net: fix testHookCanceledDial race

Loading and calling testHookCanceledDial in the function passed to
context.AfterFunc is racey, because the function runs in a separate
goroutine, and is not synchronized with the test code that restores
the original testHookCanceledDial value.

We could add a channel and wait for the "AfterFunc" to return in
the deferred function, but that's a lot of synchronization overhead
just for a bit of test code. Instead we simply load
testHookCanceledDial into a local variable synchronously. This fixes
the race without introducing any overhead.

Fixes #75474

Change-Id: If8fbd0f5f65375577c2ded64a13a15b406c45ecc
Reviewed-on: https://go-review.googlesource.com/c/go/+/704455
Reviewed-by: Dmitri Shuralyov <dmitshur@golang.org>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Auto-Submit: Dmitri Shuralyov <dmitshur@golang.org>
Reviewed-by: Damien Neil <dneil@google.com>
Auto-Submit: Damien Neil <dneil@google.com>
Reviewed-by: Dmitri Shuralyov <dmitshur@google.com>

net/http: avoid connCount underflow race

Remove a race condition in counting the number of connections per host,
which can cause a connCount underflow and a panic.

The race occurs when:

  - A RoundTrip call attempts to use a HTTP/2 roundtripper (pconn.alt != nil)
    and receives an isNoCachedConn error. The call removes the pconn from
    the idle conn pool and decrements the connCount for its host.
  - A second RoundTrip call on the same pconn succeeds,
    and delivers the pconn to a third RoundTrip waiting for a conn.
  - The third RoundTrip receives the pconn at the same moment its request
    context is canceled. It places the pconn back into the idle conn pool.

At this time, the connCount is incorrect, because the conn returned to
the idle pool is not matched by an increment in the connCount.

Fix this by not adding HTTP/2 pconns back to the idle pool in
wantConn.cancel.

Fixes #61474

Change-Id: I104d6cf85a54d0382eebf3fcf5dda99c69a7c3f6
Reviewed-on: https://go-review.googlesource.com/c/go/+/703936
Auto-Submit: Damien Neil <dneil@google.com>
Reviewed-by: Nicholas Husin <husin@google.com>
Reviewed-by: Nicholas Husin <nsh@golang.org>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>

context: don't return a non-nil from Err before Done is closed

The Context.Err documentation states that it returns nil if the
context's done channel is not closed. Fix a race condition introduced
by CL 653795 where Err could return a non-nil error slightly before
the Done channel is closed.

No impact on Err performance when returning nil.

Slows down Err when returning non-nil by about 3x,
but that's still almost 2x faster than before CL 653795
and the performance of this path is less important.
(A tight loop checking Err for doneness will be terminated
by the first Err call to return a non-nil result.)

    goos: darwin
    goarch: arm64
    pkg: context
    cpu: Apple M4 Pro
                   │ /tmp/bench.0 │            /tmp/bench.1             │
                   │    sec/op    │   sec/op     vs base                │
    ErrOK-14          1.806n ± 1%   1.774n ± 0%    -1.77% (p=0.000 n=8)
    ErrCanceled-14    1.821n ± 1%   7.525n ± 3%  +313.23% (p=0.000 n=8)
    geomean           1.813n        3.654n       +101.47%

Fixes #75533

Change-Id: Iea22781a199ace7e7f70cf65168c36e090cd2e2a
Reviewed-on: https://go-review.googlesource.com/c/go/+/705235
TryBot-Bypass: Damien Neil <dneil@google.com>
Reviewed-by: Nicholas Husin <husin@google.com>
Reviewed-by: Nicholas Husin <nsh@golang.org>
Auto-Submit: Damien Neil <dneil@google.com>

runtime: make explicit nil check in heapSetTypeSmallHeader

This is another case very similar to CL 684015 and #74375.

In spans with type headers, mallocgc always writes to the page before
returning the allocated memory. This initial write is done by
runtime.heapSetTypeSmallHeader. Prior to the write, the compiler inserts
a nil check, implemented as a dummy instruction reading from memory.

On a freshly mapped page, this read triggers a page fault, mapping the
zero page read-only. Immediately afterwards, the write triggers another
page fault, copying to a writeable page and performing a TLB flush.

This problem is exacerbated as the process scales up. At GOMAXPROCS=6,
the tile38 sweet benchmark spends around 0.1% of cycles directly
handling these page faults. On the same machine at GOMAXPROCS=192, it
spends about 2.7% of cycles directly handling these page faults.

Replacing the read with an explicit nil check reduces the direct cost of
these page faults down to around 0.1% at GOMAXPROCS=192. There are
additional positive side-effects due to reduced contention, so the
overall time spent in page faults drops from around 12.8% to 6.8%. Most
of the remaining time in page faults is spent on automatic NUMA page
migration (completely unrelated to this issue).

Impact on the tile38 benchmark results:

                    │  baseline   │             cl704755              │
                    │   sec/op    │   sec/op     vs base              │
Tile38QueryLoad-192   1.638m ± 3%   1.494m ± 5%  -8.79% (p=0.002 n=6)

                    │     baseline      │              cl704755              │
                    │ average-RSS-bytes │ average-RSS-bytes  vs base         │
Tile38QueryLoad-192        5.384Gi ± 3%        5.399Gi ± 3%  ~ (p=0.818 n=6)

                    │    baseline    │            cl704755             │
                    │ peak-RSS-bytes │ peak-RSS-bytes  vs base         │
Tile38QueryLoad-192     5.818Gi ± 1%     5.864Gi ± 2%  ~ (p=0.394 n=6)

                    │   baseline    │            cl704755            │
                    │ peak-VM-bytes │ peak-VM-bytes  vs base         │
Tile38QueryLoad-192    7.121Gi ± 1%    7.180Gi ± 2%  ~ (p=0.818 n=6)

                    │    baseline     │               cl704755                │
                    │ p50-latency-sec │ p50-latency-sec  vs base              │
Tile38QueryLoad-192       343.2µ ± 1%       313.2µ ± 3%  -8.73% (p=0.002 n=6)

                    │    baseline     │             cl704755             │
                    │ p90-latency-sec │ p90-latency-sec  vs base         │
Tile38QueryLoad-192       1.662m ± 2%       1.603m ± 5%  ~ (p=0.093 n=6)

                    │    baseline     │                cl704755                │
                    │ p99-latency-sec │ p99-latency-sec  vs base               │
Tile38QueryLoad-192       41.56m ± 8%      35.26m ± 18%  -15.17% (p=0.026 n=6)

                    │  baseline   │             cl704755              │
                    │    ops/s    │    ops/s     vs base              │
Tile38QueryLoad-192   87.89k ± 3%   96.36k ± 4%  +9.64% (p=0.002 n=6)

Updates #74375.

Change-Id: I6a6a636c1a16261b6d5076f2e1b08524a6544d33
Reviewed-on: https://go-review.googlesource.com/c/go/+/704755
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Michael Knyszek <mknyszek@google.com>
Auto-Submit: Michael Pratt <mpratt@google.com>

cmd/internal/obj/riscv: add support for Zicond instructions

This patch implement assembler for the Zicond extension: CZEROEQZ and CZERONEZ.

Follow-up to CL 631576
Updates #75350

Change-Id: Icf4be131fe61c3b7a3bde4811cf42dc807660907
GitHub-Last-Rev: 6539cc86cbf3c49c3247ed935bcbbb31bb886dea
GitHub-Pull-Request: golang/go#75408
Reviewed-on: https://go-review.googlesource.com/c/go/+/702677
Reviewed-by: Mark Freeman <markfreeman@google.com>
Reviewed-by: Joel Sing <joel@sing.id.au>
Reviewed-by: Meng Zhuo <mengzhuo1203@gmail.com>
Reviewed-by: Michael Knyszek <mknyszek@google.com>
Reviewed-by: Mark Ryan <markdryan@rivosinc.com>
TryBot-Bypass: Joel Sing <joel@sing.id.au>

cmd/compile: fix the issue of shift amount exceeding the valid range

Fixes #75479

Change-Id: I362d3e49090e94f91a840dd5a475978b59222a00
Reviewed-on: https://go-review.googlesource.com/c/go/+/704135
Reviewed-by: Mark Freeman <markfreeman@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Michael Knyszek <mknyszek@google.com>
Reviewed-by: Meidan Li <limeidan@loongson.cn>
Reviewed-by: abner chenc <chenguoqi@loongson.cn>

runtime: don't enable heap randomization if MSAN or ASAN is enabled

MSAN and ASAN do confusing things to the memory layout, which are likely
to conflict with heap base randomization, so if they are enabled,
ignore randomizedHeapBase64.

We already didn't turn it on when TSAN was enabled.

Change-Id: I41e59dfc33d8bb059c208a9595442571fb31eea3
Reviewed-on: https://go-review.googlesource.com/c/go/+/704856
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Michael Knyszek <mknyszek@google.com>
Auto-Submit: Roland Shoemaker <roland@golang.org>

runtime: fix CheckScavengedBitsCleared with randomized heap base

We cannot easily determine the base address of the arena we added the
random padding pages to, which can cause confusing
TestScavengedBitsCleared failures when the arena we padded does not have
the lowest address of all of the arenas.

Instead of attempting to determine the correct arena to ignore when
checking the scav and alloc bits, switch to just tolerating _one_ arena
having mismatches, which is expected when randomizedHeapBase64 is
enabled. Any other number of arenas having mismatches is likely a real
error.

Fixes #75502

Change-Id: Iacc445b2905824f9f71970c7abd33f187793cf39
Reviewed-on: https://go-review.googlesource.com/c/go/+/704855
Reviewed-by: Michael Knyszek <mknyszek@google.com>
Auto-Submit: Roland Shoemaker <roland@golang.org>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>

encoding/json/v2: fix typo in comment

Use the correct spelling of the Gregorian calendar.

Change-Id: I7e1d2974d38d5d3ded64f98852ee726c7b84be22
Reviewed-on: https://go-review.googlesource.com/c/go/+/704595
Reviewed-by: Joseph Tsai <joetsai@digital-static.net>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Cherry Mui <cherryyz@google.com>
Reviewed-by: Michael Knyszek <mknyszek@google.com>
Auto-Submit: Emmanuel Odeke <emmanuel@orijtech.com>

testing: use reflect.TypeAssert and reflect.TypeFor

To simplify the code.

Updates #62121
Updates #60088

Change-Id: I1cef27b19961721b89a965c708320072a052d9af
GitHub-Last-Rev: ffd07a74da87b84b457246b2c94d4ee7540c8e8b
GitHub-Pull-Request: golang/go#75494
Reviewed-on: https://go-review.googlesource.com/c/go/+/704575
Reviewed-by: Alan Donovan <adonovan@google.com>
Auto-Submit: Alan Donovan <adonovan@google.com>
Reviewed-by: Michael Knyszek <mknyszek@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>

path/filepath: fix EvalSymlinks to return ENOTDIR on plan9

CL 155597 added a test to check EvalSymlinks returns ENOTDIR
for existing path which ends with a slash.

CL 404954 added a separate evalSymlinks implementation for plan9,
which has a kludge to ensure the ENOTDIR is returned in the above case.

