cmd/compile/internal: remove incorrect riscv64 SLTI rule

The rule

(SLTI  [x] (ORI  [y] _)) && y >= 0 && int64(y) >= int64(x) => (MOVDconst [0])

is incorrect as it only generates correct code if the unknown value
being compared is >= 0. If the unknown value is < 0 the rule will
incorrectly produce a constant value of 0, whereas the code optimized
away by the rule would have produced a value of 1.

A new test that causes the faulty rule to generate incorrect code
is also added to ensure that the error does not return.

Change-Id: I69224e0776596f1b9538acf9dacf9009d305f966
Reviewed-on: https://go-review.googlesource.com/c/go/+/720220
Reviewed-by: Meng Zhuo <mengzhuo1203@gmail.com>
Reviewed-by: Junyang Shao <shaojunyang@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Keith Randall <khr@google.com>
Reviewed-by: Joel Sing <joel@sing.id.au>
Auto-Submit: Keith Randall <khr@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>

cmd/compile: fold negation into multiplication

goos: linux
goarch: riscv64
pkg: cmd/compile/internal/test
cpu: Spacemit(R) X60
        │ /root/mul.base.log │          /root/mul.new.log          │
        │       sec/op       │   sec/op     vs base                │
MulNeg           6.426µ ± 0%   4.501µ ± 0%  -29.96% (p=0.000 n=10)
Mul2Neg          9.000µ ± 0%   6.431µ ± 0%  -28.54% (p=0.000 n=10)
Mul2             1.263µ ± 0%   1.263µ ± 0%        ~ (p=1.000 n=10)
MulNeg2          1.577µ ± 0%   1.577µ ± 0%        ~ (p=0.211 n=10)
geomean          3.276µ        2.756µ       -15.89%

goos: linux
goarch: amd64
pkg: cmd/compile/internal/test
cpu: AMD EPYC 7532 32-Core Processor
        │ /root/base  │              /root/new              │
        │   sec/op    │   sec/op     vs base                │
MulNeg    691.9n ± 1%   319.4n ± 0%  -53.83% (p=0.000 n=10)
Mul2Neg   630.0n ± 0%   629.6n ± 0%   -0.07% (p=0.000 n=10)
Mul2      438.1n ± 0%   438.1n ± 0%        ~ (p=0.728 n=10)
MulNeg2   439.3n ± 0%   439.4n ± 0%        ~ (p=0.656 n=10)
geomean   538.2n        443.6n       -17.58%

Change-Id: Ice8e6c8d1e8e3009ba8a0b1b689205174e199019
Reviewed-on: https://go-review.googlesource.com/c/go/+/720180
Reviewed-by: abner chenc <chenguoqi@loongson.cn>
Reviewed-by: Keith Randall <khr@golang.org>
Reviewed-by: Junyang Shao <shaojunyang@google.com>
Reviewed-by: Joel Sing <joel@sing.id.au>
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@golang.org>

bytes: fix panic in bytes.Buffer.Peek

This change fixed the overlooked offset in bytes.Buffer.Peek.
Otherwise, it will either return wrong result or panic with
"runtime error: slice bounds out of range".

Change-Id: Ic42fd8a27fb9703c51430f298933b91cf0d45451
GitHub-Last-Rev: fb97ebc3b188959835706626f66898d6306c16fb
GitHub-Pull-Request: golang/go#76165
Reviewed-on: https://go-review.googlesource.com/c/go/+/717640
Reviewed-by: Michael Pratt <mpratt@google.com>
Auto-Submit: Michael Pratt <mpratt@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Junyang Shao <shaojunyang@google.com>

cmd/compile: optimize comparisons with single bit difference

Optimize comparisons with constants that only differ by 1 bit (i.e.
a power of 2). For example:

    x == 4 || x == 6 -> x|2 == 6
    x != 1 && x != 5 -> x|4 != 5

Change-Id: Ic61719e5118446d21cf15652d9da22f7d95b2a15
Reviewed-on: https://go-review.googlesource.com/c/go/+/719420
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Junyang Shao <shaojunyang@google.com>
Auto-Submit: Keith Randall <khr@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
Reviewed-by: Keith Randall <khr@google.com>

cmd/compile: remove unnecessary casts and types from riscv64 rules

This change shouldn't have any impact on codegen. It removes some
unnecessary int8 and int64 casts from the riscv64 rewrite rules.

