]> Cypherpunks repositories - gostls13.git/log
gostls13.git
4 years agotime: enable system zoneinfo on macOS/ARM64
Cherry Zhang [Wed, 7 Oct 2020 00:36:52 +0000 (20:36 -0400)]
time: enable system zoneinfo on macOS/ARM64

Updates #38485.

Change-Id: I4a8b509dc4ad03706235289fbe8c2a675453c871
Reviewed-on: https://go-review.googlesource.com/c/go/+/260339
Trust: Cherry Zhang <cherryyz@google.com>
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
4 years agodoc/go1.16: document GO386=387 and GO386=softfloat
Dmitri Shuralyov [Wed, 7 Oct 2020 17:42:49 +0000 (13:42 -0400)]
doc/go1.16: document GO386=387 and GO386=softfloat

Also add a few more TODOs as found by the relnote command.
It's an incomplete list due to #41849.

For #40700.

Change-Id: Id17a9be86d3338e1fcb281d26e7298ff26e92864
Reviewed-on: https://go-review.googlesource.com/c/go/+/260337
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Trust: Dmitri Shuralyov <dmitshur@golang.org>

4 years agonet/http/pprof: use Request.Context, not the deprecated CloseNotifier
Ayan George [Tue, 6 Oct 2020 18:40:40 +0000 (18:40 +0000)]
net/http/pprof: use Request.Context, not the deprecated CloseNotifier

Prior to this commit, the profiling code had a sleep() function that
waits and unblocks on either time.After() or a channel provided by an
http.CloseNotifier derived from a supplied http.ResponseWriter.

According to the documentation, http.CloseNotifier is deprecated:

  Deprecated: the CloseNotifier interface predates Go's context package.
  New code should use Request.Context instead.

This patch does just that -- sleep() now takes an *http.Request and uses
http.Request.Context() to signal when a request has been cancelled.

Change-Id: I98702314addf494f5743a4f99172dc607389dbb8
GitHub-Last-Rev: c1e37a03ca28417ed5833618d3eeddb2eecccd09
GitHub-Pull-Request: golang/go#41756
Reviewed-on: https://go-review.googlesource.com/c/go/+/259157
Reviewed-by: Bryan C. Mills <bcmills@google.com>
Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
Reviewed-by: Emmanuel Odeke <emm.odeke@gmail.com>
Trust: Bryan C. Mills <bcmills@google.com>
Trust: Emmanuel Odeke <emm.odeke@gmail.com>
Run-TryBot: Bryan C. Mills <bcmills@google.com>
TryBot-Result: Go Bot <gobot@golang.org>

4 years agosyscall: rewrite Windows makeCmdLine to use []byte
Ian Lance Taylor [Tue, 6 Oct 2020 23:31:00 +0000 (16:31 -0700)]
syscall: rewrite Windows makeCmdLine to use []byte

It's faster to append to a []byte and only convert to string at the
end then it is to build up a string by concatenating characters.

Fixes #41825

Change-Id: I45ddf77dcc62726c919f0533c95d483cee8ba366
Reviewed-on: https://go-review.googlesource.com/c/go/+/259978
Trust: Ian Lance Taylor <iant@golang.org>
Trust: Alex Brainman <alex.brainman@gmail.com>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Alex Brainman <alex.brainman@gmail.com>
4 years agocmd/compile: store call args in the call block
Cherry Zhang [Wed, 7 Oct 2020 15:32:43 +0000 (11:32 -0400)]
cmd/compile: store call args in the call block

We already do this for OpStore, but we didn't do this for OpMove.
Do the same, to ensure that no two memories are live at the same
time.

Fixes #41846.

Change-Id: Iad77ff031b3c4459d1217e0b04aeb0e692eb474d
Reviewed-on: https://go-review.googlesource.com/c/go/+/260237
Trust: Cherry Zhang <cherryyz@google.com>
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: David Chase <drchase@google.com>
4 years agocmd/go: env -w validates GOTMPDIR value
Eugene Kalinin [Mon, 24 Aug 2020 22:49:39 +0000 (01:49 +0300)]
cmd/go: env -w validates GOTMPDIR value

This change makes go env -w check if GOTMPDIR is an absolute path.

If GOTMPDIR is not an absolute and not existing path there will be an
error at every `work.Builder.Init()`. If `go env` has `-u/-w` as
argument `work.Builder.Init()` is not called.

`go env -w GOTMPDIR=` work in the same way as `go env -u GOTMPDIR`.

Fixes #40932

Change-Id: I6b0662302eeace7f20460b6d26c6e59af1111da2
Reviewed-on: https://go-review.googlesource.com/c/go/+/250198
Run-TryBot: Jay Conrod <jayconrod@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Jay Conrod <jayconrod@google.com>
Trust: Bryan C. Mills <bcmills@google.com>
Trust: Jay Conrod <jayconrod@google.com>

4 years agoruntime: restore SSE guard in asyncPreempt on 386
Cherry Zhang [Wed, 7 Oct 2020 02:07:15 +0000 (22:07 -0400)]
runtime: restore SSE guard in asyncPreempt on 386

So we don't use SSE instructions under GO386=softfloat.

Change-Id: I8ecc92340ee567f84a22501df2543ec041d25ef2
Reviewed-on: https://go-review.googlesource.com/c/go/+/260137
Trust: Cherry Zhang <cherryyz@google.com>
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
4 years agointernal/reflectlite: include Kind in ValueError message
Ori Rawlings [Fri, 29 May 2020 03:41:38 +0000 (22:41 -0500)]
internal/reflectlite: include Kind in ValueError message

The implementation has been ported from reflect, but to avoid
introducing a dependency on strconv, Kind.String() falls back to
"invalid" if the Kind is unknown rather than "kind" + strconv.Itoa(int(k))

Fixes #39286

Change-Id: I82277242a6c41d0146dabd9d20339fe72d562500
Reviewed-on: https://go-review.googlesource.com/c/go/+/235522
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Reviewed-by: Dmitri Shuralyov <dmitshur@golang.org>
Trust: Dmitri Shuralyov <dmitshur@golang.org>

4 years agocmd/link: support C-shared buildmode on macOS/ARM64
Cherry Zhang [Sun, 4 Oct 2020 03:58:29 +0000 (23:58 -0400)]
cmd/link: support C-shared buildmode on macOS/ARM64

It just works, after the plugin work.

Updates #38485.

Change-Id: I55aa11b380a33a729fccb731b77f48bc7d0dea2e
Reviewed-on: https://go-review.googlesource.com/c/go/+/259443
Trust: Cherry Zhang <cherryyz@google.com>
Reviewed-by: Than McIntosh <thanm@google.com>
4 years agocmd/link: support PIE on macOS/ARM64
Cherry Zhang [Sun, 4 Oct 2020 03:36:58 +0000 (23:36 -0400)]
cmd/link: support PIE on macOS/ARM64

On macOS/ARM64 everything must be PIE, and we already build PIE
in exe buildmode. Support PIE buildmode as well.

Updates #38485.

Change-Id: I10b68c2f6eb77714e31c26116c61a0e28bf9a358
Reviewed-on: https://go-review.googlesource.com/c/go/+/259442
Trust: Cherry Zhang <cherryyz@google.com>
Reviewed-by: Than McIntosh <thanm@google.com>
4 years agocmd/link: support plugin on macOS/ARM64
Cherry Zhang [Sat, 3 Oct 2020 20:26:37 +0000 (16:26 -0400)]
cmd/link: support plugin on macOS/ARM64

Updates #38485.

Change-Id: I8295f7fad55b1f9701162f9d2902b3499137c64d
Reviewed-on: https://go-review.googlesource.com/c/go/+/259441
Trust: Cherry Zhang <cherryyz@google.com>
Reviewed-by: Than McIntosh <thanm@google.com>
4 years agoruntime/cgo: only build xx_cgo_panicmem on iOS
Cherry Zhang [Sat, 3 Oct 2020 20:18:43 +0000 (16:18 -0400)]
runtime/cgo: only build xx_cgo_panicmem on iOS

On iOS, when running under lldb, we install xx_cgo_panicmem as
EXC_BAD_ACCESS handler so we can get a proper Go panic for
SIGSEGV. Only build it on iOS.

Updates #38485.

Change-Id: I801c477439e05920a4bb8fdf5eae6f4923ab8274
Reviewed-on: https://go-review.googlesource.com/c/go/+/259440
Trust: Cherry Zhang <cherryyz@google.com>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
4 years agoall: implement GO386=softfloat
Keith Randall [Tue, 6 Oct 2020 21:42:15 +0000 (14:42 -0700)]
all: implement GO386=softfloat

Backstop support for non-sse2 chips now that 387 is gone.

RELNOTE=yes

Change-Id: Ib10e69c4a3654c15a03568f93393437e1939e013
Reviewed-on: https://go-review.googlesource.com/c/go/+/260017
Trust: Keith Randall <khr@golang.org>
Run-TryBot: Keith Randall <khr@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
4 years agonet/http: add Transport.GetProxyConnectHeader
Brad Fitzpatrick [Tue, 6 Oct 2020 17:53:11 +0000 (10:53 -0700)]
net/http: add Transport.GetProxyConnectHeader

Fixes golang/go#41048

Change-Id: I38e01605bffb6f85100c098051b0c416dd77f261
Reviewed-on: https://go-review.googlesource.com/c/go/+/259917
Trust: Brad Fitzpatrick <bradfitz@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Damien Neil <dneil@google.com>
4 years agoall: enable more tests on macOS/ARM64
Cherry Zhang [Thu, 17 Sep 2020 19:02:26 +0000 (15:02 -0400)]
all: enable more tests on macOS/ARM64

Updates #38485.

Change-Id: Iac96f5ffe88521fcb11eab306d0df6463bdce046
Reviewed-on: https://go-review.googlesource.com/c/go/+/256920
Trust: Cherry Zhang <cherryyz@google.com>
Reviewed-by: Dmitri Shuralyov <dmitshur@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
4 years agocmd/dist: enable more tests on macOS/ARM64
Cherry Zhang [Thu, 17 Sep 2020 16:39:43 +0000 (12:39 -0400)]
cmd/dist: enable more tests on macOS/ARM64

Unlike iOS, macOS ARM64 is more of a fully featured OS. Enable
more tests.

Updates #38485.

Change-Id: I2e2240c848d21996db2b950a4a6856987f7a652c
Reviewed-on: https://go-review.googlesource.com/c/go/+/256919
Trust: Cherry Zhang <cherryyz@google.com>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
4 years agoruntime: enable more address bits on macOS/ARM64
Cherry Zhang [Thu, 17 Sep 2020 14:53:10 +0000 (10:53 -0400)]
runtime: enable more address bits on macOS/ARM64

Apparently macOS/ARM64 has 47-bit addresses, instead of 33-bit as
on ios/ARM64. Enable more address bits.