Apparently the added kludge is not working now (and maybe never worked),
as seen in [1]:

=== RUN   TestIssue29372
    path_test.go:1730: test#0: want "not a directory", got "stat /tmp/TestIssue293724085649913/001/file.txt/: not a directory: '/tmp/TestIssue293724085649913/001/file.txt/'"
--- FAIL: TestIssue29372 (0.00s)

This happens because on plan9 errors are strings and they contain the
file name after the error text, and the check assumes the "not a
directory" text is at the end of the message.

The fix is to check whether the error is somewhere in the error text,
not ends with it (as it is done in other plan9 code, for example, see
checkErrMessageContent added in CL 163058).

This should fix the above test failure.

PS perhaps the kludge should be moved up the call chain to os.Stat/Lstat?

[1]: https://ci.chromium.org/ui/p/golang/builders/try/gotip-plan9-amd64/b8704772617873749345/test-results?q=ExactID%3Apath%2Ffilepath.TestIssue29372+VHash%3Aa63f5798e9f8f502
Change-Id: I6afb68be5f65a28929b2f717247ab34ff3b4a812
Reviewed-on: https://go-review.googlesource.com/c/go/+/700775
Auto-Submit: Emmanuel Odeke <emmanuel@orijtech.com>
Reviewed-by: Michael Knyszek <mknyszek@google.com>
Reviewed-by: Dmitri Shuralyov <dmitshur@google.com>
Reviewed-by: Richard Miller <millerresearch@gmail.com>
TryBot-Bypass: Dmitri Shuralyov <dmitshur@golang.org>
Reviewed-by: Dmitri Shuralyov <dmitshur@golang.org>

go/parser: Add TestBothLineAndLeadComment

I was wondering whether is is expected that both p.lineComment
and p.leadComment might be populated at the same time.

i.e. whether parser.go:275 can be changed from:

if p.lineFor(p.pos) != endline || p.tok == token.SEMICOLON || p.tok == token.EOF

to:

if (p.tok != token.COMMENT && p.lineFor(p.pos) != endline) || p.tok == token.SEMICOLON || p.tok == token.EOF

It turns out that we cannot do so. So while i am here, add a test
case for that, since nothing else failed with that change.

Change-Id: I6a6a6964f760237c068098db8a7b4b7aaf26c651
Reviewed-on: https://go-review.googlesource.com/c/go/+/703915
Reviewed-by: Michael Knyszek <mknyszek@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Alan Donovan <adonovan@google.com>

go/ast: document that (*ast.File).Comments is sorted by position

This is the way that go/parser produces it and in every AST manipulation
that we do, we preserve such property. Also printer expects it is such
order.

Change-Id: I6a6a696424a679a2cf1f9bb7eb41bdd47523efa9
Reviewed-on: https://go-review.googlesource.com/c/go/+/704255
Reviewed-by: Michael Knyszek <mknyszek@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Alan Donovan <adonovan@google.com>
Auto-Submit: Alan Donovan <adonovan@google.com>

cmd/link: simplify PE relocations mapping

The code for mapping Windows PE relocations to Go relocations was
difficult to follow and contains some duplicated code. Also, it was
mapping IMAGE_REL_AMD64_ADDR32 to R_PCREL instead of R_ADDR.

This CL commit simplifies the code and fixes the mapping. I haven't been
able to coerce mingw-w64 to generate IMAGE_REL_AMD64_ADDR32 relocations,
so I haven't been able to test this change. However, the previous
implementation was clearly wrong.

While here, remove code supporting the unsupported windows/arm support.

Updates #71671
Updates #75485

Change-Id: Id0d6f352fa7d5df9e00509fcdf09ca0cb91ca524
Reviewed-on: https://go-review.googlesource.com/c/go/+/672155
Reviewed-by: Michael Knyszek <mknyszek@google.com>
Reviewed-by: Cherry Mui <cherryyz@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>

go/parser: simplify expectSemi

Move the outer if statement, directly into the switch.

Change-Id: I6a6a6964ff15dbcda4f4c9ae1c15423adbfb0ae5
Reviewed-on: https://go-review.googlesource.com/c/go/+/703835
Reviewed-by: Michael Knyszek <mknyszek@google.com>
Reviewed-by: Alan Donovan <adonovan@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>

internal/buildcfg: enable randomizedHeapBase64 by default

Change-Id: If2493546cd9c4e5bb6736fe3c029e60957716f6a
Reviewed-on: https://go-review.googlesource.com/c/go/+/690296
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Michael Knyszek <mknyszek@google.com>

os/user: align test file name and build tags

TestNegativeUid tests code in cgo_lookup_unix.go, so let's use
the very same go:build line for the test file, and rename
the test file accordingly.

Change-Id: Iee3dbc25aeb8a7659c734d97dde3e9c670e80fae
Reviewed-on: https://go-review.googlesource.com/c/go/+/702976
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Mark Freeman <markfreeman@google.com>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Reviewed-by: Michael Knyszek <mknyszek@google.com>

runtime/_mkmalloc: add a copy of cloneNode

cloneNode is defined in golang.org/x/tools/internal/astutil. Make a copy
of it so we can easily clone AST nodes.

Change-Id: I6a6a6964132e663e64faa00fa6037cf6e8d4cbc6
Reviewed-on: https://go-review.googlesource.com/c/go/+/703396
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Michael Knyszek <mknyszek@google.com>
Reviewed-by: Michael Matloob <matloob@google.com>

cmd/go: check pattern for utf8 validity before call regexp.MustCompile

Do not panic if the package path or the package version contains
invalid UTF-8 characters.

Fixes #75251

Change-Id: Ib787e74277cf814253857b911d378ea5e53d8824
Reviewed-on: https://go-review.googlesource.com/c/go/+/700815
Reviewed-by: Michael Matloob <matloob@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Ian Alexander <jitsu@google.com>
Reviewed-by: Michael Matloob <matloob@golang.org>

cmd/go: disable cgo by default if CC unset and  DefaultCC doesn't exist

CL 621995 disrupted the behavior
introduced by CL 450739, now restore it.

Fixes #75340

Change-Id: Icd1a0eb970876995f9446e0547ceb9e78990f6ed
Reviewed-on: https://go-review.googlesource.com/c/go/+/703555
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Michael Matloob <matloob@golang.org>
Reviewed-by: Michael Matloob <matloob@google.com>
Reviewed-by: Ian Alexander <jitsu@google.com>
Reviewed-by: Sean Liao <sean@liao.dev>

bytes,strings: remove reference to non-existent SplitFunc

Fixes #75462
Updates #72841

Change-Id: Ie2cbbdb031578a3138ecc6e60c0025807f1990e6
Reviewed-on: https://go-review.googlesource.com/c/go/+/703675
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Auto-Submit: Alan Donovan <adonovan@google.com>
Reviewed-by: Michael Knyszek <mknyszek@google.com>
Reviewed-by: Alan Donovan <adonovan@google.com>

cmd/compile: fix mips zerorange implementation

A +8 was missed when simplifying this code in CL 700936.

Fixes #75477

Change-Id: Ic7b83322dc1373432b44f0b63607141195220380
Reviewed-on: https://go-review.googlesource.com/c/go/+/703937
Reviewed-by: Keith Randall <khr@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Julian Zhu <jz531210@gmail.com>
Auto-Submit: Keith Randall <khr@google.com>
Reviewed-by: Tianon Gravi <admwiggin@gmail.com>
Reviewed-by: Michael Knyszek <mknyszek@google.com>

cmd/internal/obj/riscv: improve handling of float point moves

Translate moves from an integer register to a floating point register, or
from a floating point register to an integer register, to the appropriate
move instruction (i.e. FMVXW/FMVWX/FMVXD/FMVDX).

Add support for MOVF with a constant - we previously added support for MOVD
but not for MOVF. Add special handling for 0.0, which we can translate to
a move from the zero register to a floating point register (leveraging the
above mentioned change).

Change-Id: If8df2f5610e69b4ec0af85efb884951024685f5b
Reviewed-on: https://go-review.googlesource.com/c/go/+/703216
Reviewed-by: Meng Zhuo <mengzhuo1203@gmail.com>
Reviewed-by: Mark Freeman <markfreeman@google.com>
Reviewed-by: Michael Knyszek <mknyszek@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Mark Ryan <markdryan@rivosinc.com>

crypto/x509/internal/macos: standardize package name

Rename "macOS" to "macos".

Change-Id: I33ca7afaa14c910a97a19ead518894fb4917662c
Reviewed-on: https://go-review.googlesource.com/c/go/+/688695
Reviewed-by: Michael Knyszek <mknyszek@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Mark Freeman <markfreeman@google.com>
Reviewed-by: Daniel McCarney <daniel@binaryparadox.net>
Reviewed-by: Roland Shoemaker <roland@golang.org>

iter: document that calling yield after terminated range loop causes runtime panic

The existing documentation isn't clear about the behaviour of when
the iterator function ignores the value returned from the yield function.

The changes here update the documentation, to explicitly explain
that.

Change-Id: I24a8198c3da63429358554169697fa466345b8fd
GitHub-Last-Rev: 86c8a2dd89dca4e05631f5c4f7087b59b0e0664b
GitHub-Pull-Request: golang/go#74561
Reviewed-on: https://go-review.googlesource.com/c/go/+/687215
Auto-Submit: Sean Liao <sean@liao.dev>
Reviewed-by: Sean Liao <sean@liao.dev>
Reviewed-by: Michael Knyszek <mknyszek@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Mark Freeman <markfreeman@google.com>

cmd/link: remove support for windows/arm relocations

The windows/arm port is no longer supported. We can remove the
related code from cmd/link/internal/arm.

For #71671

Change-Id: I00de1231482cc2f28ec5fc9dc62e81f0ba3fe481
Reviewed-on: https://go-review.googlesource.com/c/go/+/703778
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Cherry Mui <cherryyz@google.com>
Reviewed-by: Than McIntosh <thanm@golang.org>
Reviewed-by: Michael Knyszek <mknyszek@google.com>

cmd/link: remove support for darwin/386 relocations

The darwin/386 port has been unsupported for years, but some
relocation handling specific to it has managed to survive till now.

Updates #37610

Change-Id: I27ae2ac5462c5f3ec219e9cb5dce6f83b037b816
Reviewed-on: https://go-review.googlesource.com/c/go/+/703777
Reviewed-by: Michael Knyszek <mknyszek@google.com>
Reviewed-by: Than McIntosh <thanm@golang.org>
Reviewed-by: Cherry Mui <cherryyz@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>

cmd/internal/obj/loong64: use the MOVVP instruction to optimize prologue

The MOVVP instruction has a larger offset range, and removes 928
instructions from the go binary on loong64.

file        before      after       Δ
go          1634208     1634064    -144
gofmt       323324      323240     -84
asm         567870      567778     -92
cgo         487694      487598     -96
compile     2500266     2500142    -124
cover       530590      530498     -92
link        723804      723692     -112
preprofile  240562      240474     -88
vet         819672      819576     -96

Change-Id: Ib0efcb006d3ae3f2bceec0d6e88f3794d5e99831
Reviewed-on: https://go-review.googlesource.com/c/go/+/702715
Reviewed-by: abner chenc <chenguoqi@loongson.cn>
Reviewed-by: Mark Freeman <markfreeman@google.com>
Reviewed-by: Meidan Li <limeidan@loongson.cn>
Reviewed-by: Michael Knyszek <mknyszek@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>

cmd/internal/obj/riscv: implement vector segment load/store instructions

https://github.com/riscv/riscv-opcodes/pull/361. After this pr was
merged, riscv-opcode can generate RVV segment load/store instructions
for Go. Implement vector segment load/store instuctions.