It also removes explicit typing where the types already match:
`(OldOp <t>) => (NewOp <t>)` is the same as `(OldOp) => (NewOp)`.

Change-Id: Ic02b65da8f548c8b9ad9ccb6627a03b7bf6f050f
Reviewed-on: https://go-review.googlesource.com/c/go/+/719220
Reviewed-by: Junyang Shao <shaojunyang@google.com>
Auto-Submit: Keith Randall <khr@golang.org>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Keith Randall <khr@golang.org>
Reviewed-by: Keith Randall <khr@google.com>
Commit-Queue: Junyang Shao <shaojunyang@google.com>

go/types, types2: swap object.color for Checker.objPathIdx

The type checker assigns types to objects. An object can be in
1 of 3 states, which are mapped to colors. This is stored as a
field on the object.

 - white : type not known, awaiting processing
 - grey  : type pending, in processing
 - black : type known, done processing

With the addition of Checker.objPathIdx, which maps objects to
a path index, presence in the map could signal grey coloring.
White and black coloring can be signaled by presence of
object.typ (a simple nil check).

This change removes the object.color field and its associated
methods, replacing it with an equivalent use of Checker.objPathIdx.
Checker.objPathIdx is integrated into the existing push and pop
methods, while slightly simplifying their signatures.

Note that the concept of object coloring remains the same - we
are merely changing and simplifying the mechanism which represents
the colors.

Change-Id: I91fb5e9a59dcb34c08ffc5b4ebc3f20a400094b6
Reviewed-on: https://go-review.googlesource.com/c/go/+/715840
Reviewed-by: Robert Griesemer <gri@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Auto-Submit: Mark Freeman <markfreeman@google.com>

cmd/link/internal/ld: make runtime.buildVersion with experiments valid

Specifically if there are experiments but no nonstandard toolchain
suffix such as "-devel", the go version will not be valid according to
go/version.IsValid. To fix that, always put the X: part into the suffix,
resulting in, for example, go1.25.0-X:foo.

Fixes #75953

Change-Id: I6a6a696468f3ba9b82b6a410fb88831428e93b58
Reviewed-on: https://go-review.googlesource.com/c/go/+/719701
Reviewed-by: Michael Matloob <matloob@google.com>
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>

test: fix tests to work with sizespecializedmalloc turned off

Cq-Include-Trybots: luci.golang.try:gotip-linux-386-nosizespecializedmalloc,gotip-linux-amd64-nosizespecializedmalloc,gotip-linux-arm64-nosizespecializedmalloc
Change-Id: I6a6a696465004b939c989afc058c4c3e1fb7134f
Reviewed-on: https://go-review.googlesource.com/c/go/+/720401
Auto-Submit: Michael Matloob <matloob@golang.org>
Reviewed-by: Michael Knyszek <mknyszek@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Michael Matloob <matloob@google.com>

cmd/trace: annotation proc start/stop with thread and proc always

In the proc view, the thread ID is useful. In the thread view, the proc
ID is useful. Add both in both cases forever more.

Change-Id: I9cb7bd67a21ee17d865c25d73b2049b3da7aefbc
Reviewed-on: https://go-review.googlesource.com/c/go/+/720402
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Michael Pratt <mpratt@google.com>
Auto-Submit: Michael Knyszek <mknyszek@google.com>

net/http: remove unused isLitOrSingle and isNotToken

isLitOrSingle and isNotToken are private and unused.