Updates #38485.

Change-Id: I8aa64ba22a3933e3d9c4fffd17d902b5f31c30e3
Reviewed-on: https://go-review.googlesource.com/c/go/+/256918
Trust: Cherry Zhang <cherryyz@google.com>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Reviewed-by: Michael Knyszek <mknyszek@google.com>
4 years agoruntime: use sigaltstack on macOS/ARM64
Cherry Zhang [Sat, 12 Sep 2020 16:33:24 +0000 (12:33 -0400)]
runtime: use sigaltstack on macOS/ARM64

Currently we don't use sigaltstack on darwin/arm64, as is not
supported on iOS. However, it is supported on macOS. Use it.
(iOS remains unchanged.)

Change-Id: Icc154c5e2edf2dbdc8ca68741ad9157fc15a72ee
Reviewed-on: https://go-review.googlesource.com/c/go/+/256917
Trust: Cherry Zhang <cherryyz@google.com>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
4 years agocompress/flate: remove unneeded zeroing of bytes array in (*huffmanBitWriter).reset
Dan Scales [Wed, 30 Sep 2020 23:34:47 +0000 (16:34 -0700)]
compress/flate: remove unneeded zeroing of bytes array in (*huffmanBitWriter).reset

There is no correctness reason to zero out the w.bytes array in (w
*huffmanBitWriter).reset, since w.nbytes is correctly set to zero. The elements of
the bytes array are always written sequentially, with nbytes indicating how many
elements have been written, and are only read up to the current value of nybytes.

We have a pprof profile of a web server that compresses its request/responses, and
the zeroing in reset() is taking up 2.6% of the CPU time of the server (and could
be causing more slowdowns elsewhere due to its effects on the cache). This
overhead may be showing up especially because there are many request/responses
that are all fairly small.

I'm not sure if the zeroing of the bytes array was intended as extra protection of
data across reset uses in the same program, but no protection is needed as long as
the huffman_bit_writer code remains correct.

Change-Id: I67f2b2f56cff9dcc38d8fc0aea885bb010aeedbf
Reviewed-on: https://go-review.googlesource.com/c/go/+/258577
Run-TryBot: Dan Scales <danscales@google.com>
Run-TryBot: Joe Tsai <joetsai@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Klaus Post <klauspost@gmail.com>
Reviewed-by: Joe Tsai <joetsai@google.com>
Trust: Joe Tsai <joetsai@google.com>
Trust: Dan Scales <danscales@google.com>

4 years agocmd/compile,cmd/internal/obj/ppc64: use mulli where possible
Lynn Boger [Fri, 2 Oct 2020 21:51:13 +0000 (17:51 -0400)]
cmd/compile,cmd/internal/obj/ppc64: use mulli where possible

This adds support to allow the use of mulli when one of the multiply
operands is a constant that fits in 16 bits.

This especially helps in the case where this instruction appears in
a loop since the load of the constant is not being moved out of the loop.

Some improvements seen in compress/flate on power9:

Decode/Digits/Huffman/1e4         259µs ± 0%     261µs ± 0%   +0.57%  (p=1.000 n=1+1)
Decode/Digits/Huffman/1e5        2.43ms ± 0%    2.45ms ± 0%   +0.79%  (p=1.000 n=1+1)
Decode/Digits/Huffman/1e6        23.9ms ± 0%    24.2ms ± 0%   +0.86%  (p=1.000 n=1+1)
Decode/Digits/Speed/1e4           278µs ± 0%     279µs ± 0%   +0.34%  (p=1.000 n=1+1)
Decode/Digits/Speed/1e5          2.80ms ± 0%    2.81ms ± 0%   +0.29%  (p=1.000 n=1+1)
Decode/Digits/Speed/1e6          28.0ms ± 0%    28.1ms ± 0%   +0.28%  (p=1.000 n=1+1)
Decode/Digits/Default/1e4         278µs ± 0%     278µs ± 0%   +0.28%  (p=1.000 n=1+1)
Decode/Digits/Default/1e5        2.68ms ± 0%    2.69ms ± 0%   +0.19%  (p=1.000 n=1+1)
Decode/Digits/Default/1e6        26.6ms ± 0%    26.6ms ± 0%   +0.21%  (p=1.000 n=1+1)
Decode/Digits/Compression/1e4     278µs ± 0%     278µs ± 0%   +0.00%  (p=1.000 n=1+1)
Decode/Digits/Compression/1e5    2.68ms ± 0%    2.69ms ± 0%   +0.21%  (p=1.000 n=1+1)
Decode/Digits/Compression/1e6    26.6ms ± 0%    26.6ms ± 0%   +0.07%  (p=1.000 n=1+1)
Decode/Newton/Huffman/1e4         322µs ± 0%     312µs ± 0%   -2.84%  (p=1.000 n=1+1)
Decode/Newton/Huffman/1e5        3.11ms ± 0%    2.91ms ± 0%   -6.41%  (p=1.000 n=1+1)
Decode/Newton/Huffman/1e6        31.4ms ± 0%    29.3ms ± 0%   -6.85%  (p=1.000 n=1+1)
Decode/Newton/Speed/1e4           282µs ± 0%     269µs ± 0%   -4.69%  (p=1.000 n=1+1)
Decode/Newton/Speed/1e5          2.29ms ± 0%    2.20ms ± 0%   -4.13%  (p=1.000 n=1+1)
Decode/Newton/Speed/1e6          22.7ms ± 0%    21.3ms ± 0%   -6.06%  (p=1.000 n=1+1)
Decode/Newton/Default/1e4         254µs ± 0%     237µs ± 0%   -6.60%  (p=1.000 n=1+1)
Decode/Newton/Default/1e5        1.86ms ± 0%    1.75ms ± 0%   -5.99%  (p=1.000 n=1+1)
Decode/Newton/Default/1e6        18.1ms ± 0%    17.4ms ± 0%   -4.10%  (p=1.000 n=1+1)
Decode/Newton/Compression/1e4     254µs ± 0%     244µs ± 0%   -3.91%  (p=1.000 n=1+1)
Decode/Newton/Compression/1e5    1.85ms ± 0%    1.79ms ± 0%   -3.10%  (p=1.000 n=1+1)
Decode/Newton/Compression/1e6    18.0ms ± 0%    17.3ms ± 0%   -3.88%  (p=1.000 n=1+1)

Change-Id: I840320fab1c4bf64c76b001c2651ab79f23df4eb
Reviewed-on: https://go-review.googlesource.com/c/go/+/259444
Run-TryBot: Lynn Boger <laboger@linux.vnet.ibm.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Paul Murphy <murp@ibm.com>
Reviewed-by: Carlos Eduardo Seo <carlos.seo@gmail.com>
Trust: Lynn Boger <laboger@linux.vnet.ibm.com>

4 years agocmd/go/internal/get: improve -insecure deprecation docs
witchard [Fri, 25 Sep 2020 14:09:42 +0000 (14:09 +0000)]
cmd/go/internal/get: improve -insecure deprecation docs

Updates #37519

Change-Id: I212607f1839b729d7da24b1258e56997b13ad830
GitHub-Last-Rev: db6d3c835bdf867a0b18f115276210e3a05902ed
GitHub-Pull-Request: golang/go#41613
Reviewed-on: https://go-review.googlesource.com/c/go/+/257157
Run-TryBot: Bryan C. Mills <bcmills@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Trust: Jay Conrod <jayconrod@google.com>
Trust: Bryan C. Mills <bcmills@google.com>
Reviewed-by: Bryan C. Mills <bcmills@google.com>
Reviewed-by: Jay Conrod <jayconrod@google.com>
4 years agocmd/go: add basic support for overlays
Michael Matloob [Tue, 11 Aug 2020 16:57:01 +0000 (12:57 -0400)]
cmd/go: add basic support for overlays

This CL adds basic support for listing packages with overlays.
The new cmd/go/internal/fs package adds an abstraction for communicating
with the file system that will open files according to their overlaid paths,
and provides functions to override those in the build context to open
overlaid files. There is also some support for executing builds on packages
with overlays. In cmd/go/internal/work.(*Builder).build, paths are mapped
to their overlaid paths before they are given as arguments to tools.

For #39958

Change-Id: I5ec0eb9ebbca303e2f1e7dbe22ec32613bc1fd17
Reviewed-on: https://go-review.googlesource.com/c/go/+/253747
Trust: Michael Matloob <matloob@golang.org>
Trust: Jay Conrod <jayconrod@google.com>
Run-TryBot: Michael Matloob <matloob@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Jay Conrod <jayconrod@google.com>
Reviewed-by: Bryan C. Mills <bcmills@google.com>
4 years agocmd/compile: correct leaf type when "selecting" singleton register-sized struct
David Chase [Fri, 2 Oct 2020 18:53:48 +0000 (14:53 -0400)]
cmd/compile: correct leaf type when "selecting" singleton register-sized struct

Two part fix:
1) bring the type "correction" forward from a later CL in the expand calls series
2) when a leaf-selwect is rewritten in place, update the type (it might have been
   changed by the type correction in 1).

Fixes #41736.

Change-Id: Id097efd10481bf0ad92aaead81a7207221c144b5
Reviewed-on: https://go-review.googlesource.com/c/go/+/259203
Trust: David Chase <drchase@google.com>
Run-TryBot: David Chase <drchase@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
4 years agocrypto/tls: fix typo in spelling of permanentError
Luca Spiller [Tue, 6 Oct 2020 08:12:45 +0000 (08:12 +0000)]
crypto/tls: fix typo in spelling of permanentError

Change-Id: I819c121ff388460ec348af773ef94b44416a2ea9
GitHub-Last-Rev: 98dd8fb25cecb73e88d107e0a35e3e63a53dfd09
GitHub-Pull-Request: golang/go#41785
Reviewed-on: https://go-review.googlesource.com/c/go/+/259517
Run-TryBot: Emmanuel Odeke <emm.odeke@gmail.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Emmanuel Odeke <emm.odeke@gmail.com>
Reviewed-by: Filippo Valsorda <filippo@golang.org>
Trust: Emmanuel Odeke <emm.odeke@gmail.com>

4 years agodoc: fix typo in contribute.html
Alberto Donizetti [Thu, 1 Oct 2020 12:00:48 +0000 (14:00 +0200)]
doc: fix typo in contribute.html

Change-Id: Ica27c4a9e4c364d94250aebfc4c2b59cff7f4a8f
Reviewed-on: https://go-review.googlesource.com/c/go/+/258679
Trust: Alberto Donizetti <alb.donizetti@gmail.com>
Reviewed-by: Dmitri Shuralyov <dmitshur@golang.org>
4 years agocmd/compile: avoid applying ARM CMP->CMN rewrite in unsigned context
David Chase [Mon, 5 Oct 2020 16:07:00 +0000 (12:07 -0400)]
cmd/compile: avoid applying ARM CMP->CMN rewrite in unsigned context

Fixes #41780.

