Reviewed-on: https://go-review.googlesource.com/c/go/+/221421
Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: David Chase <drchase@google.com>
5 years agocmd/compile: add intrinsics for runtime/internal/math on MIPS64x
Meng Zhuo [Sat, 29 Feb 2020 03:23:29 +0000 (11:23 +0800)]
cmd/compile: add intrinsics for runtime/internal/math on MIPS64x

name              old time/op  new time/op  delta
MulUintptr/small  8.42ns ± 0%  5.93ns ± 0%  -29.66%  (p=0.000 n=9+10)
MulUintptr/large  11.1ns ± 0%   7.4ns ± 0%  -33.17%  (p=0.000 n=10+9)

Change-Id: I6659a886389660461fc2c90bd248243f6e7c29d5
Reviewed-on: https://go-review.googlesource.com/c/go/+/210897
Run-TryBot: Meng Zhuo <mengzhuo1203@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
5 years agonet/http: verify RoundTripper invariants in the send function
Bryan C. Mills [Mon, 2 Mar 2020 15:16:39 +0000 (10:16 -0500)]
net/http: verify RoundTripper invariants in the send function

Issue #37598 reports a nil-panic in *Client.send that can
only occur if one of the RoundTripper invariants is violated.
Unfortunately, that condition is currently difficult to diagnose: it
manifests as a panic during a Response field access, rather than
something the user can easily associate with an specific erroneous
RoundTripper implementation.

No test because the new code paths are supposed to be unreachable.

Updates #37598

Change-Id: If0451e9c6431f6fab7137de43727297a80def05b
Reviewed-on: https://go-review.googlesource.com/c/go/+/221818
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
5 years agodoc: race condition in unsynchronized send/close
Changkun Ou [Sun, 16 Feb 2020 00:11:53 +0000 (01:11 +0100)]
doc: race condition in unsynchronized send/close

This CL documents that unsynchronized send and close operations
on a channel are detected as a race condition.

Fixes #27769

Change-Id: I7495a2d0dd834c3f3b6339f8ca18ea21ae979aa8
Reviewed-on: https://go-review.googlesource.com/c/go/+/219637
Reviewed-by: Rob Pike <r@golang.org>
5 years agointernal/cpu: use anonymous struct for CPU feature vars
Tobias Klauser [Mon, 2 Mar 2020 08:31:44 +0000 (09:31 +0100)]
internal/cpu: use anonymous struct for CPU feature vars

Like in x/sys/cpu, use anonymous structs to declare the CPU feature vars
instead of defining single-use types. Also, order the vars
alphabetically.

Change-Id: Iedd3ca51916e3cbb852d2aeed18b3a4c6613e778
Reviewed-on: https://go-review.googlesource.com/c/go/+/221757
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Reviewed-by: Martin Möhrmann <moehrmann@google.com>
5 years agoos: plan9 seek() should invalidate cached directory info
Keith Randall [Mon, 2 Mar 2020 01:40:35 +0000 (17:40 -0800)]
os: plan9 seek() should invalidate cached directory info

Update #37161

Change-Id: Iee828bbcc8436af29ca6dd9ed897cb5265a57cf8
Reviewed-on: https://go-review.googlesource.com/c/go/+/221778
Run-TryBot: Keith Randall <khr@golang.org>
Reviewed-by: Emmanuel Odeke <emm.odeke@gmail.com>
5 years agotime: use values larger than 24 for day for time.Format examples
Jean de Klerk [Sun, 1 Mar 2020 00:35:51 +0000 (17:35 -0700)]
time: use values larger than 24 for day for time.Format examples

Currently, the time.Format docs use 7 Mar 2015 as the day/month/year. In numeric
form, that is either 7/3/2015 or 3/7/2015 depending on which part of the world
you're from. This is extremely confusing.

In fact, the reference time being defined in a very US-centric way is quite
confusing for the rest of the world, too [1].

We can't change that, but we can make the time.Format docs more comprehendable
to the rest of the world without sacrificing by simply choosing a day that is
not ambiguous (a value greater than 24 for day). This CL does makes the
necessary change.

Note: this CL moves some of the padding examples into their own example, since
those examples do need a <10 day to demonstrate padding.

1: Additional context: a very old golang-nuts thread in which Rob expresses some
regret about the format being the USA standard, rather than the alternative:
https://groups.google.com/forum/m/#!msg/golang-nuts/0nQbfyNzk9E/LWbMgpRQNOgJ.

Change-Id: If0a07c5e0dab86f8420cbf59543405eb857aa7f2
Reviewed-on: https://go-review.googlesource.com/c/go/+/221612
Run-TryBot: Jean de Klerk <deklerk@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Reviewed-by: Rob Pike <r@golang.org>
5 years agoos: seek should invalidate any cached directory reads
Keith Randall [Wed, 12 Feb 2020 01:49:52 +0000 (17:49 -0800)]
os: seek should invalidate any cached directory reads

When we seek on the underlying FD, discard any directory entries
we've already read and cached. This makes sure we won't return
the same entry twice.

We already fixed this for Darwin in CL 209961.

Fixes #37161

Change-Id: I20e1ac8d751443135e67fb4c43c18d69befb643b
Reviewed-on: https://go-review.googlesource.com/c/go/+/219143
Run-TryBot: Keith Randall <khr@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Emmanuel Odeke <emm.odeke@gmail.com>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
5 years agosyscall/js: improve documentation of js.FuncOf
Torben Schinke [Sun, 1 Mar 2020 20:07:46 +0000 (20:07 +0000)]
syscall/js: improve documentation of js.FuncOf

The existing documentation is improved to be more
explicit about the lifecycle and its consequences.

Fixes #34324

Change-Id: I9969afc69f6eeb7812c11fe821a842794df5aa5b
GitHub-Last-Rev: 246a4991660927f88f48290580e96b15c16663c1
GitHub-Pull-Request: golang/go#34551
Reviewed-on: https://go-review.googlesource.com/c/go/+/197458
Reviewed-by: Richard Musiol <neelance@gmail.com>
5 years agosyscall: fix Fchdir on js/wasm
Richard Musiol [Sun, 1 Mar 2020 16:01:58 +0000 (17:01 +0100)]
syscall: fix Fchdir on js/wasm

NodeJS does not support fchdir so it has to be emulated with chdir by
saving the path when opening a directory.

However, if the path opened is relative, saving this path is not
sufficient, because after changing the working directory the path
does not resolve correctly any more, thus a subsequent fd.Chdir() fails.

This change fixes the issue by resolving a relative path when
opening the directory and saving the absolute path instead.

Fixes #37448

Change-Id: Id6bc8c4232b0019fc11e850599a526336608ce54
Reviewed-on: https://go-review.googlesource.com/c/go/+/221717
Run-TryBot: Richard Musiol <neelance@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Emmanuel Odeke <emm.odeke@gmail.com>
5 years agocmd/compile: add specialized AddArgN functions for rewrite rules
Josh Bleecher Snyder [Wed, 26 Feb 2020 19:29:34 +0000 (11:29 -0800)]
cmd/compile: add specialized AddArgN functions for rewrite rules

This shrinks the compiler without impacting performance.
(The performance-sensitive part of rewrite rules is the non-match case.)
Passes toolstash-check -all.