Change-Id: I154bb75be70c0a45e2279a75c67f68b5bb57c36e
Reviewed-on: https://go-review.googlesource.com/c/go/+/691695
Reviewed-by: Mark Freeman <markfreeman@google.com>
Reviewed-by: Michael Knyszek <mknyszek@google.com>
Reviewed-by: Meng Zhuo <mengzhuo1203@gmail.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>

cmd/asm: add double precision comparision testcases for riscv64

Change-Id: If8e03dfdb332a22ec9c6a0021d7e7955520f3ddc
Reviewed-on: https://go-review.googlesource.com/c/go/+/702136
Reviewed-by: Mark Freeman <markfreeman@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Mark Ryan <markdryan@rivosinc.com>
Reviewed-by: Joel Sing <joel@sing.id.au>
Reviewed-by: Michael Knyszek <mknyszek@google.com>

runtime: fix TestSehUnwind

The various TestSehUnwind tests recently started to fail due an
unrelated refactor (in CL 698098) that made the stack frames to
not match the expected pattern. This CL updates the tests to
be more robust to such changes.

Fixes #75467

Change-Id: I7950332bb6ca54e4bf693d13e2490b3d9d901dde
Reviewed-on: https://go-review.googlesource.com/c/go/+/703779
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Michael Knyszek <mknyszek@google.com>
Auto-Submit: Quim Muntal <quimmuntal@gmail.com>
Reviewed-by: Cherry Mui <cherryyz@google.com>

all: replace strings.Split with strings.SplitSeq

In Go 1.25+, strings.SplitSeq offers better
performance. Here are the benchmark results comparing
strings.Split and strings.SplitSeq in a for-loop, with the
benchmark code located in src/strings/iter_test.go:

goos: darwin
goarch: amd64
pkg: cmd/go/internal/auth
cpu: Intel(R) Core(TM) i7-8569U CPU @ 2.80GHz
                         │   old.txt   │               new.txt                │
                         │   sec/op    │    sec/op     vs base                │
ParseGitAuth/standard-8    281.4n ± 1%   218.0n ± 11%  -22.54% (p=0.000 n=10)
ParseGitAuth/with_url-8    549.1n ± 1%   480.5n ± 13%  -12.48% (p=0.002 n=10)
ParseGitAuth/minimal-8     235.4n ± 1%   197.3n ±  7%  -16.20% (p=0.000 n=10)
ParseGitAuth/complex-8     797.6n ± 2%   805.2n ±  4%        ~ (p=0.481 n=10)
ParseGitAuth/empty-8       87.48n ± 3%   63.25n ±  6%  -27.71% (p=0.000 n=10)
ParseGitAuth/malformed-8   228.8n ± 1%   171.2n ±  3%  -25.17% (p=0.000 n=10)
geomean                    288.9n        237.7n        -17.72%

                         │   old.txt   │              new.txt               │
                         │    B/op     │    B/op     vs base                │
ParseGitAuth/standard-8    192.00 ± 0%   96.00 ± 0%  -50.00% (p=0.000 n=10)
ParseGitAuth/with_url-8     400.0 ± 0%   288.0 ± 0%  -28.00% (p=0.000 n=10)
ParseGitAuth/minimal-8     144.00 ± 0%   80.00 ± 0%  -44.44% (p=0.000 n=10)
ParseGitAuth/complex-8      528.0 ± 0%   400.0 ± 0%  -24.24% (p=0.000 n=10)
ParseGitAuth/empty-8        32.00 ± 0%   16.00 ± 0%  -50.00% (p=0.000 n=10)
ParseGitAuth/malformed-8   176.00 ± 0%   80.00 ± 0%  -54.55% (p=0.000 n=10)
geomean                     179.0        102.1       -42.96%

                         │  old.txt   │              new.txt               │
                         │ allocs/op  │ allocs/op   vs base                │
ParseGitAuth/standard-8    3.000 ± 0%   2.000 ± 0%  -33.33% (p=0.000 n=10)
ParseGitAuth/with_url-8    4.000 ± 0%   3.000 ± 0%  -25.00% (p=0.000 n=10)
ParseGitAuth/minimal-8     3.000 ± 0%   2.000 ± 0%  -33.33% (p=0.000 n=10)
ParseGitAuth/complex-8     4.000 ± 0%   3.000 ± 0%  -25.00% (p=0.000 n=10)
ParseGitAuth/empty-8       2.000 ± 0%   1.000 ± 0%  -50.00% (p=0.000 n=10)
ParseGitAuth/malformed-8   3.000 ± 0%   2.000 ± 0%  -33.33% (p=0.000 n=10)
geomean                    3.086        2.040       -33.91%

Updates #69315.

Change-Id: Id0219edea45d9658d527b863162ebe917e7821d9
GitHub-Last-Rev: 392b315e122f2c9ef8703ca2dbce8f82ec198556
GitHub-Pull-Request: golang/go#75259
Reviewed-on: https://go-review.googlesource.com/c/go/+/701015
Reviewed-by: Keith Randall <khr@golang.org>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Michael Knyszek <mknyszek@google.com>
Reviewed-by: Emmanuel Odeke <emmanuel@orijtech.com>
Reviewed-by: Keith Randall <khr@google.com>
Auto-Submit: Emmanuel Odeke <emmanuel@orijtech.com>

all: remove redundant words in comment

Change-Id: I8e6be76f24a7b1bb21cc244d5117fa60efff03b3
Reviewed-on: https://go-review.googlesource.com/c/go/+/703317
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Auto-Submit: Sean Liao <sean@liao.dev>
Reviewed-by: Than McIntosh <thanm@golang.org>
Reviewed-by: Michael Knyszek <mknyszek@google.com>
Reviewed-by: Mark Freeman <markfreeman@google.com>
Reviewed-by: Sean Liao <sean@liao.dev>
Auto-Submit: Emmanuel Odeke <emmanuel@orijtech.com>

cmd/compile: combine doubling with shift on riscv64

Change-Id: I4bee2770fedf97e35b5a5b9187a8ba3c41f9ec2e
Reviewed-on: https://go-review.googlesource.com/c/go/+/702697
Reviewed-by: Keith Randall <khr@google.com>
Reviewed-by: Michael Knyszek <mknyszek@google.com>
Reviewed-by: Joel Sing <joel@sing.id.au>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Auto-Submit: Keith Randall <khr@google.com>

net/http: clarify panic conditions in Handle, HandleFunc, AddInsecureBypassPattern

Add explicit mention that these methods panic on both pattern conflict
and invalid syntax.

Fixes #75226

Change-Id: If7dbfc44e1ec4624ab156f0e5d7e66cee2c2fef3
GitHub-Last-Rev: acc9a9c3338373f5a9376ff2eb1decd30bbee04d
GitHub-Pull-Request: golang/go#75297
Reviewed-on: https://go-review.googlesource.com/c/go/+/701016
Reviewed-by: Michael Knyszek <mknyszek@google.com>
Auto-Submit: Sean Liao <sean@liao.dev>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Emmanuel Odeke <emmanuel@orijtech.com>
Reviewed-by: Sean Liao <sean@liao.dev>
Reviewed-by: Mark Freeman <markfreeman@google.com>

crypto/internal/cryptotest: use linux-amd64_avx512 builder for SHA-NI

Updates #74772
Fixes #69592

Change-Id: I6a6a69647ab1785ed953f3ed2dfa04cd55576f2b
Cq-Include-Trybots: luci.golang.try:gotip-linux-amd64_avx512
Reviewed-on: https://go-review.googlesource.com/c/go/+/701315
Reviewed-by: Daniel McCarney <daniel@binaryparadox.net>
Reviewed-by: Mark Freeman <markfreeman@google.com>
Auto-Submit: Filippo Valsorda <filippo@golang.org>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Michael Knyszek <mknyszek@google.com>

cmd/go,crypto/internal/fips140: prevent using FIPS 140-3 mode with purego tag

Change-Id: I6a6a696414f8d5d9dc77c65b0ac9edfc982c2798
Reviewed-on: https://go-review.googlesource.com/c/go/+/703095
Auto-Submit: Filippo Valsorda <filippo@golang.org>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Mark Freeman <markfreeman@google.com>
Reviewed-by: Daniel McCarney <daniel@binaryparadox.net>
Reviewed-by: Michael Knyszek <mknyszek@google.com>

crypto/internal/cryptotest: add MustSupportFIPS140

Change-Id: Iccb3ffdb7c1e15f8f0d889fdc4a27e62870cd58e
Reviewed-on: https://go-review.googlesource.com/c/go/+/702776
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Daniel McCarney <daniel@binaryparadox.net>
Reviewed-by: Mark Freeman <markfreeman@google.com>
Auto-Submit: Filippo Valsorda <filippo@golang.org>
Reviewed-by: Michael Knyszek <mknyszek@google.com>

cmd/internal/testdir: re-enable default all codegen flag on linux-amd64

This was limited to just the mainline linux-amd64 builder, but we don't
use that name anymore. Use the LUCI name instead, gotip-linux-amd64.

Change-Id: Ib4377ad336c529512d9939ff9dce0ea242528b74
Reviewed-on: https://go-review.googlesource.com/c/go/+/703136
Auto-Submit: Michael Knyszek <mknyszek@google.com>
Reviewed-by: Dmitri Shuralyov <dmitshur@golang.org>
Reviewed-by: Dmitri Shuralyov <dmitshur@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>

all: replace os.Getenv("GO_BUILDER_NAME") with testenv.Builder in tests

Some tests still reach for GO_BUILDER_NAME directly. This change makes
it so that they go through testenv.Builder.

There are a couple more, but changing them may also cause tests to start
failing. Done in a follow-up.

Change-Id: Id2453b7b62f5ebf3594e92fa53724a577a97440f
Reviewed-on: https://go-review.googlesource.com/c/go/+/703135
Reviewed-by: Dmitri Shuralyov <dmitshur@golang.org>
Reviewed-by: Dmitri Shuralyov <dmitshur@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>

cmd: vendor x/tools@9fccddc

(This brings in CL 703995.)

For #71859

Change-Id: I823f2da9ebbdbef943cb37123d44a7ad2e6d708b
Reviewed-on: https://go-review.googlesource.com/c/go/+/703896
Reviewed-by: Michael Matloob <matloob@golang.org>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Auto-Submit: Alan Donovan <adonovan@google.com>
Reviewed-by: Michael Matloob <matloob@google.com>

os: add (*Process).WithHandle

Implement (*Process).WithHandle, add tests for all platforms.

Fixes #70352

Change-Id: I7a8012fb4e1e1b4ce1e75a59403ff6e77504fc56
Reviewed-on: https://go-review.googlesource.com/c/go/+/699615
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Mark Freeman <markfreeman@google.com>
Auto-Submit: Kirill Kolyshkin <kolyshkin@gmail.com>
Reviewed-by: Michael Pratt <mpratt@google.com>

os: add and use errProcessReleased

This error is already used in three places, so let's define it.