Change-Id: I1dc7c19a9f057650905da3a96214c2ff4abb51be
Reviewed-on: https://go-review.googlesource.com/c/go/+/259450
Trust: David Chase <drchase@google.com>
Run-TryBot: David Chase <drchase@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
4 years agoreflect: support multiple keys in struct tags
Alexey Vilenskiy [Fri, 14 Aug 2020 08:37:31 +0000 (11:37 +0300)]
reflect: support multiple keys in struct tags

Fixes #40281

Change-Id: Ie624bce3a78a06d7ed71bba1f501e66802dffd13
Reviewed-on: https://go-review.googlesource.com/c/go/+/248341
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Trust: Dmitri Shuralyov <dmitshur@golang.org>

4 years agosrc/buildall.bash: remove linux-386-387 target
Dmitri Shuralyov [Mon, 28 Sep 2020 18:17:37 +0000 (14:17 -0400)]
src/buildall.bash: remove linux-386-387 target

Support for GO386=387 is being dropped in Go 1.16. There
is no need for the target to be available for testing on
the master branch (where Go 1.16 development is ongoing).

For #40255.

Change-Id: I4a4ee80b0c0a535b6b0b246fe991f26964eb07ca
Reviewed-on: https://go-review.googlesource.com/c/go/+/257963
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Trust: Dmitri Shuralyov <dmitshur@golang.org>

4 years agosrc/buildall.bash: remove mobile filter
Dmitri Shuralyov [Fri, 25 Sep 2020 21:10:24 +0000 (17:10 -0400)]
src/buildall.bash: remove mobile filter

Mobile targets are not supported by misc-compile trybots, as tracked in
golang.org/issue/25963, and need to be filtered out. The buildall.bash
script was created in CL 9438, back when it was a single all-compile
builder, and it was easier to filter out mobile targets in the script
than to come up with a pattern that matches all non-mobile targets.

As of CL 254740, all mobile targets (Android and iOS) have unique GOOS
values. That makes it it easy to filter them out in x/build/dashboard.
This was done in CL 258057. As a result, it's now viable to simplify
this script and perform all misc-compile target selection in x/build,
rather than having it spread it across two places.

Also, as of CL 10750, the all-compile builder has turned into multiple
misc-compile builders, so update the script description accordingly.

Updates #41610.

Change-Id: I1e33260ac18cf0a70bb68cd8e3db5587100c7e87
Reviewed-on: https://go-review.googlesource.com/c/go/+/257962
Run-TryBot: Dmitri Shuralyov <dmitshur@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Alexander Rakoczy <alex@golang.org>
Trust: Dmitri Shuralyov <dmitshur@golang.org>

4 years agocmd/compile: export notinheap annotation to object file
Keith Randall [Fri, 2 Oct 2020 22:48:50 +0000 (15:48 -0700)]
cmd/compile: export notinheap annotation to object file

In the rare case when a cgo type makes it into an object file, we need
the go:notinheap annotation to go with it.

Fixes #41761

Change-Id: I541500cb1a03de954881aef659f96fc0b7738848
Reviewed-on: https://go-review.googlesource.com/c/go/+/259297
Trust: Keith Randall <khr@golang.org>
Run-TryBot: Keith Randall <khr@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
4 years agocmd/go: use cmd/internal/pkgpath for gccgo pkgpath symbol
Ian Lance Taylor [Fri, 2 Oct 2020 23:17:30 +0000 (16:17 -0700)]
cmd/go: use cmd/internal/pkgpath for gccgo pkgpath symbol

Fixes #37272

Change-Id: I6554fd5e5400acb20c5a7e96b1d6cb1a1afb9871
Reviewed-on: https://go-review.googlesource.com/c/go/+/259299
Trust: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Than McIntosh <thanm@google.com>
4 years agocmd/cgo: split gofrontend mangling checks into cmd/internal/pkgpath
Ian Lance Taylor [Fri, 2 Oct 2020 23:03:37 +0000 (16:03 -0700)]
cmd/cgo: split gofrontend mangling checks into cmd/internal/pkgpath

This is a step toward porting https://golang.org/cl/219817 from the
gofrontend repo to the main repo.

Note that this also corrects the implementation of the v2 mangling
scheme to use ..u and ..U where appropriate.

For #37272

Change-Id: I64a1e7ca1c84348efcbf1cf62049eeb05c830ed8
Reviewed-on: https://go-review.googlesource.com/c/go/+/259298
Trust: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Than McIntosh <thanm@google.com>
4 years agocmd/go: update go_windows_test to use test go binary
Michael Matloob [Thu, 24 Sep 2020 15:26:23 +0000 (11:26 -0400)]
cmd/go: update go_windows_test to use test go binary

Most of the cmd/go tests build the cmd/go binary and run that binary to
test it, but TestAbsolutePath used the GOROOT's cmd/go instead, which
makes debugging confusing and means that make.bash has to be run in each
iteration cycle. Update TestAbsolutePath to use the same go binary as
the rest of the cmd/go tests.

Change-Id: Ib4e8ae707b66f1f75ceb346b98358f5604fd28c1
Reviewed-on: https://go-review.googlesource.com/c/go/+/256979
Trust: Michael Matloob <matloob@golang.org>
Run-TryBot: Michael Matloob <matloob@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Jay Conrod <jayconrod@google.com>
Reviewed-by: Bryan C. Mills <bcmills@google.com>
4 years agocmd/dist: test c-archive mode on ios/arm64
Cherry Zhang [Sun, 4 Oct 2020 04:25:17 +0000 (00:25 -0400)]
cmd/dist: test c-archive mode on ios/arm64

It is tested on darwin/arm64. Don't lose it when using GOOS=ios.

Updates #38485.

Change-Id: I7157d6b6f2850f2fd361e35ae310dd1ba9f31aa4
Reviewed-on: https://go-review.googlesource.com/c/go/+/259439
Trust: Cherry Zhang <cherryyz@google.com>
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
4 years agoruntime: clean up runtime.call* frame sizes on ARM64
Austin Clements [Mon, 5 Oct 2020 16:17:30 +0000 (12:17 -0400)]
runtime: clean up runtime.call* frame sizes on ARM64

ARM64 used to require that all assembly frame sizes were of the form
16*N+8 because ARM64 requires 16-byte SP alignment and the assembler
added an 8 byte LR slot. This made all of the runtime.call* frame
sizes wonky. The assembler now rounds up the frame size appropriately
after adding any additional slots it needs, so this is no longer
necessary.

This CL cleans up the frame sizes of these functions so they look the
way you'd expect and match all other architectures.

Change-Id: I47819092296b8983c43eadf2e66c7c1e0d518555
Reviewed-on: https://go-review.googlesource.com/c/go/+/259448
Trust: Austin Clements <austin@google.com>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
4 years agoruntime: define and use call16 everywhere
Austin Clements [Sun, 4 Oct 2020 00:40:49 +0000 (20:40 -0400)]
runtime: define and use call16 everywhere

Currently, runtime.call16 is defined and used only on 32-bit
architectures, while 64-bit architectures all start at call32 and go
up from there. This led to unnecessary complexity because call16's
prototype needed to be in a different file, separate from all of the
other call* prototypes, which in turn led to it getting out of sync
with the other call* prototypes. This CL adds call16 on 64-bit
architectures, bringing them all into sync, and moves the call16
prototype to live with the others.

Prior to CL 31655 (in 2016), call16 couldn't be implemented on 64-bit
architectures because it needed at least four words of argument space
to invoke "callwritebarrier" after copying back the results. CL 31655
changed the way call* invoked the write barrier in preparation for the
hybrid barrier; since the hybrid barrier had to be invoked prior to
copying back results, it needed a different solution that didn't reuse
call*'s stack space. At this point, call16 was no longer a problem on
64-bit, but we never added it. Until now.

Change-Id: Id10ade0e4f75c6ea76afa6229ddaee2b994c27dd
Reviewed-on: https://go-review.googlesource.com/c/go/+/259339
Trust: Austin Clements <austin@google.com>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
4 years agoruntime: correct signature of call16
Austin Clements [Sat, 3 Oct 2020 20:44:22 +0000 (16:44 -0400)]
runtime: correct signature of call16

The signature of call16 is currently missing the "typ" parameter. This
CL fixes this. This wasn't caught by vet because call16 is defined by
macro expansion (see #17544), and we didn't notice the mismatch with
the other call* functions because call16 is defined only on 32-bit
architectures and lives alone in stubs32.go.

Unfortunately, this means its GC signature is also wrong: the "arg"
parameter is treated as a scalar rather than a pointer, so GC won't
trace it and stack copying won't adjust it. This turns out to matter
in exactly one case right now: on 32-bit architectures (which are the
only architectures where call16 is defined), a stack-allocated defer
of a function with a 16-byte or smaller argument frame including a
non-empty result area can corrupt memory if the deferred function
grows the stack and is invoked during a panic. Whew. All other current
uses of reflectcall pass a heap-allocated "arg" frame (which happens
to be reachable from other stack roots, so tracing isn't a problem).

Curiously, in 2016, the signatures of all call* functions were wrong
in exactly this way. CL 31654 fixed all of them in stubs.go, but
missed the one in stubs32.go.

Fixes #41795.

Change-Id: I31e3c0df201f79ee5707eeb8dc4ff0d13fc10ada
Reviewed-on: https://go-review.googlesource.com/c/go/+/259338
Trust: Austin Clements <austin@google.com>
Run-TryBot: Austin Clements <austin@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
4 years agoencoding/asn1: clarify use of SET suffix
Roland Bracewell Shoemaker [Mon, 5 Oct 2020 15:46:23 +0000 (15:46 +0000)]
encoding/asn1: clarify use of SET suffix

This change clarifies the usage of the SET type name suffix. Previously
the documentation was somewhat confusing about where the suffix should
be used, and when used what it applied to. For instance the previous
language could be interpreted such that []exampleSET would be parsed as
a SEQUENCE OF SET, which is incorrect as the SET suffix only applies to
slice types, such as type exampleSET []struct{} which is parsed as a
SET OF SEQUENCE.