Executable size:

file    before    after     Δ       %
compile 20356168  20163960  -192208 -0.944%
total   115599376 115407168 -192208 -0.166%

Text size:

file                       before   after    Δ       %
cmd/compile/internal/ssa.s 3928309  3778774  -149535 -3.807%
total                      18862943 18713408 -149535 -0.793%

Memory allocated compiling package SSA:

SSA               12.7M ± 0%        12.5M ± 0%  -1.74%  (p=0.008 n=5+5)

Compiler speed impact:

name        old time/op       new time/op       delta
Template          211ms ± 1%        211ms ± 2%    ~     (p=0.832 n=49+49)
Unicode          82.8ms ± 2%       83.2ms ± 2%  +0.44%  (p=0.022 n=46+49)
GoTypes           726ms ± 1%        728ms ± 2%    ~     (p=0.076 n=46+48)
Compiler          3.39s ± 2%        3.40s ± 2%    ~     (p=0.633 n=48+49)
SSA               7.71s ± 1%        7.65s ± 1%  -0.78%  (p=0.000 n=45+44)
Flate             134ms ± 1%        134ms ± 1%    ~     (p=0.195 n=50+49)
GoParser          167ms ± 1%        167ms ± 1%    ~     (p=0.390 n=47+47)
Reflect           453ms ± 3%        452ms ± 2%    ~     (p=0.492 n=48+49)
Tar               184ms ± 3%        184ms ± 2%    ~     (p=0.862 n=50+48)
XML               248ms ± 2%        248ms ± 2%    ~     (p=0.096 n=49+47)
[Geo mean]        415ms             415ms       -0.03%

name        old user-time/op  new user-time/op  delta
Template          273ms ± 1%        273ms ± 2%    ~     (p=0.711 n=48+48)
Unicode           117ms ± 6%        117ms ± 5%    ~     (p=0.633 n=50+50)
GoTypes           972ms ± 2%        974ms ± 1%  +0.29%  (p=0.016 n=47+49)
Compiler          4.46s ± 6%        4.51s ± 6%    ~     (p=0.093 n=50+50)
SSA               10.4s ± 1%        10.3s ± 2%  -0.94%  (p=0.000 n=45+50)
Flate             166ms ± 2%        167ms ± 2%    ~     (p=0.148 n=49+48)
GoParser          202ms ± 1%        202ms ± 2%  -0.28%  (p=0.014 n=47+49)
Reflect           594ms ± 2%        594ms ± 2%    ~     (p=0.717 n=48+49)
Tar               224ms ± 2%        224ms ± 2%    ~     (p=0.805 n=50+49)
XML               311ms ± 1%        310ms ± 1%    ~     (p=0.177 n=49+48)
[Geo mean]        537ms             537ms       +0.01%

Change-Id: I562b9f349b34ddcff01771769e6dbbc80604da7a
Reviewed-on: https://go-review.googlesource.com/c/go/+/221237
Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
5 years agoruntime: deflake CGO traceback tests
Mark Pulford [Thu, 13 Feb 2020 21:34:31 +0000 (08:34 +1100)]
runtime: deflake CGO traceback tests

The CGO traceback function is called whenever CGO code is executing and
a signal is received. This occurs much more frequently now SIGURG
is used for preemption.

Disable signal preemption to significantly increase the likelihood that
a signal results in a profile sample during the test.

Updates #37201

Change-Id: Icb1a33ab0754d1a74882a4ee265b4026abe30bdc
Reviewed-on: https://go-review.googlesource.com/c/go/+/219417
Run-TryBot: Emmanuel Odeke <emm.odeke@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
5 years agocmd/compile: add a dark mode to ssa html generation which can be toggled
Bradford Lamson-Scribner [Thu, 20 Feb 2020 16:07:48 +0000 (09:07 -0700)]
cmd/compile: add a dark mode to ssa html generation which can be toggled

add a tag that when clicked, toggles a dark mode. It keeps intact
the grayed out dead values/blocks, all the highlight colors, and ensures
text is always readable.

Fixes #34325

Change-Id: I4af1e4b5f4a5b63e54c992e90f8474cc51c63465
Reviewed-on: https://go-review.googlesource.com/c/go/+/220260
Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

5 years agocmd/go, cmd/link: implement -buildmode=pie on windows
Alex Brainman [Tue, 25 Feb 2020 07:42:24 +0000 (18:42 +1100)]
cmd/go, cmd/link: implement -buildmode=pie on windows

This CL implements windows version of -buildmode=pie code in both
cmd/go and cmd/link.

Windows executables built with -buildmode=pie set (unlike the one
built with -buildmode=exe) will have extra .reloc PE section, and
will have no IMAGE_FILE_RELOCS_STRIPPED flag set. They will also
have IMAGE_DLLCHARACTERISTICS_DYNAMIC_BASE flag set, and
IMAGE_DLLCHARACTERISTICS_HIGH_ENTROPY_VA flag set for windows/amd64.

Both cgo and non-cgo versions are implemented. And TestBuildmodePIE
is extended to test both cgo and non-cgo versions on windows and
linux.

This CL used some code from CLs 152759 and 203602.

RELNOTE=yes

Fixes #27144
Updates #35192

Change-Id: I1249e4ffbd79bd4277efefb56db321c390c0f76f
Reviewed-on: https://go-review.googlesource.com/c/go/+/214397
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
5 years agotime: optimize Time.ISOWeek
Shuo [Sun, 1 Mar 2020 02:32:32 +0000 (02:32 +0000)]
time: optimize Time.ISOWeek

name       old time/op  new time/op  delta
ISOWeek-4  57.7ns ± 5%  27.9ns ±10%  -51.54%  (p=0.000 n=48+49)

Fixes #37534

Change-Id: Ic4673ced44a4b0190018e87207743ed9500fb1e0
GitHub-Last-Rev: a376c57e83a99f8e8fde297335caa85215e7aead
GitHub-Pull-Request: golang/go#36316
Reviewed-on: https://go-review.googlesource.com/c/go/+/212837
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
5 years agoruntime/pprof/internal/profile: make error message readable
Ian Lance Taylor [Thu, 27 Feb 2020 19:24:24 +0000 (11:24 -0800)]
runtime/pprof/internal/profile: make error message readable

The error message for an unrecognized type in decodeField was using
string(i) for an int type i. It was recently changed (by  me) to
string(rune(i)), but that just avoided a vet warning without fixing
the problem. This CL fixes the problem by using fmt.Errorf.

We also change the message to "unknown wire type" to match the master
copy of this code in github.com/google/pprof/profile/proto.go.

Updates #32479

Change-Id: Ia91ea6d5edbd7cd946225d1ee96bb7623b52bb44
Reviewed-on: https://go-review.googlesource.com/c/go/+/221384
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Emmanuel Odeke <emm.odeke@gmail.com>
5 years agoruntime: don't invoke t.Fatal* in goroutine in TestLibraryCtrlHandler
Emmanuel T Odeke [Sat, 29 Feb 2020 10:32:21 +0000 (02:32 -0800)]
runtime: don't invoke t.Fatal* in goroutine in TestLibraryCtrlHandler

Change-Id: I8bb06c360cab3e5a74b0b0f98bb25cca4741d66d
Reviewed-on: https://go-review.googlesource.com/c/go/+/221605
Run-TryBot: Emmanuel Odeke <emm.odeke@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
5 years agonet/textproto: pass missing argument to fmt.Sprintf
Ian Lance Taylor [Thu, 27 Feb 2020 02:34:25 +0000 (18:34 -0800)]
net/textproto: pass missing argument to fmt.Sprintf

The vet tool didn't catch this because the fmt.Sprintf format argument
was written as an expression.