Change-Id: I73565d94aebcf3d5a278201d96839d82db85a2d4
Reviewed-on: https://go-review.googlesource.com/c/go/+/702436
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Mark Freeman <markfreeman@google.com>
Reviewed-by: Sean Liao <sean@liao.dev>
Reviewed-by: Michael Pratt <mpratt@google.com>

internal/syscall/unix: add KernelVersionGE

There are a few places in the code which checks that the running kernel
is greater than or equal to x.y. The check takes a few lines and the
checking code is somewhat distracting.

Let's abstract this check into a simple function, KernelVersionGE,
and convert the users accordingly.

Add a test case (I'm not sure it has much value, can be dropped).

Change-Id: I8ec91dcc7452363361f95e46794701c0ae57d956
Reviewed-on: https://go-review.googlesource.com/c/go/+/700796
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Damien Neil <dneil@google.com>
Reviewed-by: Mark Freeman <markfreeman@google.com>

cmd/link: support race mode with MSVC clang

I couldn't make --print-file-name work with -msvc clang. (The
library name is synchronization.lib, but even with that name it
still doesn't print the full path.) Assume it always
synchronization.lib.

Change-Id: I22e8f14824f7f7e96b71b913217b1f604f1e2da7
Reviewed-on: https://go-review.googlesource.com/c/go/+/703398
Reviewed-by: Than McIntosh <thanm@golang.org>
Reviewed-by: Florian Zenker <floriank@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>

test/codegen: check zerobase for newobject on 0-sized types

This CL also adds riscv64 checks

Change-Id: I693e4e606f470615f6b49085592d6d5ca61473d3
Reviewed-on: https://go-review.googlesource.com/c/go/+/703716
Reviewed-by: Pengcheng Wang <wangpengcheng.pp@bytedance.com>
Auto-Submit: Keith Randall <khr@google.com>
Reviewed-by: Mark Freeman <markfreeman@google.com>
Reviewed-by: Joel Sing <joel@sing.id.au>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Keith Randall <khr@google.com>

net/http: add comments for transport gzip reader

Add back comments dropped in CL 510255

Updates #61353

Change-Id: Ie84610b830599e178140bb5c2a11be74b2ad461e
GitHub-Last-Rev: d6cd890dd15aa0f2f8701cf66da11a844bbd011d
GitHub-Pull-Request: golang/go#75293
Reviewed-on: https://go-review.googlesource.com/c/go/+/701395
Reviewed-by: Sean Liao <sean@liao.dev>
Reviewed-by: Mark Freeman <markfreeman@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Michael Pratt <mpratt@google.com>
Auto-Submit: Michael Pratt <mpratt@google.com>

cmd/compile,cmd/gofmt: use reflect.TypeFor

Use "reflect.TypeFor" to simplify the code.

Updates #60088

Change-Id: I93db6cbd4f02813d9a81f5d02996db8128cb81a9
GitHub-Last-Rev: 2aee64dac6e13ef869aa73f2abf236650e1c1757
GitHub-Pull-Request: golang/go#75349
Reviewed-on: https://go-review.googlesource.com/c/go/+/701676
Reviewed-by: Mark Freeman <markfreeman@google.com>
Auto-Submit: Keith Randall <khr@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
Reviewed-by: Keith Randall <khr@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>

runtime: deduplicate syscall assembly for darwin

The darwin port provides different syscall functions that only
differ on how they handle the errors, and they are all written
in assembly.

This duplication can be removed by factoring out the error handling
logic to arch-agnostic Go code and leaving the assembly functions
with the only reponsibility of making the syscall and mapping
parameters between ABIs.

Updates #51087

Cq-Include-Trybots: luci.golang.try:gotip-darwin-arm64-longtest,gotip-darwin-amd64-longtest
Change-Id: I9524377f3ef9c9a638412c7e87c8f46a33ee3453
Reviewed-on: https://go-review.googlesource.com/c/go/+/699135
Reviewed-by: Michael Pratt <mpratt@google.com>
Reviewed-by: Mark Freeman <markfreeman@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>

net: use context.AfterFunc in connect

This saves a goroutine when ctx can be but is not canceled during
the connect call.

The redundant fd.Close() call is removed, because the caller closes
the fd on error.

Change-Id: I124d7e480294a48ef74d5650d8ef0489bdfc64d6
Reviewed-on: https://go-review.googlesource.com/c/go/+/699256
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Sean Liao <sean@liao.dev>
Reviewed-by: Damien Neil <dneil@google.com>
Auto-Submit: Sean Liao <sean@liao.dev>
Reviewed-by: Mark Freeman <markfreeman@google.com>

cmd/go/internal/work: copy vet tool's stdout to our stdout

The go command connects both the stdout and stderr files of
its child commands (cmd/compile, cmd/vet, etc) to the go
command's own stderr. If the child command is supposed to
produce structure output on stderr, as is the case for
go vet -json or go fix -diff, it will be merged with the
error stream, making it useless.

This change to the go vet <-> unitchecker protocol specifies
the name of a file into which the vet tool should write its
stdout. On success, the go command will then copy the entire
content of that file to its own stdout, under a lock.
This ensures that partial writes to stdout in case of failure,
concurrent writes to stdout by parallel vet tasks, or other
junk on stderr, cannot interfere with the integrity of the
go command's structure output on stdout.

CL 702835 is the corresponding change on the x/tools side.

For #75432

Change-Id: Ib4db25b6b0095d359152d7543bd9bf692551bbfa
Reviewed-on: https://go-review.googlesource.com/c/go/+/702815
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Michael Matloob <matloob@google.com>
Auto-Submit: Alan Donovan <adonovan@google.com>

runtime: move Windows types and consts to internal/runtime/syscall/windows

This CL doesn't change any behavior, it just moves code around to reduce
the size of the runtime package and remove some duplicated symbols.

Updates #51087.

Cq-Include-Trybots: luci.golang.try:gotip-windows-arm64
Change-Id: I3d3e5f214f045c24fb5d4050d56e7b0822a6e4b5
Reviewed-on: https://go-review.googlesource.com/c/go/+/698098
Reviewed-by: Cherry Mui <cherryyz@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Michael Pratt <mpratt@google.com>

vendor: update x/tools to 3adf0e9, and other repos

Notably, the x/tools update includes CL 702835.

Also, pacify copyright_test report of missing copyright
header in generated h2_bundle.

Updates golang/go#75432

Change-Id: I428278e50dbcef5dcaa661004da0da9ab8f2c924
Reviewed-on: https://go-review.googlesource.com/c/go/+/702955
Reviewed-by: Dmitri Shuralyov <dmitshur@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Dmitri Shuralyov <dmitshur@golang.org>

cmd/link: don't count tbss section in TestFlagD

TestFlagD looks for a data-like section at the lowest address.
On OpenBSD, the .tbss section matches the current condition, which
has address 0, causing the test fail. Don't count TLS sections.

Also, print the section name on error.

Fixes #75444.

Change-Id: Ic0aa1a2bb7c6bd5c0023d4482405a482095ff68b
Reviewed-on: https://go-review.googlesource.com/c/go/+/703375
Reviewed-by: Michael Pratt <mpratt@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>

cmd/compile: optimize abi.Type.GCData loads

This fires in just one place; stkframe.go's stkobjinit. But, it does
make the struct literal entirely constants.

Change-Id: Ice76cb3cddd97adee011fdaab40597839da2e89f
Reviewed-on: https://go-review.googlesource.com/c/go/+/701300
Reviewed-by: Keith Randall <khr@golang.org>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Keith Randall <khr@google.com>
Reviewed-by: Mark Freeman <markfreeman@google.com>
Auto-Submit: Michael Pratt <mpratt@google.com>

cmd/compile, reflect: further allow inlining of TypeFor

Previous CLs optimized direct use of abi.Type, but reflect.Type is
indirected, so was not benefiting.

For TypeFor, we can use toRType directly without a nil check because the
types are statically known.

Normally, I'd think SSA would remove the nil check, but due to some
oddity (specifically, late fuse being required to remove the nil check,
but opt doesn't run that late) means that the nil check persists and
gets in the way.

Manually writing the code in this instance seems to fix the problem.

It also exposed another problem; depending on the ordering, writeType
could get to a type symbol before SSA, thereby preventing Extra from
being created on the symbol for later lookups that don't go through
TypeLinksym directly. In writeType, for non-shape types, call
TypeLinksym to ensure that the type is set up for later callers. That
change itself passed toolstash -cmp.

All up, this stack put through compilecmp shows a lot of improvement in
various reflect-using packages, and reflect itself. It is too big to fit
in the commit message but here's some info:

compilecmp master -> HEAD
master (d767064170): cmd/compile: mark abi.PtrType.Elem sym as used
HEAD (846a94c568): cmd/compile, reflect: further allow inlining of TypeFor

file      before    after     Δ       %
addr2line 3735911   3735391   -520    -0.014%
asm       6382235   6382091   -144    -0.002%
buildid   3608568   3608360   -208    -0.006%
cgo       5951816   5951480   -336    -0.006%
compile   28362080  28339772  -22308  -0.079%
cover     6668686   6661414   -7272   -0.109%
dist      4311961   4311425   -536    -0.012%
fix       3771706   3771474   -232    -0.006%
link      8686073   8684993   -1080   -0.012%
nm        3715923   3715459   -464    -0.012%
objdump   6074366   6073774   -592    -0.010%
pack      3025653   3025277   -376    -0.012%
pprof     18269485  18261653  -7832   -0.043%
test2json 3442726   3438390   -4336   -0.126%
trace     16984831  16981767  -3064   -0.018%
vet       10701931  10696355  -5576   -0.052%
total     133693951 133639075 -54876  -0.041%

runtime
runtime.stkobjinit 240 -> 165  (-31.25%)

runtime [cmd/compile]
runtime.stkobjinit 240 -> 165  (-31.25%)

reflect
reflect.Value.Seq2.func3 309 -> 245  (-20.71%)
reflect.Value.Seq2.func1.1 281 -> 198  (-29.54%)
reflect.Value.Seq.func1.1 242 -> 165  (-31.82%)
reflect.Value.Seq2.func2 360 -> 285  (-20.83%)
reflect.Value.Seq.func4 281 -> 239  (-14.95%)
reflect.Value.Seq2.func4 399 -> 284  (-28.82%)
reflect.Value.Seq.func2 271 -> 230  (-15.13%)
reflect.TypeFor[go.shape.uint64] 33 -> 18  (-45.45%)
reflect.Value.Seq.func3 219 -> 178  (-18.72%)

reflect [cmd/compile]
reflect.Value.Seq2.func2 360 -> 285  (-20.83%)
reflect.Value.Seq.func4 281 -> 239  (-14.95%)
reflect.Value.Seq.func2 271 -> 230  (-15.13%)
reflect.Value.Seq.func1.1 242 -> 165  (-31.82%)
reflect.Value.Seq2.func1.1 281 -> 198  (-29.54%)
reflect.Value.Seq2.func3 309 -> 245  (-20.71%)
reflect.Value.Seq.func3 219 -> 178  (-18.72%)
reflect.TypeFor[go.shape.uint64] 33 -> 18  (-45.45%)
reflect.Value.Seq2.func4 399 -> 284  (-28.82%)

fmt
fmt.(*pp).fmtBytes 1723 -> 1691  (-1.86%)

database/sql/driver
reflect.TypeFor[go.shape.interface 33 -> 18  (-45.45%)
database/sql/driver.init 72 -> 57  (-20.83%)

Change-Id: I9eb750cf0b7ebf532589f939431feb0a899e42ff
Reviewed-on: https://go-review.googlesource.com/c/go/+/701301
Reviewed-by: Mark Freeman <markfreeman@google.com>
Reviewed-by: Keith Randall <khr@google.com>
Auto-Submit: Michael Pratt <mpratt@google.com>
Reviewed-by: Keith Randall <khr@golang.org>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>

runtime: fix syscall9 on darwin/arm64

The aarch64 ABI says that only the first 8 arguments should be
passed as registers, subsequent  arguments should be put on
the stack.