Change-Id: I74201d9969f931f69391c236559f66cb460569ec
GitHub-Last-Rev: d0d2ddc587df4564a265c800efb9d8e204002624
GitHub-Pull-Request: golang/go#38543
Reviewed-on: https://go-review.googlesource.com/c/go/+/229078
Trust: Roland Shoemaker <roland@golang.org>
Reviewed-by: Filippo Valsorda <filippo@golang.org>
4 years agomisc/ios: fixup review comments from CL 255257
Elias Naur [Mon, 5 Oct 2020 15:51:54 +0000 (17:51 +0200)]
misc/ios: fixup review comments from CL 255257

Change-Id: I247fc9e0e26e706e6af07367f953eaa1b7e544c1
Reviewed-on: https://go-review.googlesource.com/c/go/+/259577
Trust: Elias Naur <mail@eliasnaur.com>
Run-TryBot: Elias Naur <mail@eliasnaur.com>
Reviewed-by: Dmitri Shuralyov <dmitshur@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>

4 years agocmd/dist: detect gohostarch on ios/arm64
Cherry Zhang [Sat, 3 Oct 2020 01:57:24 +0000 (21:57 -0400)]
cmd/dist: detect gohostarch on ios/arm64

Add a case for gohostos == "ios" along with "darwin".

Updates #38485.

Change-Id: Ic7310e6c97d405f78a5e5db1a639860455e61327
Reviewed-on: https://go-review.googlesource.com/c/go/+/259337
Trust: Cherry Zhang <cherryyz@google.com>
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Dmitri Shuralyov <dmitshur@golang.org>
4 years agoiostest.bash: remove
Elias Naur [Sat, 3 Oct 2020 09:27:04 +0000 (11:27 +0200)]
iostest.bash: remove

There are no tethered iOS builders left, and should they appear in
the future, they should use all.bash.

Change-Id: I3217789514ffa725e4d2584e4991d899c5fda995
Reviewed-on: https://go-review.googlesource.com/c/go/+/259278
Trust: Elias Naur <mail@eliasnaur.com>
Reviewed-by: Dmitri Shuralyov <dmitshur@golang.org>
4 years agomisc/ios: add support for running programs on the iOS simulator
Elias Naur [Wed, 16 Sep 2020 13:23:58 +0000 (15:23 +0200)]
misc/ios: add support for running programs on the iOS simulator

Update the README to mention the emulator. Remove reference to gomobile
while here; there are multiple ways to develop for iOS today, including
using the c-archive buildmode directly.

Updates #38485

Change-Id: Iccef75e646ea8e1b9bc3fc37419cc2d6bf3dfdf4
Reviewed-on: https://go-review.googlesource.com/c/go/+/255257
Run-TryBot: Elias Naur <mail@eliasnaur.com>
TryBot-Result: Go Bot <gobot@golang.org>
Trust: Elias Naur <mail@eliasnaur.com>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
4 years agocmd/link: add support for openbsd/mips64
Joel Sing [Tue, 25 Aug 2020 10:19:55 +0000 (20:19 +1000)]
cmd/link: add support for openbsd/mips64

Update #40995

Change-Id: I2cf9b85a960f479eaa59bf58081d03a0467bc2b8
Reviewed-on: https://go-review.googlesource.com/c/go/+/250582
Trust: Joel Sing <joel@sing.id.au>
Run-TryBot: Joel Sing <joel@sing.id.au>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Go Bot <gobot@golang.org>

4 years agocmd/compile: change mustHeapAlloc to return a reason why
Alberto Donizetti [Thu, 1 Oct 2020 10:03:27 +0000 (12:03 +0200)]
cmd/compile: change mustHeapAlloc to return a reason why

This change renames mustHeapAlloc to heapAllocReason, and changes it
to return the reason why the argument must escape, so we don't have to
re-deduce it in its callers just to print the escape reason. It also
embeds isSmallMakeSlice body in heapAllocReason, since the former was
only used by the latter, and deletes isSmallMakeSlice.

An outdated TODO to remove smallintconst, which the TODO claimed was
only used in one place, was also removed, since grepping shows we
currently call smallintconst in 11 different places.

Change-Id: I0bd11bf29b92c4126f5bb455877ff73217d5a155
Reviewed-on: https://go-review.googlesource.com/c/go/+/258678
Run-TryBot: Alberto Donizetti <alb.donizetti@gmail.com>
TryBot-Result: Go Bot <gobot@golang.org>
Trust: Alberto Donizetti <alb.donizetti@gmail.com>
Trust: Cuong Manh Le <cuong.manhle.vn@gmail.com>
Reviewed-by: Cuong Manh Le <cuong.manhle.vn@gmail.com>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
4 years agoruntime: update and tidy cgo callback description
Austin Clements [Thu, 1 Oct 2020 14:58:47 +0000 (10:58 -0400)]
runtime: update and tidy cgo callback description

The documentation on how cgo callbacks (C -> Go calls) works
internally has gotten somewhat stale. This CL refreshes it.

Change-Id: I1ab66225c9da52d698d97ebeb4f3c7b9b5ee97db
Reviewed-on: https://go-review.googlesource.com/c/go/+/258937
Trust: Austin Clements <austin@google.com>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
4 years agodoc/go1.16: announce netbsd/arm64 support
Tobias Klauser [Fri, 2 Oct 2020 20:13:14 +0000 (22:13 +0200)]
doc/go1.16: announce netbsd/arm64 support

netbsd/arm64 now complies with all the requirements for a port as
specified on https://golang.org/wiki/PortingPolicy

Note that this was preliminarily announced in the Go 1.13 release notes
(CL 183637) but then removed again due to the port lacking a builder at
that time (CL 192997).

Updates #30824

Change-Id: I2f40fabc84fe9cb699282e6a9d13ed9b64478e36
Reviewed-on: https://go-review.googlesource.com/c/go/+/259277
Trust: Tobias Klauser <tobias.klauser@gmail.com>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
4 years agoruntime: add debugging to TestTimePprof
Austin Clements [Thu, 1 Oct 2020 20:06:03 +0000 (16:06 -0400)]
runtime: add debugging to TestTimePprof

We've seen timeouts of TestTimePprof, but the tracebacks aren't useful
because goroutines are running on other threads. Add GOTRACEBACK=crash
to catch these in the future.

For #41120.

Change-Id: I97318172ef78d0cbab10df5e4ffcbfeadff579e3
Reviewed-on: https://go-review.googlesource.com/c/go/+/258802
Trust: Austin Clements <austin@google.com>
Run-TryBot: Austin Clements <austin@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Bryan C. Mills <bcmills@google.com>
4 years agogo/types: add Checker.walkDecl to simplify checking declarations
Rob Findley [Mon, 21 Sep 2020 12:47:34 +0000 (08:47 -0400)]
go/types: add Checker.walkDecl to simplify checking declarations

Handling ast.GenDecls while typechecking is repetitive, and a source of
diffs compared to typechecking using the cmd/compile/internal/syntax
package, which unpacks declaration groups into individual declarations.

Refactor to extract the logic for walking declarations.  This introduces
a new AST abstraction: types.decl, which comes at some minor performance
cost. However, if we are to fully abstract the AST we will be paying
this cost anyway, and benchmarking suggests that the cost is negligible.

Change-Id: If73c30c3d08053ccf7bf21ef886f0452fdbf142e
Reviewed-on: https://go-review.googlesource.com/c/go/+/256298
Run-TryBot: Robert Findley <rfindley@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Robert Griesemer <gri@golang.org>
Trust: Robert Findley <rfindley@google.com>

4 years agonet/http: remove duplicate declaration of error
Dai Jie [Fri, 2 Oct 2020 09:09:24 +0000 (09:09 +0000)]
net/http: remove duplicate declaration of error

there is no need to declare a error variable here.

Change-Id: I9ea5bcf568d800efed19c90caf751aaf9abe5555
GitHub-Last-Rev: 538d1f9cee0b8564a8bec262529f567da847f1b0
GitHub-Pull-Request: golang/go#41751
Reviewed-on: https://go-review.googlesource.com/c/go/+/259037
Reviewed-by: Rob Pike <r@golang.org>
Trust: Alberto Donizetti <alb.donizetti@gmail.com>

4 years agocrypto/dsa,crypto/x509: deprecate DSA and remove crypto/x509 support
Filippo Valsorda [Mon, 28 Sep 2020 17:29:57 +0000 (19:29 +0200)]
crypto/dsa,crypto/x509: deprecate DSA and remove crypto/x509 support

Updates #40337

Change-Id: I5c1218df3ae7e13144a1d9f7d4a4b456e4475c0a
Reviewed-on: https://go-review.googlesource.com/c/go/+/257939
Trust: Filippo Valsorda <filippo@golang.org>
Trust: Roland Shoemaker <roland@golang.org>
Reviewed-by: Roland Shoemaker <roland@golang.org>
Reviewed-by: Katie Hockman <katie@golang.org>
4 years agocmd/go: remove test checking GO386=387 invalidates cache
Alberto Donizetti [Fri, 2 Oct 2020 07:58:45 +0000 (09:58 +0200)]
cmd/go: remove test checking GO386=387 invalidates cache

Since GO386=387 is no longer supported, this change deletes a cmd/go
test checking that building something with GO386=387, and then with
sse2, invalidated the build cache.

Fixes the longtest builders.

Change-Id: I5f645ef4ddf1cddb26dcf9390cee94907fc45a70
Reviewed-on: https://go-review.googlesource.com/c/go/+/259017
Trust: Alberto Donizetti <alb.donizetti@gmail.com>
Trust: Tobias Klauser <tobias.klauser@gmail.com>
Reviewed-by: Tobias Klauser <tobias.klauser@gmail.com>
4 years agoall: drop 387 support
Keith Randall [Thu, 1 Oct 2020 21:49:33 +0000 (14:49 -0700)]
all: drop 387 support

My last 387 CL. So sad ... ... ... ... not!

Fixes #40255

Change-Id: I8d4ddb744b234b8adc735db2f7c3c7b6d8bbdfa4
Reviewed-on: https://go-review.googlesource.com/c/go/+/258957
Trust: Keith Randall <khr@golang.org>
Run-TryBot: Keith Randall <khr@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
4 years agoreflect: add test for variadic reflect.Type.Method
Ian Lance Taylor [Thu, 1 Oct 2020 21:55:11 +0000 (14:55 -0700)]
reflect: add test for variadic reflect.Type.Method

For #41737

Change-Id: Id065880dd7da54dec1b45662c202aeb7f8397c60
Reviewed-on: https://go-review.googlesource.com/c/go/+/258819
Trust: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
4 years agocmd/go/internal/modfetch: always extract module directories in place
Jay Conrod [Thu, 1 Oct 2020 17:37:06 +0000 (13:37 -0400)]
cmd/go/internal/modfetch: always extract module directories in place

Previously by default, we extracted modules to a temporary directory,
then renamed it into place. This failed with ERROR_ACCESS_DENIED on
Windows if another process (usually an anti-virus scanner) opened
files in the temporary directory.