Fixes #37467

Change-Id: I72c20ba45e3f42c195fa5e68adcdb9837c7d7ad5
Reviewed-on: https://go-review.googlesource.com/c/go/+/221297
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Emmanuel Odeke <emm.odeke@gmail.com>
5 years agocmd/compile: lower float to uint conversions on s390x
Ruixin(Peter) Bao [Tue, 26 Nov 2019 15:52:43 +0000 (10:52 -0500)]
cmd/compile: lower float to uint conversions on s390x

Add rules for lowering float <-> unsigned int on s390x.

During compilation,
Cvt64Uto64F rule triggers around 80 times,
Cvt64Fto64U rule triggers around 20 times,
Cvt64Uto32F rule triggers around 5 times.

Change-Id: If4c9d128b9132fce8c0bea9abc09cb43a5df7989
Reviewed-on: https://go-review.googlesource.com/c/go/+/209177
Reviewed-by: Michael Munday <mike.munday@ibm.com>
Run-TryBot: Michael Munday <mike.munday@ibm.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

5 years agocmd/compile: add more amd64 constant simplifications
Josh Bleecher Snyder [Fri, 28 Feb 2020 02:56:28 +0000 (18:56 -0800)]
cmd/compile: add more amd64 constant simplifications

More minor optimization opportunities from CL 220499.

Change-Id: Ic4f34c41ed8ab0fce227ac194731c1be12c602db
Reviewed-on: https://go-review.googlesource.com/c/go/+/221608
Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
5 years agocmd/compile: constant fold SSA bool to int conversions
Josh Bleecher Snyder [Sat, 29 Feb 2020 15:07:56 +0000 (07:07 -0800)]
cmd/compile: constant fold SSA bool to int conversions

Shaves off a few instructions here and there.

file                        before   after    Δ       %
go/types.s                  322118   321851   -267    -0.083%
go/internal/gcimporter.s    34937    34909    -28     -0.080%
go/internal/gccgoimporter.s 56493    56474    -19     -0.034%
cmd/compile/internal/ssa.s  3926994  3927177  +183    +0.005%
total                       18862670 18862539 -131    -0.001%

Change-Id: I724f32317b946b5138224808f85709d9c097a247
Reviewed-on: https://go-review.googlesource.com/c/go/+/221428
Reviewed-by: Keith Randall <khr@golang.org>
5 years agocmd/compile: use correct types in phiopt
Josh Bleecher Snyder [Sat, 29 Feb 2020 01:04:16 +0000 (17:04 -0800)]
cmd/compile: use correct types in phiopt

We try to preserve type correctness of generic ops.
phiopt modified a bool to be an int without a conversion.
Add a conversion. There are a few random fluctations in the
generated code as a result, but nothing noteworthy or systematic.

no binary size changes

file                        before   after    Δ       %
math.s                      35966    35961    -5      -0.014%
debug/dwarf.s               108141   108147   +6      +0.006%
crypto/dsa.s                6047     6044     -3      -0.050%
image/png.s                 42882    42885    +3      +0.007%
go/parser.s                 80281    80278    -3      -0.004%
cmd/internal/obj.s          115116   115113   -3      -0.003%
go/types.s                  322130   322118   -12     -0.004%
cmd/internal/obj/arm64.s    151679   151685   +6      +0.004%
go/internal/gccgoimporter.s 56487    56493    +6      +0.011%
cmd/test2json.s             1650     1647     -3      -0.182%
cmd/link/internal/loadelf.s 35442    35443    +1      +0.003%
cmd/go/internal/work.s      305039   305035   -4      -0.001%
cmd/link/internal/ld.s      544835   544834   -1      -0.000%
net/http.s                  558777   558774   -3      -0.001%
cmd/compile/internal/ssa.s  3926551  3926994  +443    +0.011%
cmd/compile/internal/gc.s   1552320  1552321  +1      +0.000%
total                       18862241 18862670 +429    +0.002%

Change-Id: I4289e773be6be534ea3f907d68f614441b8f9b46
Reviewed-on: https://go-review.googlesource.com/c/go/+/221607
Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com>
Reviewed-by: Keith Randall <khr@golang.org>
5 years agoruntime: do not exit(2) if a Go built DLL receives a signal
martin [Fri, 13 Dec 2019 00:03:04 +0000 (16:03 -0800)]
runtime: do not exit(2) if a Go built DLL receives a signal

Fixes #35965

Change-Id: I172501fc0b29595e59b058f6e30f31efe5f6d1f9
Reviewed-on: https://go-review.googlesource.com/c/go/+/211139
Run-TryBot: Emmanuel Odeke <emm.odeke@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Alex Brainman <alex.brainman@gmail.com>
Reviewed-by: Emmanuel Odeke <emm.odeke@gmail.com>
5 years agointernal/cpu: add MIPS64x feature detection
Meng Zhuo [Fri, 28 Feb 2020 12:14:18 +0000 (20:14 +0800)]
internal/cpu: add MIPS64x feature detection

Change-Id: Iacdad1758aa15e4703fccef38c08ecb338b95fd7
Reviewed-on: https://go-review.googlesource.com/c/go/+/200579
Run-TryBot: Meng Zhuo <mengzhuo1203@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
5 years agocmd/go: avoid matching wildcards rooted outside of available modules
Bryan C. Mills [Fri, 28 Feb 2020 20:03:54 +0000 (15:03 -0500)]
cmd/go: avoid matching wildcards rooted outside of available modules

To avoid confusion, also distinguish between packages and dirs in
search.Match results.

No test because this is technically only a performance optimization:
it would be very difficult to write such a test so that it would not
be flaky. (However, tested the change manually.)

Fixes #37521

Change-Id: I17b443699ce6a8f3a63805a7ef0be806f695a4b3
Reviewed-on: https://go-review.googlesource.com/c/go/+/221544
Run-TryBot: Bryan C. Mills <bcmills@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Jay Conrod <jayconrod@google.com>
5 years agodoc: update Effective Go with 1.14 language changes
Felix Cornelius [Fri, 28 Feb 2020 20:15:02 +0000 (20:15 +0000)]
doc: update Effective Go with 1.14 language changes

Fixes #37560

Change-Id: Iccb8e53254c45d203c1b42ea9b4d8509b93dd7a9
GitHub-Last-Rev: 5972b67e5d7b3b36ce1854ee9365197e78f654cd
GitHub-Pull-Request: golang/go#37563
Reviewed-on: https://go-review.googlesource.com/c/go/+/221429
Reviewed-by: Rob Pike <r@golang.org>
5 years agocmd/go: rationalize errors in internal/load and internal/modload
Bryan C. Mills [Mon, 8 Jul 2019 22:13:23 +0000 (18:13 -0400)]
cmd/go: rationalize errors in internal/load and internal/modload

This change is a non-minimal fix for #32917, but incidentally fixes
several other bugs and makes the error messages much more ergonomic.