Syscall9 is not putting the 9th argument on the stack, and it should.

The standard library hasn't hit this issue because it uses Syscall9
for functions that only require 7 or 8 parameters.

Change-Id: I1fafca5b16f977ea856e3f08b4ff3d0a2a7a4dfe
Reviewed-on: https://go-review.googlesource.com/c/go/+/702297
Reviewed-by: Michael Pratt <mpratt@google.com>
Reviewed-by: Cherry Mui <cherryyz@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>

internal/goexperiment: add a sizespecializedmalloc goexperiment setting

Change-Id: I6a6a696497f2a0b0d403bbb11d7502f62edec78b
Reviewed-on: https://go-review.googlesource.com/c/go/+/696535
Auto-Submit: Michael Matloob <matloob@google.com>
Reviewed-by: Michael Knyszek <mknyszek@google.com>
Reviewed-by: Michael Matloob <matloob@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>

cmd/link: support MSVC clang

Currently on Windows, for cgo, we support MinGW-based C toolchain,
that is, with a -windows-gnu target. This CL makes it work with
clang with a -windows-msvc target. The LLVM toolchain bundled in
MSVC (https://learn.microsoft.com/en-us/cpp/build/clang-support-msbuild)
is such an example.

Currently it is expecting lld-link as the C linker, which is also
bundled in MSVC, can be requested with -fuse-ld=lld, but is not
the default.

This is the first step, which makes it generate a working cgo
binary. There are still more work to do, e.g. there are some
linker warnings, and the binary doesn't have symbol table.
all.bat doesn't pass with this setting.

Change-Id: I54d33f7dd5f5eeeafa0735cd52f4127fe4865636
Reviewed-on: https://go-review.googlesource.com/c/go/+/703055
Reviewed-by: Than McIntosh <thanm@golang.org>
Reviewed-by: Florian Zenker <floriank@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>

internal/testenv: update Builder docs with LUCI builder names

Also include some advice to avoid using this when possible. It makes
tests more prone to breaking under infrastructure changes.

Change-Id: Ifb2848742347eb0c937547dd888ab2cad3343f2d
Reviewed-on: https://go-review.googlesource.com/c/go/+/703115
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Dmitri Shuralyov <dmitshur@golang.org>
Reviewed-by: Dmitri Shuralyov <dmitshur@google.com>

net,internal/poll: skip TestAllocs when race is enabled on Windows

The Windows implementation of several network protocols make use of
sync.Pool, which randomly drops cached items when race is enabled.

While here, zero out the control buffer to allow it to be garbage
collected.

Fixes #75341

Change-Id: Ie20e21adef2edc02ca7b4a78012dd5f3a9f03bee
Reviewed-on: https://go-review.googlesource.com/c/go/+/703195
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Dmitri Shuralyov <dmitshur@google.com>
Reviewed-by: Cherry Mui <cherryyz@google.com>

cmd/asm, cmd/internal/obj: add riscv64 generic CSR ops

Support is added for the generic RISC-V CSR operations; CSRRC, CSRRCI,
CSRRS, CSRRSI, CSRRW, CSRRWI.  These instructions require special
handling as their second operand is a symbolic CSR register name and
not an immediate value or a register.  CSR names are implemented as
special operands.

RISC-V CSRs are not currently saved and restored when a go routine is
asynchronously pre-empted so it is only safe to use these instructions
in hand written assembler.  Note that CSRRS was already partially
supported by the assembler so this restriction predates this commit.
We mention it here as this commit makes CSRRS much easier to use.

Change-Id: I9ff8d804328b418a879d463e7d9cc31f489c7a00
Reviewed-on: https://go-review.googlesource.com/c/go/+/630519
Reviewed-by: Junyang Shao <shaojunyang@google.com>
Reviewed-by: Joel Sing <joel@sing.id.au>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Meng Zhuo <mengzhuo1203@gmail.com>
Reviewed-by: Michael Pratt <mpratt@google.com>

runtime: move mksizeclasses.go to runtime/_mkmalloc

This will allow us to share code with the specialized malloc code
generator.

Change-Id: I6a6a696450a5039a957811fb06228122d494ddce
Reviewed-on: https://go-review.googlesource.com/c/go/+/700495
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Michael Knyszek <mknyszek@google.com>
Reviewed-by: Michael Matloob <matloob@google.com>

cmd/go: run cgo and cgo compiles in their own actions

This change splits package builds further into even more actions.

Between the cache action (and cover action, if present) and the actual
build action we insert three layers of actions:

              Check Cache Action
              |      |
              |      V
              | Run Cover Action (if cover requested)
              |      |
              |      V
 New Layer 1  | Run Cgo Action (if package is built with cgo)
              |      | | | (multiple arrows representing fan-out to
              |      V V V  multiple actions)
 New Layer 2  | gcc Compile Action for each C/C++/S/F/M file (if cgo)
              |      |/    (arrow represeting fan-in to single action)
              |      V
 New Layer 3  | Cgo collect action (if cgo)
              |      |
              \      V
                Build action

The first run cgo action takes the input source files and runs cgo on
them to produce the cgo-processed go files, which are given to the
compiler, and to produce additional C files to compile and headers to
use in the following compilations. The action also takes care of running
SWIG before running cgo if there are swig files. This will produce
additional cgo sources that are inputs to cgo.

The run cgo action action fans out to multiple actions to do each of the
C/C++/Obj-C/assembly/Fortran compilations for the non-Go files in the
package, as well as those produced as outputs by the cgo command
invocation.

These actions then join into a single noop "collect" action which
primarily exists so that we don't pollute the build action's
dependencies with a bunch of non-go compile actions. (The build action
expects its dependencies to mostly be other build actions).

All of this work in the new actions was previously being done in the
build action itself. There's still a remnant of the original cgo logic
left in the build action to run the cgo command with -dynimport to
produce a go file to be built with the rest of the package, and do some
checks.

Most of this CL consists of moving code around. Just like the previous
CL breaking out the coverage logic into a separate action, we don't
cache the outputs of the cgo actions, and just treat all the actions
used to build a single package as one cacheable unit. This makes things
a bit simpler. If we decide in a future CL to cache the outputs
separately, we could remove the dependency on the cover action on the
check cache action (which in turn depends on all the package's
dependencies) and could start non-go compilation pretty much as early as
we want in the build.

The 'cgoAction' function in action.go takes care of creating the layers
of cgo action dependencies, which are inserted as dependencies of the
build action. It's mostly straightforward, except for the fact that we
need to tell each non-go compile action which non-go file to compile, so
we need to compute the names of the generated files. (Alternatively we
could give each action a number and have it build the nth file produced
by the run cgo action, but that seems even more complicated). The actors
produced to run the action logic are pretty light wrappers around the
execution logic in exec.go.

In the 'build' function in exec.go, most of the new code mainly checks
for the information from the cgo actions to use instead of running it,
and then passes it to the processCgoOutputs function. The only other
weird thing in the build functian is that we we call the logic to
compute the nonGoOverlay separately just for the C files that are being
built with gccgo. We compute the overlay for the non-go files used in a
cgo build in the run cgo action.

The 'cgo' function that previously ran the cgo logic for the build has
now been split into three: the first half, which runs cgo is now in the
runCgo function, the center part, which compiles the non-go files is now
partly in creating the invididual non-go compile actions, as well as the
cgoCompileActor's Act function. And the final part, which runs
cgo -dynimport is now in processCgoOutputs. These parts communicate with
each other through the providers that are set on the cgo actions.

One further improvement we can make to this change in the future is to
compile the dynimport file separately from the build action: its output
is only needed by the linker. This would remove any dependencies from
dependent packages' build actions on the cgo compile actions, allowing
more flexibility for scheduling actions.

Fixes #9887

Change-Id: Ie3c70bbf985148ba73094cddfc78c39dc6faad6e
Reviewed-on: https://go-review.googlesource.com/c/go/+/694475
Reviewed-by: Michael Pratt <mpratt@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Michael Matloob <matloob@google.com>

testing: exit B.Loop early upon saturation

There's a cap of 1 billion benchmark iterations because more than that
is usually not going to give more useful data. Unfortunately, the
existing implementation neglected to check whether the 1e9 cap had
already been exceeded when it adjusted the number of iterations in the
B.Loop slow path (stopOrScaleBLoop), since it's only when that cap is hit
that it needed to terminate early.

As a result, for _very_ cheap benchmarks (e.g. testing assembly
implementations with just a few instructions), the B.Loop would stop
incrementing the number of iterations, but wouldn't terminate early,
making it re-enter the slow-path _every_ iteration until the benchmark
time was exhausted.

This wasn't normally visible with the default -benchtime 2s, but when
raised to 5s, it would cause benchmarks that took <5ns/op to be reported
as exactly 5ns/op. (which looks a bit suspicious)

Notably, one can use -count for larger groupings to compute statistics.
golang.org/x/perf/cmd/benchstat is valuable for coalescing larger
run-counts from -count into more useful statistics.

Add a test which allows for fewer iterations on slow/contended
platforms but guards against reintroducing a bug of this nature.

Fixes #75210

Change-Id: Ie7f0b2e6c737b064448434f3ed565bfef8c4f020
Reviewed-on: https://go-review.googlesource.com/c/go/+/700275
Reviewed-by: Junyang Shao <shaojunyang@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Damien Neil <dneil@google.com>
Reviewed-by: Sean Liao <sean@liao.dev>
Auto-Submit: Sean Liao <sean@liao.dev>

cmd/go/internal/work: remove deps[1]="fmt" vet hack

The Builder.vet operation has not needed an artificial
dependency on the fmt package since CL 176439 in 2019.
Remove it now.

Change-Id: I398a6d2d57175c12843520d9f19ffd023e676123
Reviewed-on: https://go-review.googlesource.com/c/go/+/702856
Auto-Submit: Alan Donovan <adonovan@google.com>
Reviewed-by: Michael Matloob <matloob@golang.org>
Reviewed-by: Michael Matloob <matloob@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>

cmd/link: allow one to specify the data section in the internal linker

Fixes #74945

Change-Id: Ia73a8dcdf707222e822522daaa7f31a38b1c31e6
GitHub-Last-Rev: da1526ad8cebd5cfa2f979d49d86f3424d192ce0
GitHub-Pull-Request: golang/go#75117
Reviewed-on: https://go-review.googlesource.com/c/go/+/698355
Reviewed-by: Mark Freeman <markfreeman@google.com>
Reviewed-by: Cherry Mui <cherryyz@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>

encoding/gob: make use of reflect.TypeAssert

Use "reflect.TypeAssert" to simplify the code.