Since Go 1.15, users have been able to set
GODEBUG=modcacheunzipinplace=1 to opt into new behavior: we extract
modules at their final location, and we create and later delete a
.partial file to prevent the directory from being used if we crash.
.partial files are recognized by Go 1.14.2 and later.

With this change, the new behavior is the only behavior.
modcacheunzipinplace is no longer recognized.

Fixes #36568

Change-Id: Iff19fca5cd6eaa3597975a69fa05c4cb1b834bd6
Reviewed-on: https://go-review.googlesource.com/c/go/+/258798
Run-TryBot: Jay Conrod <jayconrod@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Trust: Jay Conrod <jayconrod@google.com>
Reviewed-by: Bryan C. Mills <bcmills@google.com>
4 years agocmd/compile: allow directory specification for GOSSAFUNC output
David Chase [Mon, 31 Aug 2020 18:29:58 +0000 (14:29 -0400)]
cmd/compile: allow directory specification for GOSSAFUNC output

This was useful for debugging failures occurring during make.bash.
The added flush also ensures that any hints in the GOSSAFUNC output
are flushed before fatal exit.

The environment variable GOSSADIR specifies where the SSA html debugging
files should be placed.  To avoid collisions, each one is written into
the [package].[functionOrMethod].html, where [package] is the filepath
separator separated package name, function is the function name, and method
is either (*Type).Method, or Type.Method, as appropriate.  Directories
are created as necessary to make this work.

Change-Id: I420927426b618b633bb1ffc51cf0f223b8f6d49c
Reviewed-on: https://go-review.googlesource.com/c/go/+/252338
Trust: David Chase <drchase@google.com>
Run-TryBot: David Chase <drchase@google.com>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
4 years agocmd/compile: activate revert switch for late call expansion regression
David Chase [Thu, 1 Oct 2020 20:22:19 +0000 (16:22 -0400)]
cmd/compile: activate revert switch for late call expansion regression

Not a fix, but things will work while I fix it.
Credit @andybons "for we revert switches for scary stuff".

Updates #41736

Change-Id: I55f90860eae919765aac4f6d9f108a54139027e1
Reviewed-on: https://go-review.googlesource.com/c/go/+/258897
Trust: David Chase <drchase@google.com>
Run-TryBot: David Chase <drchase@google.com>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Go Bot <gobot@golang.org>

4 years agoruntime: align 12-byte objects to 8 bytes on 32-bit systems
Michael Anthony Knyszek [Thu, 10 Sep 2020 21:20:46 +0000 (21:20 +0000)]
runtime: align 12-byte objects to 8 bytes on 32-bit systems

Currently on 32-bit systems 8-byte fields in a struct have an alignment
of 4 bytes, which means that atomic instructions may fault. This issue
is tracked in #36606.

Our current workaround is to allocate memory and put any such atomically
accessed fields at the beginning of the object. This workaround fails
because the tiny allocator might not align the object right. This case
specifically only happens with 12-byte objects because a type's size is
rounded up to its alignment. So if e.g. we have a type like:

type obj struct {
    a uint64
    b byte
}

then its size will be 12 bytes, because "a" will require a 4 byte
alignment. This argument may be extended to all objects of size 9-15
bytes.

So, make this workaround work by specifically aligning such objects to 8
bytes on 32-bit systems. This change leaves a TODO to remove the code
once #36606 gets resolved. It also adds a test which will presumably no
longer be necessary (the compiler should enforce the right alignment)
when it gets resolved as well.

Fixes #37262.

Change-Id: I3a34e5b014b3c37ed2e5e75e62d71d8640aa42bc
Reviewed-on: https://go-review.googlesource.com/c/go/+/254057
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Reviewed-by: Austin Clements <austin@google.com>
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Trust: Michael Knyszek <mknyszek@google.com>

4 years agocmd/compile,cmd/internal/obj/ppc64: fix some shift rules due to a regression
Lynn Boger [Mon, 28 Sep 2020 22:20:12 +0000 (18:20 -0400)]
cmd/compile,cmd/internal/obj/ppc64: fix some shift rules due to a regression

A recent change to improve shifts was generating some
invalid cases when the rule was based on an AND. The
extended mnemonics CLRLSLDI and CLRLSLWI only allow
certain values for the operands and in the mask case
those values were not being checked properly. This
adds a check to those rules to verify that the
'b' and 'n' values used when an AND was part of the rule
have correct values.

There was a bug in some diag messages in asm9. The
message expected 3 values but only provided 2. Those are
corrected here also.

The test/codegen/shift.go was updated to add a few more
cases to check for the case mentioned here.

Some of the comments that mention the order of operands
in these extended mnemonics were wrong and those have been
corrected.

Fixes #41683.

Change-Id: If5bb860acaa5051b9e0cd80784b2868b85898c31
Reviewed-on: https://go-review.googlesource.com/c/go/+/258138
Run-TryBot: Lynn Boger <laboger@linux.vnet.ibm.com>
Reviewed-by: Paul Murphy <murp@ibm.com>
Reviewed-by: Carlos Eduardo Seo <carlos.seo@gmail.com>
TryBot-Result: Go Bot <gobot@golang.org>
Trust: Lynn Boger <laboger@linux.vnet.ibm.com>

4 years agocmd/compile: late call expansion for go func and simple defer func
David Chase [Tue, 11 Aug 2020 12:39:17 +0000 (08:39 -0400)]
cmd/compile: late call expansion for go func and simple defer func

Passes run.bash and race.bash on darwin/amd64.

Change-Id: Icbccaa2f2e7c3eac7c328c5253f331e598e11542
Reviewed-on: https://go-review.googlesource.com/c/go/+/247898
Trust: David Chase <drchase@google.com>
Run-TryBot: David Chase <drchase@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
4 years agocmd/compile: enable late expansion for closure calls
David Chase [Mon, 10 Aug 2020 22:30:11 +0000 (18:30 -0400)]
cmd/compile: enable late expansion for closure calls

This works for "normal" calls.
Defer func() and Go func() still pending.
RT calls still pending.

Change-Id: I29cbdad8c877d12c08bbf7f3f0696611de877da9
Reviewed-on: https://go-review.googlesource.com/c/go/+/247771
Trust: David Chase <drchase@google.com>
Run-TryBot: David Chase <drchase@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
4 years agocmd/compile: enable late expansion for interface calls
David Chase [Sat, 8 Aug 2020 02:46:43 +0000 (22:46 -0400)]
cmd/compile: enable late expansion for interface calls

Includes a few tweaks to Value.copyOf(a) (make it a no-op for
a self-copy) and new pattern hack "___" (3 underscores) is
like ellipsis, except the replacement doesn't need to have
matching ellipsis/underscores.

Moved the arg-length check in generated pattern-matching code
BEFORE the args are probed, because not all instances of
variable length OpFoo will have all the args mentioned in
some rule for OpFoo, and when that happens, the compiler
panics without the early check.

Change-Id: I66de40672b3794a6427890ff96c805a488d783f4
Reviewed-on: https://go-review.googlesource.com/c/go/+/247537
Trust: David Chase <drchase@google.com>
Run-TryBot: David Chase <drchase@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
4 years agocmd/compile: enable late expansion for address-of static calls
David Chase [Wed, 5 Aug 2020 15:42:44 +0000 (11:42 -0400)]
cmd/compile: enable late expansion for address-of static calls

passes run.bash and race.bash (on Darwin-amd64)

Change-Id: I2abda9636b681d050e85e88fc357ebe5220d2ba2
Reviewed-on: https://go-review.googlesource.com/c/go/+/246938
Trust: David Chase <drchase@google.com>
Run-TryBot: David Chase <drchase@google.com>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
4 years agocmd/compile: fix late call expansion for SSA-able aggregate results and arguments
David Chase [Mon, 27 Jul 2020 20:46:35 +0000 (16:46 -0400)]
cmd/compile: fix late call expansion for SSA-able aggregate results and arguments

This change incorporates the decision that it should be possible to
run call expansion relatively late in the optimization chain, so that
(1) calls themselves can be exposed to useful optimizations
(2) the effect of selectors on aggregates is seen at the rewrite,
    so that assignment of parts into registers is less complicated
    (at least I hope it works that way).

That means that selectors feeding into SelectN need to be processed,
and Make* feeding into call parameters need to be processed.

This does however require that call expansion run before decompose
builtins.

This doesn't yet handle rewrites of strings, slices, interfaces,
and complex numbers.

Passes run.bash and race.bash

Change-Id: I71ff23d3c491043beb30e926949970c4f63ef1a4
Reviewed-on: https://go-review.googlesource.com/c/go/+/245133
Trust: David Chase <drchase@google.com>
Run-TryBot: David Chase <drchase@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
4 years agoMerge "cmd/link: merge branch 'dev.link' into master"
Gerrit Code Review [Thu, 1 Oct 2020 15:16:43 +0000 (15:16 +0000)]
Merge "cmd/link: merge branch 'dev.link' into master"

4 years agodoc/go1.16: fix crypto typo
Roland Shoemaker [Thu, 1 Oct 2020 14:57:00 +0000 (07:57 -0700)]
doc/go1.16: fix crypto typo

Change-Id: Icf962098cc22f16b0acf75db1e82eaddb9fa0c80
Reviewed-on: https://go-review.googlesource.com/c/go/+/258777
Trust: Roland Shoemaker <roland@golang.org>
Run-TryBot: Roland Shoemaker <roland@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
4 years agocmd/link: enable ASLR on windows binaries built with -buildmode=c-shared
Quim [Wed, 16 Sep 2020 23:59:14 +0000 (01:59 +0200)]
cmd/link: enable ASLR on windows binaries built with -buildmode=c-shared

Windows binaries built with -buildmode=c-shared set will have
IMAGE_DLLCHARACTERISTICS_DYNAMIC_BASE flag set, and
IMAGE_DLLCHARACTERISTICS_HIGH_ENTROPY_VA flag set for windows/amd64.

ASLR can be disabled on windows by using the new linker -aslr flag.