Updates #32917
Updates #27122
Updates #28459
Updates #29280
Updates #30590
Updates #37214
Updates #36173
Updates #36587
Fixes #36008
Fixes #30992

Change-Id: Iedb26d2e0963697c130df5d0f72e7f83ec2dcf06
Reviewed-on: https://go-review.googlesource.com/c/go/+/185345
Reviewed-by: Michael Matloob <matloob@golang.org>
Reviewed-by: Jay Conrod <jayconrod@google.com>
5 years agocmd/go/internal/search: consolidate package-pattern predicates into Match methods
Bryan C. Mills [Thu, 27 Feb 2020 20:52:32 +0000 (15:52 -0500)]
cmd/go/internal/search: consolidate package-pattern predicates into Match methods

This change consolidates predicates currently scattered throughout
various parts of the package and module loader into methods on the
search.Match type.

That not only makes them more concise, but also encourages
consistency, both in the code and in reasoning about the kinds of
patterns that need to be handled. (For example, the IsLocal predicate
was previously two different calls, either of which could be easily
forgotten at a given call site.)

Factored out from CL 185344 and CL 185345.

Updates #32917

Change-Id: Ifa450ffaf6101f673e0ed69ced001a487d6f9335
Reviewed-on: https://go-review.googlesource.com/c/go/+/221458
Run-TryBot: Bryan C. Mills <bcmills@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Michael Matloob <matloob@golang.org>
Reviewed-by: Jay Conrod <jayconrod@google.com>
5 years agocmd/go/internal/search: record errors in the Match struct
Bryan C. Mills [Mon, 8 Jul 2019 22:11:23 +0000 (18:11 -0400)]
cmd/go/internal/search: record errors in the Match struct

Previously, we would either invoke base.Fatalf (which is too aggressive),
or log.Print (which is too passive).

Updates #32917

Change-Id: I5475e873e76948de7df65dca08bc0ce67a7fc827
Reviewed-on: https://go-review.googlesource.com/c/go/+/185344
Run-TryBot: Bryan C. Mills <bcmills@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Jay Conrod <jayconrod@google.com>
5 years agocmd/go/internal/modload: make AmbiguousImportError an ImportPathError
Michael Matloob [Thu, 27 Feb 2020 21:18:56 +0000 (16:18 -0500)]
cmd/go/internal/modload: make AmbiguousImportError an ImportPathError

AmbiguousImportErrors will now be formatted like other ImportPathErrors:
this means that now the ambiguously imported package won't be printed
twice. Whereas the error message looked like the following:

can't load package: package example.com/m/importy: ambiguous import: found package example.com/m/importy in multiple directories:
$WORK/importy
$WORK/vendor/example.com/m/importy

It now looks like this:

can't load package: ambiguous import: found package example.com/m/importy in multiple directories:
$WORK/importy
$WORK/vendor/example.com/m/importy

Change-Id: I52a2074a6b3f5eb7d78d331d0852b7ea6b3735e6
Reviewed-on: https://go-review.googlesource.com/c/go/+/221457
Run-TryBot: Michael Matloob <matloob@golang.org>
Run-TryBot: Bryan C. Mills <bcmills@google.com>
Reviewed-by: Bryan C. Mills <bcmills@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

5 years agonet: report port number correctly in Plan 9 error
Ian Lance Taylor [Thu, 27 Feb 2020 15:36:39 +0000 (07:36 -0800)]
net: report port number correctly in Plan 9 error

The code was incorrectly using a string conversion of a numeric port
to display the port number.

No test because as far as I can tell this code is only executed if
there is some error in a /net file.

Updates #32479

Change-Id: I0b8deebbf3c0b7cb1e1eee0fd059505f3f4c1623
Reviewed-on: https://go-review.googlesource.com/c/go/+/221377
Run-TryBot: Ian Lance Taylor <iant@golang.org>
Reviewed-by: Bryan C. Mills <bcmills@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

5 years agocmd/compile: add dedicated ARM64BitField aux type
Josh Bleecher Snyder [Tue, 18 Feb 2020 01:47:34 +0000 (17:47 -0800)]
cmd/compile: add dedicated ARM64BitField aux type

The goal here is improved AuxInt printing in ssa.html.
Instead of displaying an inscrutable encoded integer,
it displays something like

v25 (28) = UBFX <int> [lsb=4,width=8] v52

which is much nicer for debugging.

Change-Id: I40713ff7f4a857c4557486cdf73c2dff137511ca
Reviewed-on: https://go-review.googlesource.com/c/go/+/221420
Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
5 years agocmd/compile: improve SignExt32to64 on riscv64
Joel Sing [Tue, 25 Feb 2020 17:00:10 +0000 (04:00 +1100)]
cmd/compile: improve SignExt32to64 on riscv64

SignExt32to64 can be implemented with a single ADDIW instruction, rather than
the two shifts that are in use currently.

Change-Id: Ie1bbaef4018f1ba5162773fc64fa5a887457cfc9
Reviewed-on: https://go-review.googlesource.com/c/go/+/220922
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

5 years agocmd/gofmt, go/format: sync internal.go
Dmitri Shuralyov [Fri, 28 Feb 2020 03:53:18 +0000 (22:53 -0500)]
cmd/gofmt, go/format: sync internal.go

Apply CL 40930 to src/cmd/gofmt/internal.go to bring
it into sync with src/go/format/internal.go.

Also revert '\n' back to "\n\n" in one of the comments,
because the previous text was more accurate.
Gofmt replaces the "; " part of "package p; func _() {"
input with two newline characters, not one.

Updates #11844

Change-Id: I6bb8155a931b793311991d3cd8e006a2931b167a
Reviewed-on: https://go-review.googlesource.com/c/go/+/221497
Reviewed-by: Robert Griesemer <gri@golang.org>
5 years agocmd/compile: add ellipsis rule diagnostics to rulegen
Josh Bleecher Snyder [Thu, 23 Jan 2020 22:33:26 +0000 (14:33 -0800)]
cmd/compile: add ellipsis rule diagnostics to rulegen

These detect opportunities to convert a rule to use an ellipsis,
and provide better error messages when something goes wrong.

This change was used to generate all the preceding changes
converting rules to use ellipses. This change is at the end of those
changes rather than the beginning in order to avoid log spam during rule
generation (say during a git bisection).

The preceding changes collectively shrink the cmd/compile binary by ~2.2%.

Part of this detection is also warning when the presence of an
unmentioned aux or auxint could cause conversion to an ellipsis
rule to change the sematics of the rule.

For example:

(Div64 x y) -> (DIV x y)

looks like a promising rule for an ellipsis. However, Div64 has an auxint,
and (on most platforms) DIV does not. An ellipsis rule would keep the
auxint intact, rather than zeroing it, which can infere with CSE.
So this change flags this rule as doing implicit zeroing;
it should be replaced by

(Div64 [a] x y) -> (DIV x y)

which makes it clear that the auxint is being zeroed.

This detection is not foolproof, but it currently has no false positives.
If false positives arise in the future, we will need to gate the output.

Change-Id: Ie21f284579e5d6e75aa304d0deb024d41ede528b
Reviewed-on: https://go-review.googlesource.com/c/go/+/217014
Run-TryBot: Daniel Martí <mvdan@mvdan.cc>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Daniel Martí <mvdan@mvdan.cc>
5 years agocmd/go: version command should error when given bad args
Daniel Martí [Thu, 27 Feb 2020 16:19:06 +0000 (16:19 +0000)]
cmd/go: version command should error when given bad args

For example, 'go version -m' happily gives you Go's own version, even
though the -m flag only makes sense when grabbing the version of a
binary on disk.