There are also some performance gains:

goarch: arm64
pkg: encoding/gob
cpu: Apple M4
                         │     old     │                 new                 │
                         │   sec/op    │   sec/op     vs base                │
EncodeComplex128Slice-10   1.048µ ± 3%   1.048µ ± 1%        ~ (p=0.986 n=10)
EncodeFloat64Slice-10      481.5n ± 0%   538.1n ± 0%  +11.75% (p=0.000 n=10)
EncodeInt32Slice-10        560.0n ± 1%   562.2n ± 1%        ~ (p=0.239 n=10)
EncodeStringSlice-10       713.1n ± 2%   690.1n ± 1%   -3.24% (p=0.000 n=10)
EncodeInterfaceSlice-10    16.10µ ± 3%   16.89µ ± 3%   +4.94% (p=0.004 n=10)
DecodeComplex128Slice-10   5.507µ ± 2%   5.488µ ± 2%        ~ (p=0.617 n=10)
DecodeFloat64Slice-10      3.359µ ± 1%   3.365µ ± 1%        ~ (p=0.403 n=10)
DecodeInt32Slice-10        3.296µ ± 1%   3.290µ ± 2%        ~ (p=0.926 n=10)
DecodeStringSlice-10       8.397µ ± 2%   8.459µ ± 1%        ~ (p=0.796 n=10)
DecodeStringsSlice-10      18.47µ ± 1%   11.14µ ± 1%  -39.69% (p=0.000 n=10)
DecodeBytesSlice-10        5.038µ ± 1%   5.039µ ± 1%        ~ (p=0.956 n=10)
DecodeInterfaceSlice-10    40.14µ ± 1%   40.60µ ± 1%   +1.16% (p=0.001 n=10)
DecodeMap-10               43.43µ ± 1%   44.09µ ± 1%   +1.51% (p=0.000 n=10)
geomean                    4.451µ        4.335µ        -2.62%

                         │      old       │                  new                   │
                         │      B/op      │     B/op      vs base                  │
EncodeComplex128Slice-10    1.0000 ±  ?      0.5000 ±  ?        ~ (p=0.350 n=10)
EncodeFloat64Slice-10        0.000 ± 0%       0.000 ± 0%        ~ (p=1.000 n=10) ¹
EncodeInt32Slice-10          0.000 ± 0%       0.000 ± 0%        ~ (p=1.000 n=10) ¹
EncodeStringSlice-10         0.000 ± 0%       0.000 ± 0%        ~ (p=1.000 n=10) ¹
EncodeInterfaceSlice-10      19.00 ± 5%       20.00 ± 5%   +5.26% (p=0.002 n=10)
DecodeComplex128Slice-10   24.55Ki ± 0%     24.53Ki ± 0%   -0.10% (p=0.000 n=10)
DecodeFloat64Slice-10      10.56Ki ± 0%     10.54Ki ± 0%   -0.22% (p=0.000 n=10)
DecodeInt32Slice-10        9.539Ki ± 0%     9.516Ki ± 0%   -0.25% (p=0.000 n=10)
DecodeStringSlice-10       38.18Ki ± 0%     38.16Ki ± 0%   -0.06% (p=0.000 n=10)
DecodeStringsSlice-10      63.96Ki ± 0%     40.51Ki ± 0%  -36.65% (p=0.000 n=10)
DecodeBytesSlice-10        22.58Ki ± 0%     22.58Ki ± 0%        ~ (p=1.000 n=10)
DecodeInterfaceSlice-10    80.47Ki ± 0%     80.47Ki ± 0%        ~ (p=1.000 n=10)
DecodeMap-10               48.81Ki ± 0%     48.81Ki ± 0%        ~ (p=1.000 n=10) ¹
geomean                                 ²                  -8.15%                ²
¹ all samples are equal
² summaries must be >0 to compute geomean

                         │      old      │                  new                  │
                         │   allocs/op   │  allocs/op   vs base                  │
EncodeComplex128Slice-10    0.000 ± 0%      0.000 ± 0%        ~ (p=1.000 n=10) ¹
EncodeFloat64Slice-10       0.000 ± 0%      0.000 ± 0%        ~ (p=1.000 n=10) ¹
EncodeInt32Slice-10         0.000 ± 0%      0.000 ± 0%        ~ (p=1.000 n=10) ¹
EncodeStringSlice-10        0.000 ± 0%      0.000 ± 0%        ~ (p=1.000 n=10) ¹
EncodeInterfaceSlice-10     0.000 ± 0%      0.000 ± 0%        ~ (p=1.000 n=10) ¹
DecodeComplex128Slice-10    149.0 ± 0%      148.0 ± 0%   -0.67% (p=0.000 n=10)
DecodeFloat64Slice-10       150.0 ± 0%      149.0 ± 0%   -0.67% (p=0.000 n=10)
DecodeInt32Slice-10         149.0 ± 0%      148.0 ± 0%   -0.67% (p=0.000 n=10)
DecodeStringSlice-10       1.150k ± 0%     1.149k ± 0%   -0.09% (p=0.000 n=10)
DecodeStringsSlice-10      2.158k ± 0%     1.158k ± 0%  -46.34% (p=0.000 n=10)
DecodeBytesSlice-10         149.0 ± 0%      149.0 ± 0%        ~ (p=1.000 n=10) ¹
DecodeInterfaceSlice-10    3.158k ± 0%     3.158k ± 0%        ~ (p=1.000 n=10) ¹
DecodeMap-10                160.0 ± 0%      160.0 ± 0%        ~ (p=1.000 n=10) ¹
geomean                                ²                 -4.83%                ²
¹ all samples are equal
² summaries must be >0 to compute geomean

Updates #62121

Change-Id: I1d3534e5001ca185bc8ba5a7ed4ddbc00f85a17e
GitHub-Last-Rev: c0209f8c50e14477069592bfeb161011e7f9e80a
GitHub-Pull-Request: golang/go#75409
Reviewed-on: https://go-review.googlesource.com/c/go/+/702735
Reviewed-by: Mark Freeman <markfreeman@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Auto-Submit: Michael Pratt <mpratt@google.com>
Reviewed-by: Daniel Martí <mvdan@mvdan.cc>
Reviewed-by: Michael Pratt <mpratt@google.com>

archive/tar: fix typo in benchmark name

This was accidentally introduced in CL 662835.

Change-Id: I5c7ac67337e33e82037414377912b57d2a45be91
Reviewed-on: https://go-review.googlesource.com/c/go/+/702275
Reviewed-by: Dmitri Shuralyov <dmitshur@golang.org>
Reviewed-by: Dmitri Shuralyov <dmitshur@google.com>
Auto-Submit: Dmitri Shuralyov <dmitshur@golang.org>
Reviewed-by: Mark Freeman <markfreeman@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>

syscall: actually remove unreachable code

CL 702415 claimed to remove unreachable code, but in reality merely hid
it from vet's unreachable pass. Since the unreachable code isn't serving
an active role in the test, do remove it to simplify code.

Change-Id: I5905b8b566e4ca013bdd1202d1492e3eae6a5ede
Reviewed-on: https://go-review.googlesource.com/c/go/+/702575
Reviewed-by: Ian Lance Taylor <iant@golang.org>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Mark Freeman <markfreeman@google.com>
Auto-Submit: Dmitri Shuralyov <dmitshur@golang.org>
Reviewed-by: Dmitri Shuralyov <dmitshur@google.com>

crypto/rsa: don't test CL 687836 against v1.0.0 FIPS 140-3 module

Fixes #75343 (again)
Updates #74115

Change-Id: I6a6a696431d12e45ec9e302e63ed18990c5ee9d9
Reviewed-on: https://go-review.googlesource.com/c/go/+/702615
Auto-Submit: Filippo Valsorda <filippo@golang.org>
Reviewed-by: Michael Pratt <mpratt@google.com>
Reviewed-by: Daniel McCarney <daniel@binaryparadox.net>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Mark Freeman <markfreeman@google.com>

debug/macho: filter non-external symbols when reading imported symbols without LC_DYSYMTAB

File.ImportedSymbols will return symbols with a type that has one of the
N_STAB (0xe0) bits set and no section. That's not the expected behavior,
as those symbols might not be external.

We should expand the type check to also account for the N_EXT bit.
The section check is not necessary, as N_EXT symbols never have it set.

I have found this issue in the wild by running "go tool cgo -dynimport",
but unfortuantely I couldn't get a minimal C code that generates
N_STAB symbols without section, so this CL doesn't add any new test.

Change-Id: Ib0093ff66b50c7cc2f39d83936314fc293236917
Reviewed-on: https://go-review.googlesource.com/c/go/+/702296
Reviewed-by: Cherry Mui <cherryyz@google.com>
Reviewed-by: David Chase <drchase@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>

internal/runtime/maps: remove redundant package docs

This CL removes redundant package docs comments under different files
as all of them will be shown in pkgsite.

As of now, package docs 'Package maps implements Go's builtin map type'
appears 3 times in pkgsite.

See: https://pkg.go.dev/internal/runtime/maps

Change-Id: I2f0696dd4f860afea5cc346532bca59135bec798
Reviewed-on: https://go-review.googlesource.com/c/go/+/702135
Reviewed-by: Michael Pratt <mpratt@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Mark Freeman <markfreeman@google.com>

runtime/internal/maps: only conditionally clear groups when sparse

We only want to do the work of clearing slots
if they are full. But we also don't want to do too
much work to figure out whether a slot is full or not,
especially if clearing a slot is cheap.
 1) We decide group-by-group instead of slot-by-slot.
    If any slot in a group is full, we zero the whole group.
 2) If groups are unlikely to be empty, don't bother
    testing for it.
 3) If groups are 50%/50% likely to be empty, also don't
    bother testing, as it confuses the branch predictor. See #75097.
 4) But if a group is really large, do the test anyway, as
    clearing is expensive.

Fixes #75097

Change-Id: I9191865dd3e0fe887751cffe6082ac27d8d8439c
Reviewed-on: https://go-review.googlesource.com/c/go/+/697876
Reviewed-by: Keith Randall <khr@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Auto-Submit: Keith Randall <khr@google.com>
Reviewed-by: Michael Pratt <mpratt@google.com>
Reviewed-by: Youlin Feng <fengyoulin@live.com>

internal/runtime/maps: speed up Clear

We don't need to know the actual full slots, just whether there
are any or not.

The test for any full slots is simpler on amd64. We don't have to
use PMOVMSKB and do the intreg->floatreg transfer.

Fixes #75097

Change-Id: Iace1c100618d7fc2ac5ddd5fe9e8fe5c9595243f
Reviewed-on: https://go-review.googlesource.com/c/go/+/697875
Reviewed-by: Youlin Feng <fengyoulin@live.com>
Auto-Submit: Keith Randall <khr@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Michael Pratt <mpratt@google.com>
Reviewed-by: Keith Randall <khr@google.com>

cmd: delete some more windows/arm remnants

Updates #71671

Amends CL 699176, CL 698415, CL 698036, CL 648795.