RELNOTE=yes

Fixes #41421

Change-Id: I62bd88c6d7e0f87173b093a0ad8e1a4d269ec790
Reviewed-on: https://go-review.googlesource.com/c/go/+/255259
Reviewed-by: Alex Brainman <alex.brainman@gmail.com>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Trust: Alex Brainman <alex.brainman@gmail.com>
Trust: Cherry Zhang <cherryyz@google.com>
Run-TryBot: Alex Brainman <alex.brainman@gmail.com>
TryBot-Result: Go Bot <gobot@golang.org>

4 years agocmd/link: merge branch 'dev.link' into master
Jeremy Faller [Thu, 1 Oct 2020 14:24:31 +0000 (10:24 -0400)]
cmd/link: merge branch 'dev.link' into master

no conflicts.

Change-Id: Ic40ad81f1e491d6ceb72fbb66ad7919ad2b34a3e

4 years agogo/types: turn TestBenchmark into a normal benchmark
Rob Findley [Mon, 28 Sep 2020 15:13:14 +0000 (11:13 -0400)]
go/types: turn TestBenchmark into a normal benchmark

TestBenchmark doesn't use the -bench flag, so that it can format custom
output -- the number of checked lines per second. This is a barrier both
to discoverability, and to piping benchmark output into analysis tools
such as benchstat.

Using testing.B.ReportMetric and a bit of manual timing, we can achieve
similar results while conforming to normal benchmark output. Do this,
and rename the test func to BenchmarkCheck (for symmetry with
TestCheck).

Change-Id: Ie8f2259c1ca9e6986f0137287acf8eb2843f96b8
Reviewed-on: https://go-review.googlesource.com/c/go/+/257958
Run-TryBot: Robert Findley <rfindley@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Trust: Robert Findley <rfindley@google.com>
Reviewed-by: Robert Griesemer <gri@golang.org>
4 years agosyscall: use correct cmsg alignment for netbsd/arm64
Tobias Klauser [Thu, 1 Oct 2020 08:25:20 +0000 (10:25 +0200)]
syscall: use correct cmsg alignment for netbsd/arm64

netbsd/arm64 requires 128-bit alignment for cmsgs.

Re-submit of CL 258437 which was dropped due to #41718.

Change-Id: I898043d79f513bebe1a5eb931e7ebd8e291a5aec
Reviewed-on: https://go-review.googlesource.com/c/go/+/258677
Trust: Tobias Klauser <tobias.klauser@gmail.com>
Trust: Benny Siegert <bsiegert@gmail.com>
Run-TryBot: Tobias Klauser <tobias.klauser@gmail.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Benny Siegert <bsiegert@gmail.com>
4 years agoio: make clear that EOF should not be wrapped
Russ Cox [Wed, 30 Sep 2020 19:48:14 +0000 (15:48 -0400)]
io: make clear that EOF should not be wrapped

For #40827.

Change-Id: Ifd108421abd8d0988dd7b985e4f9e2bd5356964a
Reviewed-on: https://go-review.googlesource.com/c/go/+/258524
Trust: Russ Cox <rsc@golang.org>
Reviewed-by: Rob Pike <r@golang.org>
4 years agointernal/poll: use ignoringEINTR in Darwin Fsync
Ian Lance Taylor [Thu, 1 Oct 2020 00:37:24 +0000 (17:37 -0700)]
internal/poll: use ignoringEINTR in Darwin Fsync

Also add comment explaining why we don't use ignoringEINTR around
call to close.

Fixes #41115

Change-Id: Ia7bbe01eaf26003f70d184b7e82803efef2b2c18
Reviewed-on: https://go-review.googlesource.com/c/go/+/258542
Trust: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Bryan C. Mills <bcmills@google.com>
4 years ago[dev.link] Merge branch 'master' into dev.link
Jeremy Faller [Wed, 30 Sep 2020 21:57:14 +0000 (17:57 -0400)]
[dev.link] Merge branch 'master' into dev.link

2 conflicts, that make sense.
src/cmd/internal/obj/objfile.go
src/cmd/link/internal/loader/loader.go

Change-Id: Ib224e2d248cb568fa1e888af79dd908b2f5e05ff

4 years ago[dev.link] cmd/link: use generator symbols for the rest of pclntab
Jeremy Faller [Tue, 18 Aug 2020 17:38:04 +0000 (13:38 -0400)]
[dev.link] cmd/link: use generator symbols for the rest of pclntab

Move the rest of pclntab creation to generator symbols. Any savings in
pclntab generation CPU time is eaten by the generators run in Asmb
phase.

Stats for Darwin, cmd/compile:

alloc/op:
Pclntab_GC                   13.9MB ± 0%     6.4MB ± 0%    -53.68%  (p=0.000 n=10+10)

allocs/op
Pclntab_GC                    86.5k ± 0%     61.5k ± 0%    -28.90%  (p=0.000 n=10+10)

liveB:
Pclntab_GC                    24.3M ± 0%     22.9M ± 0%     -5.57%  (p=0.000 n=10+10)

Timing:

Pclntab                   32.1ms ± 2%    24.2ms ± 2%    -24.35%  (p=0.000 n=9+9)
Asmb                      18.3ms ±14%    27.4ms ± 9%    +49.55%  (p=0.000 n=10+10)
TotalTime                  351ms ± 2%     347ms ± 3%       ~     (p=0.200 n=9+8)

Change-Id: I5c6b6df5953f6f255240e07578f1c9f8c5f68500
Reviewed-on: https://go-review.googlesource.com/c/go/+/249023
Trust: Jeremy Faller <jeremy@golang.org>
Run-TryBot: Jeremy Faller <jeremy@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Austin Clements <austin@google.com>
4 years agoruntime: code cleanup about map
mengxiaodong [Mon, 28 Sep 2020 09:38:13 +0000 (17:38 +0800)]
runtime: code cleanup about map

1.Revise ambiguous comments: "all current buckets" means buckets in hmap.buckets, actually current bucket and all the overflow buckets connected to it are full
2.All the pointer address add use src/runtime/stubs.go:add, keep the code style uniform

Change-Id: Idc7224dbe6c391e1b03bf5d009c3734bc75187ce
Reviewed-on: https://go-review.googlesource.com/c/go/+/257979
Reviewed-by: Austin Clements <austin@google.com>
Reviewed-by: Keith Randall <khr@golang.org>
Run-TryBot: Austin Clements <austin@google.com>
TryBot-Result: Go Bot <gobot@golang.org>

4 years ago[dev.link] cmd/link run generators in parallel
Jeremy Faller [Mon, 28 Sep 2020 18:11:02 +0000 (14:11 -0400)]
[dev.link] cmd/link run generators in parallel

Small runtime win:

Stats for darwin, building cmd/compile:

Asmb                      20.7ms ±14%    18.3ms ±14%  -11.54%  (p=0.002 n=10+10)
TotalTime                  365ms ±10%     351ms ± 2%     ~     (p=0.211 n=10+9)

Change-Id: Ia8afdf6948111d59b0c89e52cb50557a10f33c40
Reviewed-on: https://go-review.googlesource.com/c/go/+/257964
Trust: Jeremy Faller <jeremy@golang.org>
Run-TryBot: Jeremy Faller <jeremy@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Austin Clements <austin@google.com>
4 years agosrc/go.mod, net/http: update bundled and latest golang.org/x/net
Emmanuel T Odeke [Wed, 30 Sep 2020 07:47:48 +0000 (00:47 -0700)]
src/go.mod, net/http: update bundled and latest golang.org/x/net

Updates x/net/http2 to git rev 5d4f7005572804eaf7f5ecdd2473a62557f733ba

    http2: send WINDOW_UPDATE on a body's write failure
    https://golang.org/cl/245158 (fixes #40423)

also updates the vendored version of golang.org/x/net as per

$ go get golang.org/x/net@5d4f700557
$ go mod tidy
$ go mod vendor
$ go generate -run bundle std

For #40423.

Change-Id: I3270d0fb6f28889266596f7365d36d30ef2bb368
Reviewed-on: https://go-review.googlesource.com/c/go/+/258359
Run-TryBot: Emmanuel Odeke <emm.odeke@gmail.com>
TryBot-Result: Go Bot <gobot@golang.org>
Trust: Emmanuel Odeke <emm.odeke@gmail.com>
Reviewed-by: Dmitri Shuralyov <dmitshur@golang.org>
4 years agocmd/go: test more commands in mod_build_info_error
Bryan C. Mills [Tue, 29 Sep 2020 14:23:07 +0000 (10:23 -0400)]
cmd/go: test more commands in mod_build_info_error

For #26909
For #41688

Change-Id: I22f28d426ce499fce6f0f1295dbde425998042aa
Reviewed-on: https://go-review.googlesource.com/c/go/+/258219
Trust: Bryan C. Mills <bcmills@google.com>
Trust: Jay Conrod <jayconrod@google.com>
Run-TryBot: Bryan C. Mills <bcmills@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Jay Conrod <jayconrod@google.com>
Reviewed-by: Michael Matloob <matloob@golang.org>
4 years agocmd/go: add yet another test case for ambiguous arguments to 'go get'
Bryan C. Mills [Wed, 23 Sep 2020 20:32:59 +0000 (16:32 -0400)]
cmd/go: add yet another test case for ambiguous arguments to 'go get'

For #37438

Change-Id: Ie40971ff677d36ddadbf9834bba2d366a0fc34d0
Reviewed-on: https://go-review.googlesource.com/c/go/+/256922
Trust: Bryan C. Mills <bcmills@google.com>
Trust: Jay Conrod <jayconrod@google.com>
Run-TryBot: Bryan C. Mills <bcmills@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Michael Matloob <matloob@golang.org>
Reviewed-by: Jay Conrod <jayconrod@google.com>
4 years agocrypto/x509: enforce SAN IA5String encoding restrictions
Roland Shoemaker [Sat, 23 May 2020 17:15:46 +0000 (10:15 -0700)]
crypto/x509: enforce SAN IA5String encoding restrictions

Extends the IA5String encoding restrictions that are currently applied
to name constraints to dNSName, rfc822Name, and
uniformResourceIdentifier elements of the SAN. The utility function
isIA5String is updated to use unicode.MaxASCII rather than utf8.RuneSelf
as it is somewhat more readable.

Certificates that include these badly encoded names do exist, but are
exceedingly rare. zlint and other linters enforce this encoding and
searching censys.io reveals only three currently trusted certificates
with this particular encoding issue.