Similarly, if any of the directly named files can't be found, the tool
would succeed. That's acceptable if an error is encountered while
walking a large directory, but not when locating a path directly given
by the user.

These added test cases run even in short mode, as 'go build' is not
needed for them.

Change-Id: I7bb40b72853799e31d9f86cc5e999c8d57813eef
Reviewed-on: https://go-review.googlesource.com/c/go/+/221397
Run-TryBot: Daniel Martí <mvdan@mvdan.cc>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Bryan C. Mills <bcmills@google.com>
5 years agocmd/compile: optimize unsigned comparisons to 0/1 on amd64
Josh Bleecher Snyder [Wed, 1 Jan 2020 07:12:33 +0000 (23:12 -0800)]
cmd/compile: optimize unsigned comparisons to 0/1 on amd64

Plus a bonus optimization I noticed while working on this.

There are no functions (besides the rewrite rules) whose text size
increases as a result of this change.

Updates #21439

The following per-package text size stats were generated by parsing the
output of compiling with -S and summing the function size reported on the
STEXT line. This gives a far more accurate picture of the impact
on generated code than merely looking at the object file size changes
or the resulting binary size changes. The latter are below, for reference.

file                                          before  after   Δ       %
runtime.s                                     477257  476417  -840    -0.176%
math.s                                        35985   35976   -9      -0.025%
vendor/golang.org/x/net/dns/dnsmessage.s      87314   87232   -82     -0.094%
debug/dwarf.s                                 108444  108432  -12     -0.011%
regexp.s                                      64535   64467   -68     -0.105%
internal/xcoff.s                              23175   22945   -230    -0.992%
cmd/vendor/golang.org/x/arch/arm/armasm.s     45263   45260   -3      -0.007%
cmd/vendor/golang.org/x/arch/arm64/arm64asm.s 118140  118135  -5      -0.004%
cmd/internal/obj/arm64.s                      151502  151498  -4      -0.003%
cmd/compile/internal/ssa.s                    6061483 6063120 +1637   +0.027%
total                                         9321728 9322112 +384    +0.004%

file      before    after     Δ       %
go        15188916  15184820  -4096   -0.027%
addr2line 4315984   4311888   -4096   -0.095%
cgo       4836088   4831992   -4096   -0.085%
compile   24506008  24493720  -12288  -0.050%
doc       4680952   4676856   -4096   -0.088%
link      6605336   6601240   -4096   -0.062%
pprof     14776756  14772660  -4096   -0.028%
total     135250956 135214092 -36864  -0.027%

Change-Id: I1243a098a08db452f7d1eb0998e241c9b199e2b4
Reviewed-on: https://go-review.googlesource.com/c/go/+/213058
Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
5 years agoflag: update comment to refer to Output, not out
Ian Lance Taylor [Thu, 27 Feb 2020 19:11:48 +0000 (11:11 -0800)]
flag: update comment to refer to Output, not out

The out method was renamed to Output in CL 70391 for #17628 and #21888.

Fixes #37514

Change-Id: I99be47b5030ccbbf10a056df9fcc3c97cb99b015
Reviewed-on: https://go-review.googlesource.com/c/go/+/221383
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Rob Pike <r@golang.org>
5 years agonet/textproto: close channel to signal pipeline event completion
Tim Cooper [Thu, 27 Feb 2020 18:26:36 +0000 (12:26 -0600)]
net/textproto: close channel to signal pipeline event completion

Change-Id: I7e4827b3428b48c67060789a528586a8907ca3db
Reviewed-on: https://go-review.googlesource.com/c/go/+/221418
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
5 years agocmd/go/testdata/script: fix path separator in test to be os-independent
Michael Matloob [Thu, 27 Feb 2020 22:24:48 +0000 (17:24 -0500)]
cmd/go/testdata/script: fix path separator in test to be os-independent

Use ${:} instead of : so to be Windows-friendly.

Change-Id: I5e660af49b0a2ac2da868583cbcb55cf8bf89b68
Reviewed-on: https://go-review.googlesource.com/c/go/+/221461
Run-TryBot: Bryan C. Mills <bcmills@google.com>
Reviewed-by: Bryan C. Mills <bcmills@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

5 years agocmd/go: roll forward "convert TestShadowingLogic to the script framework"
Michael Matloob [Thu, 20 Feb 2020 00:35:58 +0000 (00:35 +0000)]
cmd/go: roll forward "convert TestShadowingLogic to the script framework"

This rolls forward the change golang.org/cl/214431, which was reverted
in golang.org/cl/220217. The cl was broken because
TestVersionControlErrorMessageIncludesCorrectDirectory, which is going
to be removed in golang.org/cl/214429 hadn't been submitted yet.

Original change description:

Part of converting all tests to script framework to improve
test parallelism.

Updates #36320
Updates #17751

Change-Id: I87b3f9acb8575fbcbd58d454b5f9bac4923429b3
Reviewed-on: https://go-review.googlesource.com/c/go/+/220178
Run-TryBot: Michael Matloob <matloob@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Bryan C. Mills <bcmills@google.com>
5 years agocmd/go: convert testCDAndGOPATHAreDifferent to the script framework
Michael Matloob [Mon, 13 Jan 2020 20:24:56 +0000 (15:24 -0500)]
cmd/go: convert testCDAndGOPATHAreDifferent to the script framework

This is a bit complex. There's a driver program to run go with modifications
to the GOPATH used to test Windows.

Also remove the cd method on testgoData, because this was the last function
that used it.

Part of converting all tests to script framework to improve
test parallelism.

Updates #36320
Updates #17751

Change-Id: I3e8e27f37fd3701bd36b6365b128dd73b69181c0
Reviewed-on: https://go-review.googlesource.com/c/go/+/214578
Reviewed-by: Jay Conrod <jayconrod@google.com>
5 years agocmd/go: convert TestLegacyModGet to the script framework
Michael Matloob [Thu, 9 Jan 2020 23:05:38 +0000 (18:05 -0500)]
cmd/go: convert TestLegacyModGet to the script framework

I think this test needs to be split up eventually. It's one of
the longest tests.

Part of converting all tests to script framework to improve
test parallelism.

Updates #36320
Updates #17751

Change-Id: If2168fac040d78fd0ec3dcbdef2affd2a8f48f6d
Reviewed-on: https://go-review.googlesource.com/c/go/+/214158
Reviewed-by: Jay Conrod <jayconrod@google.com>
5 years agocmd/go: convert TestLinkXImportPathEscape to the script framework
Michael Matloob [Mon, 13 Jan 2020 20:39:20 +0000 (15:39 -0500)]
cmd/go: convert TestLinkXImportPathEscape to the script framework

Part of converting all tests to script framework to improve
test parallelism.

Updates #36320
Updates #17751

Change-Id: Ib386838081abad8bc6b01c1f0a4656553d0b6ff3
Reviewed-on: https://go-review.googlesource.com/c/go/+/214579
Reviewed-by: Jay Conrod <jayconrod@google.com>
5 years agocmd/go: convert some tests in vendor_test to the script framework
Michael Matloob [Mon, 6 Jan 2020 20:33:38 +0000 (15:33 -0500)]
cmd/go: convert some tests in vendor_test to the script framework

Part of converting all tests to script framework to improve
test parallelism.