Change-Id: I69adff73d1c9631e07df63fc84c80ec0204d49d5
Reviewed-on: https://go-review.googlesource.com/c/go/+/700835
Reviewed-by: Michael Pratt <mpratt@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Dmitri Shuralyov <dmitshur@google.com>
Reviewed-by: Dmitri Shuralyov <dmitshur@golang.org>

runtime: don't artificially limit TestReadMetricsSched

TestReadMetricsSched/running can take some time to enter in steady state
on busy systems. We currently only allow 1 second for that, we should
let it run unlimitedly until success or the test time's out.

Fixes #75049

Change-Id: I452059e1837caf12a2d2d9cae1f70a0ef2d4f518
Reviewed-on: https://go-review.googlesource.com/c/go/+/702295
Auto-Submit: Damien Neil <dneil@google.com>
Reviewed-by: Damien Neil <dneil@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Michael Knyszek <mknyszek@google.com>

cmd/compile: when CSEing two values, prefer the statement marked one

Fixes #75249

Change-Id: Ifd61bf5341f23ce2c9735e607e00d987489caacf
Reviewed-on: https://go-review.googlesource.com/c/go/+/701295
Reviewed-by: Keith Randall <khr@google.com>
Reviewed-by: David Chase <drchase@google.com>
Reviewed-by: Alexander Musman <alexander.musman@gmail.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>

types2: better documentation for resolve()

Change-Id: Iece109dfbdc98d436b845148612f4943598697fe
Reviewed-on: https://go-review.googlesource.com/c/go/+/697697
Reviewed-by: Robert Findley <rfindley@google.com>
Commit-Queue: Mark Freeman <markfreeman@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>

internal/trace/raw: use strings.Cut instead of strings.SplitN 2

Replace strings.SplitN with strings.Cut for better performance and readability.

Change-Id: Ia245db62d8c2d1686887cb455f492db15606b57a
GitHub-Last-Rev: e00e164688f79d85d34fdf0d4ef126387ec6c0a0
GitHub-Pull-Request: golang/go#75257
Reviewed-on: https://go-review.googlesource.com/c/go/+/700915
Reviewed-by: Michael Knyszek <mknyszek@google.com>
Auto-Submit: Michael Pratt <mpratt@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Michael Pratt <mpratt@google.com>

Revert "cmd/compile: improve stp merging for non-sequent cases"

This reverts commit 4c63d798cb947a3cdd5a5b68f254a73d83eb288f.

Reason for revert: Causes miscompilations. See issue 75365.

Change-Id: Icd1fcfeb23d2ec524b16eb556030f43875e1c90d
Reviewed-on: https://go-review.googlesource.com/c/go/+/702455
Reviewed-by: Keith Randall <khr@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Cherry Mui <cherryyz@google.com>
Auto-Submit: Keith Randall <khr@golang.org>
Reviewed-by: Mark Freeman <markfreeman@google.com>

go/token, syscall: annotate if blocks that defeat vet's unreachable pass

Since putting code in an "if true" block is unusual, make it easier
for readers to understand the purpose of doing this.

For #73998.

Change-Id: I3fd8d65130211c7c01d424366a3c662482d80add
Reviewed-on: https://go-review.googlesource.com/c/go/+/702416
Reviewed-by: Alan Donovan <adonovan@google.com>
Auto-Submit: Dmitri Shuralyov <dmitshur@golang.org>
Reviewed-by: Dmitri Shuralyov <dmitshur@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>

syscall: remove unreachable code

Reported by go vet:

$ go vet syscall
# syscall_test
# [syscall_test]
syscall/env_unix_test.go:100:4: unreachable code

The TestVetStdlib test in golang.org/x/tools/go/analysis/unitchecker
also ran into this.

Fixes #73998.

Change-Id: I7f2842a42835a38163433a09a3311be9c30f8a14
Cq-Include-Trybots: luci.golang.try:x_tools-gotip-linux-amd64-longtest
Reviewed-on: https://go-review.googlesource.com/c/go/+/702415
Auto-Submit: Dmitri Shuralyov <dmitshur@golang.org>
Reviewed-by: Dmitri Shuralyov <dmitshur@google.com>
Reviewed-by: Alan Donovan <adonovan@google.com>
TryBot-Bypass: Dmitri Shuralyov <dmitshur@golang.org>

Revert "crypto/internal/fips140: update frozen module version to "v1.0.0""

This reverts CL 701518. This should land just before CL 701520 to avoid
breaking the builders for long.

Fixes #75343

Change-Id: If1ae1fe933fe443ca9776888967d80875b02f41a
Reviewed-on: https://go-review.googlesource.com/c/go/+/702255
Reviewed-by: Keith Randall <khr@google.com>
Reviewed-by: Keith Randall <khr@golang.org>
Reviewed-by: Dmitri Shuralyov <dmitshur@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>

encoding/json/v2: document context annotation with SemanticError

When the json package calls
Marshaler, MarshalerTo, Unmarshaler, or UnmarshalerFrom methods
and a SemanticError is returned, it will automatically
annotate the error with context.

Document this behavior.

Change-Id: Id8e775a7c1c2a6ffc29ea518913591915e8aff87
Reviewed-on: https://go-review.googlesource.com/c/go/+/701956
Reviewed-by: Damien Neil <dneil@google.com>
Reviewed-by: Dmitri Shuralyov <dmitshur@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>

runtime: when using cgo on 386, call C sigaction function

On 386 the C sigaction function assumes that the caller does not set
the SA_RESTORER flag. It does not copy the C sa_restorer field to
the kernel sa_restorer field. The effect is that the kernel sees
the SA_RESTORER flag but a NULL sa_restorer field, and the program
crashes when returning from a signal handler.

On the other hand, the C sigaction function will return the SA_RESTORER
flag and the sa_restorer field stored in the kernel.

This means that if the Go runtime installs a signal handler,
with SA_RESTORER as is required when calling the kernel,
and the Go program calls C code that calls the C sigaction function
to query the current signal handler, that C code will get a result
that it can't pass back to sigaction.

This CL fixes the problem by using the C sigaction function
for 386 programs that use cgo. This reuses the functionality
used on amd64 and other GOARCHs to support the race detector.

See #75253, or runtime/testdata/testprogcgo/eintr.go, for sample
code that used to fail on 386. No new test case is required,
we just remove the skip we used to have for eintr.go.

Fixes #75253

Change-Id: I803059b1fb9e09e9fbb43f68eccb6a59a92c2991
Reviewed-on: https://go-review.googlesource.com/c/go/+/701375
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Cherry Mui <cherryyz@google.com>
Reviewed-by: Dmitri Shuralyov <dmitshur@google.com>
Auto-Submit: Ian Lance Taylor <iant@golang.org>

runtime: remove duff support for riscv64

Change-Id: I987d9f49fbd2650eef4224f72271bf752c54d39c
Reviewed-on: https://go-review.googlesource.com/c/go/+/700538
Reviewed-by: Mark Freeman <markfreeman@google.com>
Reviewed-by: Keith Randall <khr@golang.org>
Reviewed-by: Keith Randall <khr@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Mark Ryan <markdryan@rivosinc.com>

cmd/compile: use generated loops instead of DUFFCOPY on riscv64

MemmoveKnownSize112-4            632.1Mi ± 1%   1288.5Mi ± 0%  +103.85% (p=0.000 n=10)
MemmoveKnownSize128-4            636.1Mi ± 0%   1280.9Mi ± 1%  +101.36% (p=0.000 n=10)
MemmoveKnownSize192-4            645.3Mi ± 0%   1306.9Mi ± 1%  +102.53% (p=0.000 n=10)
MemmoveKnownSize248-4            650.2Mi ± 2%   1312.5Mi ± 1%  +101.87% (p=0.000 n=10)
MemmoveKnownSize256-4            650.7Mi ± 0%   1303.6Mi ± 1%  +100.33% (p=0.000 n=10)
MemmoveKnownSize512-4            658.2Mi ± 1%   1293.9Mi ± 0%   +96.60% (p=0.000 n=10)
MemmoveKnownSize1024-4           662.1Mi ± 0%   1312.6Mi ± 0%   +98.26% (p=0.000 n=10)

Change-Id: I43681ca029880025558b33ddc4295da3947c9b28
Reviewed-on: https://go-review.googlesource.com/c/go/+/700537
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Keith Randall <khr@google.com>
Reviewed-by: Keith Randall <khr@golang.org>
Reviewed-by: Mark Freeman <markfreeman@google.com>

cmd/compile: use generated loops instead of DUFFZERO on riscv64

MemclrKnownSize112-4          5.602Gi ± 0%    5.601Gi ± 0%         ~ (p=0.363 n=10)
MemclrKnownSize128-4          6.933Gi ± 1%    6.545Gi ± 1%    -5.59% (p=0.000 n=10)
MemclrKnownSize192-4          8.055Gi ± 1%    7.804Gi ± 0%    -3.12% (p=0.000 n=10)
MemclrKnownSize248-4          8.489Gi ± 0%    8.718Gi ± 0%    +2.69% (p=0.000 n=10)
MemclrKnownSize256-4          8.762Gi ± 0%    8.763Gi ± 0%         ~ (p=0.494 n=10)
MemclrKnownSize512-4          9.514Gi ± 1%    9.514Gi ± 0%         ~ (p=0.529 n=10)
MemclrKnownSize1024-4         9.940Gi ± 0%    9.939Gi ± 1%         ~ (p=0.989 n=10)
ClearFat3-4                   1.300Gi ± 0%    1.301Gi ±  0%         ~ (p=0.447 n=10)
ClearFat4-4                   3.902Gi ± 0%    3.902Gi ±  0%         ~ (p=0.971 n=10)
ClearFat5-4                   665.8Mi ± 0%   1331.5Mi ±  0%  +100.01% (p=0.000 n=10)
ClearFat6-4                   665.8Mi ± 0%   1330.5Mi ±  0%   +99.82% (p=0.000 n=10)
ClearFat7-4                   490.7Mi ± 0%   1331.9Mi ±  0%  +171.45% (p=0.000 n=10)
ClearFat8-4                   5.201Gi ± 0%    5.202Gi ±  0%         ~ (p=0.123 n=10)
ClearFat9-4                   856.1Mi ± 0%   1331.6Mi ±  0%   +55.54% (p=0.000 n=10)
ClearFat10-4                  887.8Mi ± 0%   1331.9Mi ±  0%   +50.03% (p=0.000 n=10)
ClearFat11-4                  915.3Mi ± 0%   1331.1Mi ±  0%   +45.42% (p=0.000 n=10)
ClearFat12-4                  5.202Gi ± 0%    5.202Gi ±  0%         ~ (p=0.481 n=10)
ClearFat13-4                  961.5Mi ± 0%   1331.8Mi ±  0%   +38.50% (p=0.000 n=10)
ClearFat14-4                  981.0Mi ± 0%   1331.8Mi ±  0%   +35.76% (p=0.000 n=10)
ClearFat15-4                  951.3Mi ± 0%   1331.4Mi ±  0%   +39.96% (p=0.000 n=10)
ClearFat16-4                  1.600Gi ± 0%    5.202Gi ±  0%  +225.10% (p=0.000 n=10)
ClearFat18-4                  1.018Gi ± 0%    1.300Gi ±  0%   +27.77% (p=0.000 n=10)
ClearFat20-4                  2.601Gi ± 0%    4.938Gi ± 12%   +89.87% (p=0.000 n=10)
ClearFat24-4                  2.601Gi ± 0%    5.201Gi ±  0%   +99.96% (p=0.000 n=10)
ClearFat32-4                  1.982Gi ± 0%    5.203Gi ±  0%  +162.55% (p=0.000 n=10)
ClearFat40-4                  3.467Gi ± 0%    4.338Gi ±  0%   +25.11% (p=0.000 n=10)
ClearFat48-4                  3.671Gi ± 0%    5.201Gi ±  0%   +41.69% (p=0.000 n=10)
ClearFat56-4                  3.640Gi ± 0%    5.201Gi ±  0%   +42.88% (p=0.000 n=10)
ClearFat64-4                  2.250Gi ± 0%    5.202Gi ±  0%  +131.25% (p=0.000 n=10)
ClearFat72-4                  4.064Gi ± 0%    5.201Gi ±  0%   +27.97% (p=0.000 n=10)
ClearFat128-4                 4.496Gi ± 0%    5.203Gi ±  0%   +15.71% (p=0.000 n=10)
ClearFat256-4                 4.756Gi ± 0%    5.201Gi ±  0%    +9.36% (p=0.000 n=10)
ClearFat512-4                 2.512Gi ± 0%    5.201Gi ±  0%  +107.03% (p=0.000 n=10)
ClearFat1024-4                4.255Gi ± 0%    5.202Gi ±  0%   +22.26% (p=0.000 n=10)
ClearFat1032-4                4.260Gi ± 0%    5.201Gi ±  0%   +22.09% (p=0.000 n=10)
ClearFat1040-4                4.285Gi ± 1%    5.203Gi ±  0%   +21.41% (p=0.000 n=10)
geomean                       2.005Gi         3.020Gi         +50.58%