Fixes #26362

Change-Id: I7a4f3e165a1754e5b4bfaeabc03e01eb7367f3c9
Reviewed-on: https://go-review.googlesource.com/c/go/+/235078
Run-TryBot: Roland Shoemaker <roland@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>
Trust: Roland Shoemaker <roland@golang.org>
Reviewed-by: Filippo Valsorda <filippo@golang.org>
4 years agocmd/go: add another test case for package/module ambiguity in 'go get'
Bryan C. Mills [Wed, 23 Sep 2020 16:28:29 +0000 (12:28 -0400)]
cmd/go: add another test case for package/module ambiguity in 'go get'

For #37438

Change-Id: Iae00ef7f97144e85f4f710cdb3087c2548b4b8f0
Reviewed-on: https://go-review.googlesource.com/c/go/+/256799
Trust: Bryan C. Mills <bcmills@google.com>
Trust: Jay Conrod <jayconrod@google.com>
Run-TryBot: Bryan C. Mills <bcmills@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Michael Matloob <matloob@golang.org>
Reviewed-by: Jay Conrod <jayconrod@google.com>
4 years agocrypto/x509: return errors instead of panicking
Paschalis Tsilias [Wed, 2 Sep 2020 10:44:36 +0000 (13:44 +0300)]
crypto/x509: return errors instead of panicking

Eliminate a panic in x509.CreateCertificate when passing templates with unknown ExtKeyUsage; return an error instead.

Fixes #41169

Change-Id: Ia229d3b0d4a1bdeef05928439d97dab228687b3c
Reviewed-on: https://go-review.googlesource.com/c/go/+/252557
Reviewed-by: Roland Shoemaker <roland@golang.org>
Reviewed-by: Filippo Valsorda <filippo@golang.org>
Run-TryBot: Roland Shoemaker <roland@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>

4 years agocrypto/ecdsa: use FillBytes on s390x
Ruixin Bao [Tue, 29 Sep 2020 19:55:19 +0000 (15:55 -0400)]
crypto/ecdsa: use FillBytes on s390x

Originally, zeroExtendAndCopy is used to pad src with leading zeros and
copy the padded src into the destination. It is no longer needed after
CL 230397 introduced FillBytes. We can simply use that and remove the
zeroExtendAndCopy function. It is cleaner and reduces some allocation.

In addition, this patch tries to avoid calling hashToInt function in
both Sign and Verify function so some allocation is reduced.

Benchmarks:
name             old alloc/op   new alloc/op   delta
SignP256-8         1.60kB ± 0%    1.49kB ± 0%    -7.23%  (p=0.000 n=20+20)
SignP384-8         1.74kB ± 0%    1.59kB ± 0%    -8.50%  (p=0.000 n=20+18)
VerifyP256-8         176B ± 0%        0B       -100.00%  (p=0.000 n=20+20)
KeyGeneration-8      640B ± 0%      640B ± 0%      ~     (all equal)

name             old allocs/op  new allocs/op  delta
SignP256-8           22.0 ± 0%      17.0 ± 0%   -22.73%  (p=0.000 n=20+20)
SignP384-8           22.0 ± 0%      17.0 ± 0%   -22.73%  (p=0.000 n=20+20)
VerifyP256-8         7.00 ± 0%      0.00       -100.00%  (p=0.000 n=20+20)
KeyGeneration-8      13.0 ± 0%      13.0 ± 0%      ~     (all equal)

Change-Id: Ic4c95191eded55deb3420d97db501689f3b173c9
Reviewed-on: https://go-review.googlesource.com/c/go/+/232297
Reviewed-by: Michael Munday <mike.munday@ibm.com>
Run-TryBot: Michael Munday <mike.munday@ibm.com>
TryBot-Result: Go Bot <gobot@golang.org>
Trust: Filippo Valsorda <filippo@golang.org>

4 years agocmd/go: error if -modfile used with 'go install pkg@version'
Jay Conrod [Tue, 29 Sep 2020 20:40:57 +0000 (16:40 -0400)]
cmd/go: error if -modfile used with 'go install pkg@version'

'go install pkg@version' runs without a main module or a module root
directory. The -modfile flag cannot be used to set the module root
directory or to substitute a different go.mod file.

This error won't be reported if -modfile is set in
GOFLAGS. Unsupported flags there are generally ignored.

For #40276

Change-Id: I0b39b1fa9184c15c6e863b647d43c328710920f4
Reviewed-on: https://go-review.googlesource.com/c/go/+/258297
Trust: Jay Conrod <jayconrod@google.com>
Run-TryBot: Jay Conrod <jayconrod@google.com>
Reviewed-by: Bryan C. Mills <bcmills@google.com>
TryBot-Result: Go Bot <gobot@golang.org>

4 years agoruntime: don't crash if vsyscall and vdso are disabled on x86_64
Andrei Vagin [Tue, 29 Sep 2020 04:45:55 +0000 (04:45 +0000)]
runtime: don't crash if vsyscall and vdso are disabled on x86_64

If vdso is disabled, the goruntime calls gettimeofday from vsyscall,
but if vsyscall is disabled too, all golang binaries crash:

SIGSEGV {si_signo=SIGSEGV, si_code=SEGV_MAPERR, si_addr=0xffffffffff600000} ---
killed by SIGSEGV (core dumped) ++

vsyscall doesn't work as it was designed for a long time due to security
reasons and now vsyscall is a little more expensive than real syscalls:
https://github.com/torvalds/linux/commit/5cec93c216db

This patch reworks the code to call syscalls if the vdso library isn't
available.

Change-Id: I16cbf3f49871bea91e26af1f49aa0ae2fbd3a01d
GitHub-Last-Rev: 1d133cd30a5dee1fea9aee0fb4ea0b07e0e87f2a
GitHub-Pull-Request: golang/go#41681
Reviewed-on: https://go-review.googlesource.com/c/go/+/257982
Run-TryBot: Tobias Klauser <tobias.klauser@gmail.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Reviewed-by: Michael Pratt <mpratt@google.com>
Trust: Michael Pratt <mpratt@google.com>

4 years agocmd/compile: report type loop for invalid recursive types
Matthew Dempsky [Tue, 29 Sep 2020 09:11:10 +0000 (02:11 -0700)]
cmd/compile: report type loop for invalid recursive types

Similar to how we report initialization loops in initorder.go and type
alias loops in typecheck.go, this CL updates align.go to warn about
invalid recursive types. The code is based on the loop code from
initorder.go, with minimal changes to adapt from detecting
variable/function initialization loops to detecting type declaration
loops.

Thanks to Cuong Manh Le for investigating this, helping come up with
test cases, and exploring solutions.

Fixes #41575
Updates #41669.

Change-Id: Idb2cb8c5e1d645e62900e178fcb50af33e1700a1
Reviewed-on: https://go-review.googlesource.com/c/go/+/258177
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Robert Griesemer <gri@golang.org>
Reviewed-by: Cuong Manh Le <cuong.manhle.vn@gmail.com>
Trust: Matthew Dempsky <mdempsky@google.com>
Trust: Cuong Manh Le <cuong.manhle.vn@gmail.com>

4 years agocrypto/x509: prioritize potential parents in chain building
Roland Shoemaker [Fri, 8 May 2020 22:57:25 +0000 (15:57 -0700)]
crypto/x509: prioritize potential parents in chain building

When building a x509 chain the algorithm currently looks for parents
that have a subject key identifier (SKID) that matches the child
authority key identifier (AKID), if it is present, and returns all
matches. If the child doesn't have an AKID, or there are no parents
with matching SKID it will instead return all parents that have a
subject DN matching the child's issuer DN. Prioritizing AKID/SKID
matches over issuer/subject matches means that later in buildChains we
have to throw away any pairs where these DNs do not match. This also
prevents validation when a child has a SKID with two possible parents,
one with matching AKID but mismatching subject DN, and one with a
matching subject but missing AKID. In this case the former will be
chosen and the latter ignored, meaning a valid chain cannot be built.

This change alters how possible parents are chosen. Instead of doing a
two step search it instead only consults the CertPool.byName subject DN
map, avoiding issues where possible parents may be shadowed by parents
that have SKID but bad subject DNs. Additionally it orders the list of
possible parents by the likelihood that they are in fact a match. This
ordering follows this pattern:
* AKID and SKID match
* AKID present, SKID missing / AKID missing, SKID present
* AKID and SKID don't match

In an ideal world this should save a handful of cycles when there are
multiple possible matching parents by prioritizing parents that have
the highest likelihood. This does diverge from past behavior in that
it also means there are cases where _more_ parents will be considered
than in the past. Another version of this change could just retain the
past behavior, and only consider parents where both the subject and
issuer DNs match, and if both parent and child have SKID and AKID also
compare those, without any prioritization of the candidate parents.

This change removes an existing test case as it assumes that the
CertPool will return a possible candidate where the issuer/subject DNs
do not match.

Fixes #30079

Change-Id: I629f579cabb0b3d0c8cae5ad0429cc5a536b3e58
Reviewed-on: https://go-review.googlesource.com/c/go/+/232993
Trust: Roland Shoemaker <roland@golang.org>
Run-TryBot: Roland Shoemaker <roland@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Filippo Valsorda <filippo@golang.org>
4 years agoruntime: correct arguments to pipe2 syscall in pipe on netbsd/arm64
Tobias Klauser [Tue, 29 Sep 2020 13:13:17 +0000 (15:13 +0200)]
runtime: correct arguments to pipe2 syscall in pipe on netbsd/arm64

Follow the implementation of pipe2 to pass arguments and handle return
values.

With this fix the runtime tests pass on netbsd/arm64 (tested using
gomote on the netbsd-arm64-bsiegert builder).

Update #30824

Change-Id: I346db68fc9dde8dc7f11351af05c1ad3105f4a32
Reviewed-on: https://go-review.googlesource.com/c/go/+/258000
Trust: Tobias Klauser <tobias.klauser@gmail.com>
Trust: Benny Siegert <bsiegert@gmail.com>
Run-TryBot: Tobias Klauser <tobias.klauser@gmail.com>
Reviewed-by: Benny Siegert <bsiegert@gmail.com>
TryBot-Result: Go Bot <gobot@golang.org>

4 years agocmd/compile: mark type namedata symbols content-addressable
Cherry Zhang [Mon, 28 Sep 2020 17:10:30 +0000 (13:10 -0400)]
cmd/compile: mark type namedata symbols content-addressable

Type namedata symbols are for type/field/method names and package
paths. We can use content-addressable symbol mechanism for them.

Change-Id: I923fda17b7094c7a0e46aad7c450622eb3826294
Reviewed-on: https://go-review.googlesource.com/c/go/+/257960
Trust: Cherry Zhang <cherryyz@google.com>
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Than McIntosh <thanm@google.com>
Reviewed-by: Jeremy Faller <jeremy@golang.org>
4 years agocmd/link: retain only used interface methods
Cherry Zhang [Tue, 22 Sep 2020 00:44:53 +0000 (20:44 -0400)]
cmd/link: retain only used interface methods

Currently, in the linker's deadcode pass, when an interface type
is live, the linker thinks all its methods are live, and uses
them to match methods on concrete types. The interface method may
never be used, though.