Updates #36320
Updates #17751

Change-Id: I601e0fcee32b8c5bf2107b520d1dfbe12a19ad3f
Reviewed-on: https://go-review.googlesource.com/c/go/+/213223
Reviewed-by: Jay Conrod <jayconrod@google.com>
5 years agocmd/go: convert TestGoBuildGOPATHOrder to the script framework
Michael Matloob [Mon, 13 Jan 2020 18:51:14 +0000 (13:51 -0500)]
cmd/go: convert TestGoBuildGOPATHOrder to the script framework

It looks like TestGoBuildGOPATHOrderBroken has been fixed so I've converted
that too, without the skip.

Part of converting all tests to script framework to improve
test parallelism.

Updates #36320
Updates #17751

Change-Id: I7ee77f22fb888811c175bcdc5eb814c80fbec420
Reviewed-on: https://go-review.googlesource.com/c/go/+/214432
Reviewed-by: Jay Conrod <jayconrod@google.com>
5 years agocmd/go: convert more tests to the script framework
Michael Matloob [Mon, 13 Jan 2020 16:30:52 +0000 (11:30 -0500)]
cmd/go: convert more tests to the script framework

Part of converting all tests to script framework to improve
test parallelism.

Updates #36320
Updates #17751

Change-Id: I394844da1ffc0dcde7f5862c41ed8efa7c5ca088
Reviewed-on: https://go-review.googlesource.com/c/go/+/214429
Reviewed-by: Jay Conrod <jayconrod@google.com>
5 years agocmd/go: convert TestGoTestRaceInstallsCgo to script framework
Michael Matloob [Mon, 13 Jan 2020 16:11:06 +0000 (11:11 -0500)]
cmd/go: convert TestGoTestRaceInstallsCgo to script framework

Part of converting all tests to script framework to improve
test parallelism.

Updates #36320
Updates #17751

Change-Id: I9a99aa5d37300c83a2f95fb906949cb4c1d5356f
Reviewed-on: https://go-review.googlesource.com/c/go/+/214426
Reviewed-by: Jay Conrod <jayconrod@google.com>
5 years agocmd/go: convert TestGoBuildARM to the script framework
Michael Matloob [Mon, 13 Jan 2020 15:54:34 +0000 (10:54 -0500)]
cmd/go: convert TestGoBuildARM to the script framework

Part of converting all tests to script framework to improve
test parallelism.

Updates #36320
Updates #17751

Change-Id: If1e591f28d6399a07b37ed7f4a1419bf7cd915eb
Reviewed-on: https://go-review.googlesource.com/c/go/+/214425
Reviewed-by: Jay Conrod <jayconrod@google.com>
5 years agocmd/go: convert TestGoTestBuildsAnXtestContainingOnly... to the script framework
Michael Matloob [Mon, 13 Jan 2020 15:43:09 +0000 (10:43 -0500)]
cmd/go: convert TestGoTestBuildsAnXtestContainingOnly... to the script framework

The name of the test is too long to fit on the first line. It's
TestGoTestBuildsAnXtestContainingOnlyNonRunnableExamples.

Part of converting all tests to script framework to improve
test parallelism.

Updates #36320
Updates #17751

Change-Id: I490748271b10a85cbe1d34f9dbecb86ccf0101a7
Reviewed-on: https://go-review.googlesource.com/c/go/+/214423
Reviewed-by: Jay Conrod <jayconrod@google.com>
5 years agocmd/go: convert TestPackageNotStaleWithTrailingSlash to script framework
Michael Matloob [Fri, 10 Jan 2020 23:15:38 +0000 (18:15 -0500)]
cmd/go: convert TestPackageNotStaleWithTrailingSlash to script framework

Part of converting all tests to script framework to improve
test parallelism.

Updates #36320
Updates #17751

Change-Id: I1020feaa4ddb40ff52c46728bc4973cea4c7b066
Reviewed-on: https://go-review.googlesource.com/c/go/+/214391
Reviewed-by: Jay Conrod <jayconrod@google.com>
5 years agocmd/compile: ignore div/mod in prove on non-x86 architectures
Josh Bleecher Snyder [Mon, 24 Feb 2020 21:53:53 +0000 (13:53 -0800)]
cmd/compile: ignore div/mod in prove on non-x86 architectures

Instead of writing AuxInt during prove and then zeroing it during lower,
just don't write it in the first place.

Passes toolstash-check -all.

Change-Id: Iea4b555029a9d69332e835536f9cf3a42b8223db
Reviewed-on: https://go-review.googlesource.com/c/go/+/220682
Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
5 years agocmd/compile: document Move's type
Josh Bleecher Snyder [Tue, 25 Feb 2020 00:16:27 +0000 (16:16 -0800)]
cmd/compile: document Move's type

Fixes #37381

Change-Id: I8abf07d6342c10fc8d52e11c6a70fb0ec09220d2
Reviewed-on: https://go-review.googlesource.com/c/go/+/220683
Reviewed-by: Cherry Zhang <cherryyz@google.com>
5 years agocmd/go/internal/modload: make PackageNotInModuleError reasonable for the Target module
Bryan C. Mills [Wed, 3 Jul 2019 21:12:32 +0000 (17:12 -0400)]
cmd/go/internal/modload: make PackageNotInModuleError reasonable for the Target module

Updates #28459
Updates #32917

Change-Id: Iced562cb7c2e0ac075d8345f1e4ad3b073842dcf
Reviewed-on: https://go-review.googlesource.com/c/go/+/185343
Run-TryBot: Bryan C. Mills <bcmills@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Jay Conrod <jayconrod@google.com>
5 years agocmd/compile: make inlining intrinsics cost one
Josh Bleecher Snyder [Fri, 14 Feb 2020 19:42:58 +0000 (11:42 -0800)]
cmd/compile: make inlining intrinsics cost one

We were assigning a cost of 2 for intrinsics:
One when we recognized an intrinsic,
and one for the OCALLFUNC node.

I believe that the intent was that intrinsics should
cost 1, since they are typically an arithmetic op,
and because they tend to occur in performance-sensitive code.
(Not that any of this is particularly principled right now.)

Stop charging when we recognize an intrinsic;
let the OCALLFUNC node cover the cost.

This has a negligible impact on std+cmd.

Change-Id: Ie6ae5c18da3ae7e40aec425aed95c6999c2831a1
Reviewed-on: https://go-review.googlesource.com/c/go/+/221357
Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
5 years agocmd/compile: check rune type bounds as int32, not uint32
Ian Lance Taylor [Thu, 27 Feb 2020 19:08:30 +0000 (11:08 -0800)]
cmd/compile: check rune type bounds as int32, not uint32

Also, avoid string(i) where i has type int.

Updates #32479

Change-Id: If3c6edc8523860082726e034ef9e887b5f7fabd1
Reviewed-on: https://go-review.googlesource.com/c/go/+/221382
Run-TryBot: Ian Lance Taylor <iant@golang.org>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

5 years agoall: fix incorrect channel and API usage in some unit tests
Ziheng Liu [Thu, 13 Feb 2020 21:20:30 +0000 (16:20 -0500)]
all: fix incorrect channel and API usage in some unit tests

This CL changes some unit test functions, making sure that these tests (and goroutines spawned during test) won't block.
Since they are just test functions, I use one CL to fix them all. I hope this won't cause trouble to reviewers and can save time for us.
There are three main categories of incorrect logic fixed by this CL:
1. Use testing.Fatal()/Fatalf() in spawned goroutines, which is forbidden by Go's document.
2. Channels are used in such a way that, when errors or timeout happen, the test will be blocked and never return.
3. Channels are used in such a way that, when errors or timeout happen, the test can return but some spawned goroutines will be leaked, occupying resource until all other tests return and the process is killed.