Change-Id: Iea1da734ff8eaf1b5a2822ae2bdb7f4fd9b65651
Reviewed-on: https://go-review.googlesource.com/c/go/+/699635
Reviewed-by: Mark Ryan <markdryan@rivosinc.com>
Reviewed-by: Keith Randall <khr@google.com>
Reviewed-by: Keith Randall <khr@golang.org>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Mark Freeman <markfreeman@google.com>

cmd/compile: simplify zerorange on riscv64

Drop large zeroing cases, part of removing duff support.

Change-Id: Ia2936f649901886f3eb1d7ba1f90e3bf40ea2dee
Reviewed-on: https://go-review.googlesource.com/c/go/+/697615
Reviewed-by: Keith Randall <khr@golang.org>
Reviewed-by: Julian Zhu <jz531210@gmail.com>
Reviewed-by: Mark Ryan <markdryan@rivosinc.com>
Reviewed-by: Keith Randall <khr@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Mark Freeman <markfreeman@google.com>
Reviewed-by: Joel Sing <joel@sing.id.au>

encoding/json: use reflect.TypeAssert

Updates #62121

Change-Id: Ic3c4fe84a5dacfd8270aba0d5dd59f83f0a9030f
Reviewed-on: https://go-review.googlesource.com/c/go/+/701955
Reviewed-by: Mark Freeman <markfreeman@google.com>
Reviewed-by: Damien Neil <dneil@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>

cmd/cgo: run gcc to get errors and debug info in parallel

This change kicks off the work to load the debug info when processing
each file, and then waits for all the files to be processed before
starting the single-goroutined part that processes them. The processing
is very order dependent so we won't try to make it concurrent. Though
in a later CL we can wait for only the relevant package to have been
processed concurrently before doing the single-goroutined processing for
it instead of waiting for all packages to be processed concurrently
before the single goroutine section.

We use a par.Queue to make sure we're not running too many gcc compiles
at the same time. The change to cmd/dist makes the par package available
to cgo.

Fixes #75167

Change-Id: I6a6a6964fb7f3a3684118b5ee66f1ad856b3ee59
Reviewed-on: https://go-review.googlesource.com/c/go/+/699020
Reviewed-by: Michael Matloob <matloob@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Reviewed-by: Michael Pratt <mpratt@google.com>

runtime: lock mheap_.speciallock when allocating synctest specials

Avoid racing use of mheap_.specialBubbleAlloc.

Fixes #75134

Change-Id: I0c9140c18d2bca1e1c3387cd81230f0e8c9ac23e
Reviewed-on: https://go-review.googlesource.com/c/go/+/699255
Reviewed-by: Michael Knyszek <mknyszek@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>

runtime: don't negate eventfd errno

The Linux syscall package does this for us.

Fixes #75337.

Change-Id: I6a6a636c9bb5fe25fdc6f80dc8b538ebed60d00b
Reviewed-on: https://go-review.googlesource.com/c/go/+/701796
Reviewed-by: Michael Knyszek <mknyszek@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Auto-Submit: Michael Pratt <mpratt@google.com>

syscall: optimise cgo clearenv

For programs with very large environments, calling unsetenv(3) for each
environment variable can be very expensive because of CGo overhead, but
clearenv(3) is much faster. The only thing we have to track is whether
GODEBUG is being unset by the operation, which can be done very quickly
without resorting to doing unsetenv(3) for every variable.

This change makes syscall.Clearenv() >98% faster when run in an
environment with as little as 100 environment variables. (Note that due
to golang/go#27217, it is necessary to modify BenchmarkClearenv to use
t.StopTimer() and -benchtime=100x in order to get these benchmark times
-- otherwise syscall.Setenv() time is included and the benchmarks give a
more pessimistic 50% performance improvement.)

    goos: linux
    goarch: amd64
    pkg: syscall
    cpu: AMD Ryzen 7 7840U w/ Radeon  780M Graphics
                      │      before      │                after                │
                      │      sec/op      │   sec/op     vs base                │
    Clearenv/100-16        22276.5n ± 5%   285.8n ± 3%  -98.72% (p=0.000 n=10)
    Clearenv/1000-16     1414104.0n ± 1%   783.1n ± 8%  -99.94% (p=0.000 n=10)
    Clearenv/10000-16   143827.554µ ± 1%   7.591µ ± 5%  -99.99% (p=0.000 n=10)
    geomean                  1.655m        1.193µ       -99.93%

The above benchmarks are CGo builds, which require CGo overhead for
every setenv(2). If you run the same benchmarks for a non-CGo package
(i.e., outside of the "syscall" package), you get slightly more modest
performance improvements:

    goos: linux
    goarch: amd64
    pkg: clearenv_nocgo
    cpu: AMD Ryzen 7 7840U w/ Radeon  780M Graphics
                      │     before     │                after                 │
                      │     sec/op     │    sec/op     vs base                │
    Clearenv/100-16       1106.0n ± 3%   230.7n ±  8%  -79.14% (p=0.000 n=10)
    Clearenv/1000-16     11222.0n ± 1%   305.4n ±  6%  -97.28% (p=0.000 n=10)
    Clearenv/10000-16   195676.5n ± 6%   759.9n ± 10%  -99.61% (p=0.000 n=10)
    geomean                13.44µ        376.9n        -97.20%

(As above, this requires modifying the benchmarks to use t.StopTimer()
and -benchtime=100x.)

Change-Id: I53b96a75f189e91affbde423c907888b7e0fafcd
GitHub-Last-Rev: f8d7a8140d8490189d726eb380522dccacc5f176
GitHub-Pull-Request: golang/go#70672
Reviewed-on: https://go-review.googlesource.com/c/go/+/633515
Reviewed-by: Mark Freeman <markfreeman@google.com>
Reviewed-by: Keith Randall <khr@golang.org>
Reviewed-by: Kirill Kolyshkin <kolyshkin@gmail.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Keith Randall <khr@google.com>

crypto/rsa: check PrivateKey.D for consistency with Dp and Dq

This unfortunately nearly doubles the runtime of
NewPrivateKeyWithPrecomputation. It would be nice to find an alternative
way to check it.

fips140: off
goos: darwin
goarch: arm64
pkg: crypto/rsa
cpu: Apple M2
                            │ 6aeb841faf  │             62ec3e34f3              │
                            │   sec/op    │    sec/op     vs base               │
ParsePKCS8PrivateKey/2048-8   70.28µ ± 0%   116.16µ ± 0%  +65.28% (p=0.002 n=6)

Fixes #74115

Change-Id: I6a6a6964091817d9aee359cc48932167e55184b9
Reviewed-on: https://go-review.googlesource.com/c/go/+/687836
Auto-Submit: Filippo Valsorda <filippo@golang.org>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Daniel McCarney <daniel@binaryparadox.net>
Reviewed-by: Mark Freeman <markfreeman@google.com>
Reviewed-by: Michael Pratt <mpratt@google.com>

crypto/rsa: check for post-Precompute changes in Validate

Updates #74115

Change-Id: I6a6a6964be55cff5131d99235f621b4ff2a93d2b
Reviewed-on: https://go-review.googlesource.com/c/go/+/687835
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Mark Freeman <markfreeman@google.com>
Reviewed-by: Michael Pratt <mpratt@google.com>
Reviewed-by: Daniel McCarney <daniel@binaryparadox.net>
Auto-Submit: Filippo Valsorda <filippo@golang.org>

crypto/internal/fips140: update frozen module version to "v1.0.0"

We are re-sealing the .zip file anyway for another reason, might as well
take the opportunity to fix the "v1.0" mistake.

Change-Id: I6a6a69646b3188984c865031ff9393ccaaaa9479
Reviewed-on: https://go-review.googlesource.com/c/go/+/701518
Reviewed-by: Daniel McCarney <daniel@binaryparadox.net>
Auto-Submit: Filippo Valsorda <filippo@golang.org>
Reviewed-by: Mark Freeman <markfreeman@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Michael Pratt <mpratt@google.com>

crypto/ecdsa: deprecate direct use of big.Int fields in keys

Updates #63963

Change-Id: I6a6a69645e625cde1ac1c6abf698a5fd3d52b4cf
Reviewed-on: https://go-review.googlesource.com/c/go/+/701516
Auto-Submit: Filippo Valsorda <filippo@golang.org>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Daniel McCarney <daniel@binaryparadox.net>
Reviewed-by: Mark Freeman <markfreeman@google.com>
Reviewed-by: Michael Pratt <mpratt@google.com>

cmd/compile/internal/inline: ignore superfluous slicing

When slicing, ignore expressions which could be elided, as in slicing
starting at 0 or ending at len(v).

Fixes #75278

Change-Id: I9c18e29c3d4da9bef89bd25bb261d3cb60e66392
Reviewed-on: https://go-review.googlesource.com/c/go/+/701216
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Keith Randall <khr@google.com>
Reviewed-by: Keith Randall <khr@golang.org>
Auto-Submit: Keith Randall <khr@golang.org>
Reviewed-by: Mark Freeman <markfreeman@google.com>