This CL changes it to only keep used interface methods, for
matching concrete type methods. To do that, when an interface
method is used, the compiler generates a mark relocation. The
linker uses the marker relocations to mark used interface
methods, and only the used ones.

binary size    before      after
cmd/compile   18887400   18812200
cmd/go        13470652   13470492

Change-Id: I3cfd9df4a53783330ba87735853f2a0ec3c42802
Reviewed-on: https://go-review.googlesource.com/c/go/+/256798
Trust: Cherry Zhang <cherryyz@google.com>
Reviewed-by: Than McIntosh <thanm@google.com>
Reviewed-by: Jeremy Faller <jeremy@golang.org>
4 years agocmd/compile: convert last amd64 rules to typed aux
Alberto Donizetti [Sun, 27 Sep 2020 17:13:24 +0000 (19:13 +0200)]
cmd/compile: convert last amd64 rules to typed aux

Passes

  gotip build -toolexec 'toolstash -cmp' -a std

Change-Id: I196d3bdef4a4b650534a4ddd3053e65e0846fdcc
Reviewed-on: https://go-review.googlesource.com/c/go/+/257898
Reviewed-by: Keith Randall <khr@golang.org>
Trust: Alberto Donizetti <alb.donizetti@gmail.com>

4 years agocmd/link, runtime: use a sentinel value for unreachable method
Cherry Zhang [Thu, 24 Sep 2020 20:54:31 +0000 (16:54 -0400)]
cmd/link, runtime: use a sentinel value for unreachable method

In the method table, the method's code pointer is stored as an
offset from the start of the text section. Currently, for an
unreachable method, the offset is left as 0, which resolves to
the start of the text section at run time. It is possible that
there is valid code there. If an unreachable method is ever
reached (due to a compiler or linker bug), the execution will
jump to a wrong location but may continue to run for a while,
until it fails with a seemingly unrelated error.

This CL changes it to use -1 for unreachable method instead. At
run time this will resolve to an invalid address, which makes it
fail immediately if it is ever reached.

Change-Id: Ied6ed7f1833c4f3b991fdf55d8810d70d307b2e6
Reviewed-on: https://go-review.googlesource.com/c/go/+/257203
Trust: Cherry Zhang <cherryyz@google.com>
Run-TryBot: Cherry Zhang <cherryyz@google.com>
Reviewed-by: Than McIntosh <thanm@google.com>
4 years agocrypto/tls: fix TestLinkerGC test
Cherry Zhang [Mon, 28 Sep 2020 21:43:44 +0000 (17:43 -0400)]
crypto/tls: fix TestLinkerGC test

A test that checks if "tls.(*Conn)" appears in any symbol's name.
tls.Conn is a type, so the string "tls.(*Conn)" can only appear
in the name of a method of Conn. But the test code doesn't use
any of the methods. Not sure why this needs to be live. In
particular, the linker is now able to prune all methods of Conn.
Remove this requirement. In fact, just drop the only_conn test
case, as simply allocating a type doesn't necessarily bring
anything live.

Change-Id: I754291b75d38e1465b5291b4dea20806615d21b3
Reviewed-on: https://go-review.googlesource.com/c/go/+/257973
Trust: Cherry Zhang <cherryyz@google.com>
Trust: Tobias Klauser <tobias.klauser@gmail.com>
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Filippo Valsorda <filippo@golang.org>
Reviewed-by: Than McIntosh <thanm@google.com>
Reviewed-by: Jeremy Faller <jeremy@golang.org>
4 years agocmd/go: fix doc math for build cache hashing
Daniel Martí [Sat, 26 Sep 2020 20:00:37 +0000 (21:00 +0100)]
cmd/go: fix doc math for build cache hashing

The function takes five 24-bit chunks from the hash, resulting in 120
bits. When base-64 encoded, this results in a 20-byte output string,
which is confirmed by "var dst [chunks * 4]byte".

It seems like the documented math could have been written for a previous
implementation with shorter hashes, using 4 chunks instead of 5, as then
the math checks out.

Since this code has been working correctly for over three years, let's
fix the documentation to reflect the code.

Change-Id: I9e908e6bafb5dcc1e3c23915e2b6c8843ed444d6
Reviewed-on: https://go-review.googlesource.com/c/go/+/257646
Trust: Daniel Martí <mvdan@mvdan.cc>
Run-TryBot: Daniel Martí <mvdan@mvdan.cc>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Bryan C. Mills <bcmills@google.com>
4 years agoencoding/asn1: error instead of panic on invalid value to Unmarshal
KimMachineGun [Tue, 29 Sep 2020 10:03:38 +0000 (10:03 +0000)]
encoding/asn1: error instead of panic on invalid value to Unmarshal

Changes Unmarshal to return an error, instead of
panicking when its value is nil or not a pointer.

This change matches the behavior of other encoding
packages like json.

Fixes #41509.

Change-Id: I92c3af3a960144566e4c2b55d00c3a6fe477c8d5
GitHub-Last-Rev: c668b6e4ad826f84542c2675eb31ccfb010c45bb
GitHub-Pull-Request: golang/go#41485
Reviewed-on: https://go-review.googlesource.com/c/go/+/255881
Run-TryBot: Emmanuel Odeke <emm.odeke@gmail.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Emmanuel Odeke <emm.odeke@gmail.com>
Reviewed-by: Filippo Valsorda <filippo@golang.org>
Trust: Emmanuel Odeke <emm.odeke@gmail.com>

4 years agocrypto/tls: make config.Clone return nil if the source is nil
Chen.Zhidong [Tue, 29 Sep 2020 09:05:41 +0000 (09:05 +0000)]
crypto/tls: make config.Clone return nil if the source is nil

Fixes #40565

Change-Id: I13a67be193f8cd68df02b8729529e627a73d364b
GitHub-Last-Rev: b03d2c04fd88db909b40dfd7bd08fe13d8994ab9
GitHub-Pull-Request: golang/go#40566
Reviewed-on: https://go-review.googlesource.com/c/go/+/246637
Run-TryBot: Emmanuel Odeke <emm.odeke@gmail.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Filippo Valsorda <filippo@golang.org>
Reviewed-by: Emmanuel Odeke <emm.odeke@gmail.com>
Trust: Emmanuel Odeke <emm.odeke@gmail.com>

4 years agocmd/compile: convert more amd64 rules to typed aux
Alberto Donizetti [Sun, 27 Sep 2020 16:32:18 +0000 (18:32 +0200)]
cmd/compile: convert more amd64 rules to typed aux

Passes

  gotip build -toolexec 'toolstash -cmp' -a std

Change-Id: I2927283e444e7075e155cf29680553b92d471667
Reviewed-on: https://go-review.googlesource.com/c/go/+/257897
Trust: Alberto Donizetti <alb.donizetti@gmail.com>
Reviewed-by: Keith Randall <khr@golang.org>
4 years agocrypto/x509: define certDirectories per GOOS
Tobias Klauser [Mon, 17 Aug 2020 16:26:00 +0000 (18:26 +0200)]
crypto/x509: define certDirectories per GOOS

Split the list of CA certificate directory locations in root_unix.go by
GOOS (aix, *bsd, js, linux, solaris).

On solaris, also include /etc/certs/CA as documented here:
https://docs.oracle.com/cd/E37838_01/html/E61024/kmf-cacerts.html

Same as CL 2208 did for certFiles.

Change-Id: Id24822d6a674bbbbf4088ebb8fe8437edad232b7
Reviewed-on: https://go-review.googlesource.com/c/go/+/248762
Trust: Tobias Klauser <tobias.klauser@gmail.com>
Run-TryBot: Tobias Klauser <tobias.klauser@gmail.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Filippo Valsorda <filippo@golang.org>
4 years agocmd/go/internal/lockedfile/internal/filelock: remove stale TODO comment
Tobias Klauser [Mon, 28 Sep 2020 22:45:12 +0000 (00:45 +0200)]
cmd/go/internal/lockedfile/internal/filelock: remove stale TODO comment

This was addressed by CL 255258.

Updates #35618

Change-Id: I8dd5b30a846f2d16a3d4752304861d7d2178d1cf
Reviewed-on: https://go-review.googlesource.com/c/go/+/257940
Trust: Tobias Klauser <tobias.klauser@gmail.com>
Run-TryBot: Tobias Klauser <tobias.klauser@gmail.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Bryan C. Mills <bcmills@google.com>
4 years agoruntime: initialise cpu.HWCap on netbsd/arm64
Tobias Klauser [Mon, 28 Sep 2020 08:11:06 +0000 (10:11 +0200)]
runtime: initialise cpu.HWCap on netbsd/arm64

NetBSD does not supply AT_HWCAP, however we still need to initialise
cpu.HWCaps.  For now specify the bare minimum until we add some form of
capabilities detection. See
https://golang.org/issue/30824#issuecomment-494901591

Follows CL 174129 which did the same for openbsd/arm64.

Updates #30824

Change-Id: I43a86b583bc60d259a66772703de06970124bb7f
Reviewed-on: https://go-review.googlesource.com/c/go/+/257998
Trust: Tobias Klauser <tobias.klauser@gmail.com>
Trust: Benny Siegert <bsiegert@gmail.com>
Run-TryBot: Tobias Klauser <tobias.klauser@gmail.com>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Reviewed-by: Benny Siegert <bsiegert@gmail.com>
TryBot-Result: Go Bot <gobot@golang.org>

4 years agocmd/link: consider interface conversions only in reachable code
Cherry Zhang [Mon, 21 Sep 2020 03:29:20 +0000 (23:29 -0400)]
cmd/link: consider interface conversions only in reachable code

The linker prunes methods that are not directly reachable if the
receiver type is never converted to interface. Currently, this
"never" is too strong: it is invalidated even if the interface
conversion is in an unreachable function. This CL improves it by
only considering interface conversions in reachable code. To do
that, we introduce a marker relocation R_USEIFACE, which marks
the target symbol as UsedInIface if the source symbol is reached.

binary size    before      after
cmd/compile   18897528   18887400
cmd/go        13607372   13470652

Change-Id: I66c6b69eeff9ae02d84d2e6f2bc7f1b29dd53910
Reviewed-on: https://go-review.googlesource.com/c/go/+/256797
Trust: Cherry Zhang <cherryyz@google.com>
Reviewed-by: Jeremy Faller <jeremy@golang.org>
Reviewed-by: Than McIntosh <thanm@google.com>