Change-Id: I3df931ec380794a0cf1404e632c1dd57c65d63e8
Reviewed-on: https://go-review.googlesource.com/c/go/+/219380
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
5 years agocmd/compile: handle more cases in isNonNegative
Josh Bleecher Snyder [Fri, 20 Dec 2019 21:42:39 +0000 (13:42 -0800)]
cmd/compile: handle more cases in isNonNegative

The gains from this aren't particularly impressive.
Still, it is cheap and easy, and
it will keep me from wondering about whether it
might help to add X every time I look at this function.

This updated function is pretty exhaustive;
I examined every op encountered in a call to isNonNegative
when compiling all the stuff hanging around in my GOPATH,
for both 386 and amd64.

(32 bit architectures were somewhat neglected before.)

Object file size impact, 64 bit:

file                       before    after     Δ       %
archive/zip.a              359352    359284    -68     -0.019%
cmd/compile/internal/ssa.a 30715960  30717526  +1566   +0.005%
cmd/internal/obj/arm64.a   2972532   2972440   -92     -0.003%
cmd/internal/obj/riscv.a   297714    297672    -42     -0.014%
debug/dwarf.a              656336    655346    -990    -0.151%
debug/gosym.a              183352    183122    -230    -0.125%
encoding/gob.a             901130    900798    -332    -0.037%
image/gif.a                171884    171890    +6      +0.003%
internal/trace.a           506930    507270    +340    +0.067%
math.a                     233506    233490    -16     -0.007%
reflect.a                  1431740   1431476   -264    -0.018%
runtime.a                  3854480   3854332   -148    -0.004%
unicode/utf16.a            8920      8980      +60     +0.673%
total                      133000610 133000400 -210    -0.000%

Object file size impact, 32 bit:

file                                          before    after     Δ       %
archive/zip.a                                 330794    329640    -1154   -0.349%
cmd/compile/internal/gc.a                     8090204   8090026   -178    -0.002%
cmd/compile/internal/ssa.a                    29392460  29393890  +1430   +0.005%
cmd/internal/goobj2.a                         189512    189492    -20     -0.011%
cmd/internal/obj/arm64.a                      2444942   2444860   -82     -0.003%
cmd/internal/obj/riscv.a                      272848    272806    -42     -0.015%
cmd/link/internal/loader.a                    388548    388544    -4      -0.001%
cmd/link/internal/loadpe.a                    158776    158684    -92     -0.058%
cmd/vendor/golang.org/x/arch/ppc64/ppc64asm.a 511824    511316    -508    -0.099%
cmd/vendor/golang.org/x/arch/x86/x86asm.a     512812    512704    -108    -0.021%
cmd/vendor/golang.org/x/sys/unix.a            942422    942218    -204    -0.022%
compress/bzip2.a                              88768     88680     -88     -0.099%
crypto/tls.a                                  1655542   1655396   -146    -0.009%
debug/dwarf.a                                 608520    605822    -2698   -0.443%
debug/gosym.a                                 168282    168276    -6      -0.004%
debug/pe.a                                    173146    173108    -38     -0.022%
encoding/gob.a                                797978    797724    -254    -0.032%
encoding/hex.a                                44080     44020     -60     -0.136%
image/gif.a                                   152142    152148    +6      +0.004%
internal/xcoff.a                              186480    185834    -646    -0.346%
math.a                                        257866    257854    -12     -0.005%
net/http.a                                    3588246   3588150   -96     -0.003%
net/textproto.a                               162384    162120    -264    -0.163%
reflect.a                                     1316204   1316058   -146    -0.011%
regexp.a                                      373346    373248    -98     -0.026%
runtime/pprof.a                               345318    345088    -230    -0.067%
runtime.a                                     3513902   3513714   -188    -0.005%
syscall.a                                     781406    781018    -388    -0.050%
time.a                                        483814    483750    -64     -0.013%
unicode/utf16.a                               8394      8364      -30     -0.357%
vendor/golang.org/x/crypto/cryptobyte.a       287100    286706    -394    -0.137%
vendor/golang.org/x/net/route.a               175042    174724    -318    -0.182%
total                                         121677354 121670234 -7120   -0.006%

Change-Id: Ie672752feb5e94dd151836f852181980710e820d
Reviewed-on: https://go-review.googlesource.com/c/go/+/212777
Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
5 years agonet: fix typo in unexported variable name
Mark Rushakoff [Thu, 27 Feb 2020 05:21:22 +0000 (05:21 +0000)]
net: fix typo in unexported variable name

Change-Id: Idf64716ef8cc3ceee2fc94462cba33ba60049f7b
GitHub-Last-Rev: ba550c5d136688e5aeaf3c9d4713b7f29bd17492
GitHub-Pull-Request: golang/go#36237
Reviewed-on: https://go-review.googlesource.com/c/go/+/212300
Run-TryBot: Emmanuel Odeke <emm.odeke@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Emmanuel Odeke <emm.odeke@gmail.com>
5 years agoruntime: don't panic on racy use of timers
Ian Lance Taylor [Wed, 26 Feb 2020 04:23:15 +0000 (20:23 -0800)]
runtime: don't panic on racy use of timers

If we see a racy use of timers, as in concurrent calls to Timer.Reset,
do the operations in an unpredictable order, rather than crashing.

Fixes #37400

Change-Id: Idbac295df2dfd551b6d762909d5040fc532c1b34
Reviewed-on: https://go-review.googlesource.com/c/go/+/221077
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Michael Knyszek <mknyszek@google.com>
5 years agocmd/dist: enable cgo and PIE tests on android/arm64
Cherry Zhang [Wed, 26 Feb 2020 15:24:39 +0000 (10:24 -0500)]
cmd/dist: enable cgo and PIE tests on android/arm64

Now that android/arm64 supports internal linking PIE, enable the
test. While here, I realized that some cgo tests are also not
enabled on android/arm64. Enable them as well. Let's see if it
works.

Change-Id: Ibf186fe402ebf0bbec82873fd56d0eb752b48180
Reviewed-on: https://go-review.googlesource.com/c/go/+/221099
Run-TryBot: Cherry Zhang <cherryyz@google.com>
Reviewed-by: Elias Naur <mail@eliasnaur.com>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
5 years agocmd/link: reenable internal linking PIE on linux/amd64 and linux/arm64
Cherry Zhang [Wed, 26 Feb 2020 15:08:30 +0000 (10:08 -0500)]
cmd/link: reenable internal linking PIE on linux/amd64 and linux/arm64

It was enabled in CL 207877, but then accidentally disabled in
CL 207299 due to a bad rebase. Reenable.

Change-Id: I147bf724a4263d4aae54576a36cc7e1cad5e8a2c
Reviewed-on: https://go-review.googlesource.com/c/go/+/221098
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Elias Naur <mail@eliasnaur.com>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
5 years agoio/ioutil: reject path separators in TempDir, TempFile pattern
Constantin Konstantinidis [Wed, 25 Dec 2019 16:24:07 +0000 (17:24 +0100)]
io/ioutil: reject path separators in TempDir, TempFile pattern

Fixes #33920

Change-Id: I2351a1caa80c086ff5a8e02aad70d996be7aac35
Reviewed-on: https://go-review.googlesource.com/c/go/+/212597
Reviewed-by: Emmanuel Odeke <emm.odeke@gmail.com>
Reviewed-by: Robert Griesemer <gri@golang.org>
Run-TryBot: Emmanuel Odeke <emm.odeke@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

5 years agobuild: in clean.bash, look for 'go' in GOROOT/bin instead of GOBIN
Nikson Kanti Paul [Tue, 21 Jan 2020 00:50:27 +0000 (01:50 +0100)]
build: in clean.bash, look for 'go' in GOROOT/bin instead of GOBIN

Updates #14340
Updates #32674
Fixes #36659

Change-Id: I5bfaba4e53dab894d113dd5065794d66f1a25f2b
Reviewed-on: https://go-review.googlesource.com/c/go/+/215478
Reviewed-by: Bryan C. Mills <bcmills@google.com>
5 years agocmd/compile: use immediates for MOVO of readonly sym
Josh Bleecher Snyder [Sun, 23 Feb 2020 06:05:05 +0000 (22:05 -0800)]
cmd/compile: use immediates for MOVO of readonly sym

file      before    after     Δ       %
addr2line 4307760   4303616   -4144   -0.096%
api       5968600   5968568   -32     -0.001%
asm       5109928   5122120   +12192  +0.239%
buildid   2843752   2839608   -4144   -0.146%
cgo       4823768   4819624   -4144   -0.086%
compile   20687848  20691800  +3952   +0.019%
cover     5259896   5255752   -4144   -0.079%
dist      3665176   3661032   -4144   -0.113%
doc       4668648   4668600   -48     -0.001%
fix       3368792   3368744   -48     -0.001%
link      6613328   6609200   -4128   -0.062%
nm        4253312   4253280   -32     -0.001%
objdump   4655360   4655312   -48     -0.001%
pack      2294312   2294280   -32     -0.001%
pprof     14747332  14747284  -48     -0.000%
test2json 2819416   2815272   -4144   -0.147%
trace     11669436  11665292  -4144   -0.036%
vet       8274184   8270040   -4144   -0.050%
total     116030848 116009424 -21424  -0.018%

Change-Id: Ice37222c4d76540b3591459f605321cbf142872d
Reviewed-on: https://go-review.googlesource.com/c/go/+/220690
Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
5 years agocmd/compile: mark Lsyms as readonly earlier
Josh Bleecher Snyder [Mon, 17 Feb 2020 01:00:52 +0000 (17:00 -0800)]
cmd/compile: mark Lsyms as readonly earlier

The SSA backend has rules to read the contents of readonly Lsyms.
However, this rule was failing to trigger for many readonly Lsyms.
This is because the readonly attribute that was set on the Node.Name
was not propagated to its Lsym until the dump globals phase, after SSA runs.

To work around this phase ordering problem, introduce Node.SetReadonly,
which sets Node.Name.Readonly and also configures the Lsym
enough that SSA can use it.

This change also fixes a latent problem in the rewrite rule function,
namely that reads past the end of lsym.P were treated as entirely zero,
instead of merely requiring padding with trailing zeros.

This change also adds an amd64 rule needed to fully optimize
the results of this change. It would be better not to need this,
but the zero extension that should handle this for us
gets optimized away too soon (see #36897 for a similar problem).
I have not investigated whether other platforms also need new
rules to take full advantage of the new optimizations.

Compiled code for (interface{})(true) on amd64 goes from:

LEAQ type.bool(SB), AX
MOVBLZX ""..stmp_0(SB), BX
LEAQ runtime.staticbytes(SB), CX
ADDQ CX, BX

to

LEAQ type.bool(SB), AX
LEAQ runtime.staticbytes+1(SB), BX

Prior to this change, the readonly symbol rewrite rules
fired a total of 884 times during make.bash.
Afterwards they fire 1807 times.

file    before    after     Δ       %
cgo     4827832   4823736   -4096   -0.085%
compile 24907768  24895656  -12112  -0.049%
fix     3376952   3368760   -8192   -0.243%
pprof   14751700  14747604  -4096   -0.028%
total   120343528 120315032 -28496  -0.024%

Change-Id: I59ea52138276c37840f69e30fb109fd376d579ec
Reviewed-on: https://go-review.googlesource.com/c/go/+/220499
Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
5 years agocmd/compile: make clobber variadic
Josh Bleecher Snyder [Tue, 25 Feb 2020 01:30:44 +0000 (17:30 -0800)]
cmd/compile: make clobber variadic

There are often many values to clobber.
Allow passing them all in at once.
The goal is increased rule readability.
As a bonus, it shrinks cmd/compile by ~97k, almost half a percent.
Package SSA requires 1.2% less memory to compile.

The single-line changes were make via regex,
and the remaining multi-line clobbers were manually combined.

Passes toolstash-check -all.

Change-Id: Ib310e9265d3616211f8192c9040b4c8933824d19
Reviewed-on: https://go-review.googlesource.com/c/go/+/220691
Reviewed-by: Michael Munday <mike.munday@ibm.com>
5 years agocmd/link/internal/ld: bump NetBSD ABI version to 7.0
Tobias Klauser [Tue, 25 Feb 2020 09:18:55 +0000 (10:18 +0100)]
cmd/link/internal/ld: bump NetBSD ABI version to 7.0

According to https://golang.org/wiki/NetBSD, NetBSD 7.0 is supported as
of Go 1.3 (with Go 1.5 recommended). NetBSD 6.0 was last supported in Go
1.9.7. Thus, bump the minimal ABI version to NetBSD 7.0

Suggested by Benny Siegert in CL 212461.

Change-Id: I11dd14c6b835ac9fc156880b2551f71893c7f267
Reviewed-on: https://go-review.googlesource.com/c/go/+/220428
Reviewed-by: Benny Siegert <bsiegert@gmail.com>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
5 years agocmd/compile: constant fold SHLxconst of a constant on amd64
Josh Bleecher Snyder [Sun, 23 Feb 2020 05:47:49 +0000 (21:47 -0800)]
cmd/compile: constant fold SHLxconst of a constant on amd64

These rules fire in particular when comparing to a constant
string of length two. They should trigger even more after CL 220499.

file    before    after     Δ       %
compile 20639976  20635880  -4096   -0.020%
total   116003456 115999360 -4096   -0.004%

Change-Id: I21c1c02cf32d710d7a4eb12efab00f02796ccb84
Reviewed-on: https://go-review.googlesource.com/c/go/+/220694
Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
5 years agocmd/compile: merge const into storeidx on amd64
Josh Bleecher Snyder [Sun, 23 Feb 2020 04:05:12 +0000 (20:05 -0800)]
cmd/compile: merge const into storeidx on amd64

file    before    after     Δ       %
compile 20652264  20639976  -12288  -0.059%
trace   11673532  11669436  -4096   -0.035%
total   116019840 116003456 -16384  -0.014%

Change-Id: Id0522e08f10e77c885fba1d0d9b65f8981a647ac
Reviewed-on: https://go-review.googlesource.com/c/go/+/220693
Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>