Change-Id: I07718d4496e92d5f75ed74986e174a8aa1f70a88
GitHub-Last-Rev: 722c4dccd85dca5d28a52e95a4f9efbab2b11807
GitHub-Pull-Request: golang/go#76216
Reviewed-on: https://go-review.googlesource.com/c/go/+/718700
Reviewed-by: Damien Neil <dneil@google.com>
Auto-Submit: Sean Liao <sean@liao.dev>
Reviewed-by: Junyang Shao <shaojunyang@google.com>
Reviewed-by: Sean Liao <sean@liao.dev>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>

cmd/go: fix flaky TestScript/mod_get_direct

Use our local git server instead of cloud.google.com/go,
which may go down or otherwise reject our traffic.

I built and installed git 2.23.0 and confirmed that this
updated test still fails if CL 196961 is rolled back.
So the test is still accurate at detecting the problem.

Change-Id: I58011f6cc62af2f21fbbcc526ba5401f4186eeaf
Reviewed-on: https://go-review.googlesource.com/c/go/+/720322
Reviewed-by: Michael Matloob <matloob@golang.org>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Michael Matloob <matloob@google.com>

net/http/cgi: accept INCLUDED as protocol for server side includes

The existing protocol check for fcgi/cgi requests did not properly
account for Apache SSI (Server-Side Includes) SERVER_PROTOCOL value of
INCLUDED.

Added check for well-known INCLUDED value for proper implementation of
the CGI Spec as specified in RFC 3875 - section 4.1.16.

The SERVER_PROTOCOL section of the specification is outlined at
https://www.rfc-editor.org/rfc/rfc3875.html#section-4.1.16

Fixes #70416

Change-Id: I129e606147e16d1daefb49ed6c13a561a88ddeb6
Reviewed-on: https://go-review.googlesource.com/c/go/+/715680
Reviewed-by: Damien Neil <dneil@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Junyang Shao <shaojunyang@google.com>
Reviewed-by: Sean Liao <sean@liao.dev>
Auto-Submit: Sean Liao <sean@liao.dev>

go/types: minor simplification

Change-Id: Ia315cef2022197ec740023b67827bc810219b868
Reviewed-on: https://go-review.googlesource.com/c/go/+/715361
Reviewed-by: Robert Griesemer <gri@google.com>
Auto-Submit: Mark Freeman <markfreeman@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>

go/types: generate cycles.go

Change-Id: I09a31a1ccec082f5538736c48b211e4d567ee80c
Reviewed-on: https://go-review.googlesource.com/c/go/+/715420
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Robert Griesemer <gri@google.com>
Auto-Submit: Mark Freeman <markfreeman@google.com>

go/types, types2: rename definedType to declaredType and clarify docs

declaredType seems a better name for this function because it is reached
when processing a (Named or Alias) type declaration. In both cases, the
fromRHS field (rather than the underlying field) will be set.

Change-Id: Ibb1cc338e3b0632dc63f9cf2fd9d64cef6f1aaa5
Reviewed-on: https://go-review.googlesource.com/c/go/+/695955
Auto-Submit: Mark Freeman <markfreeman@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Robert Griesemer <gri@google.com>

go/types, types2: check for direct cycles as a separate phase

A direct cycle is the most basic form of cycle, where no type literal or
predeclared type is reached. It is formed by a series of only TypeNames.

To illustrate, type T T is a direct cycle, but type T [1]T and type T *T
are not. Likewise, the below is also a direct cycle:

  type A B
  type B C
  type C = A

Direct cycles are handled explicitly as part of resolveUnderlying, since
they are the only cycle which can prevent reaching an underlying type.
If we move this check to an earlier compiler phase, we can simplify
resolveUnderlying.

This is the first of (hopefully) several cycle kinds to be moved into a
preliminary phase, with the goal of simplifying the main type-checking
pass. For that reason, the bulk of the logic is placed in cycles.go.

CL based on an earlier version by Mark Freeman.

Change-Id: I3044c383278deb6acb8767c498d8cb68099ba8ef
Reviewed-on: https://go-review.googlesource.com/c/go/+/717343
Auto-Submit: Robert Griesemer <gri@google.com>
Reviewed-by: Mark Freeman <markfreeman@google.com>
Reviewed-by: Robert Griesemer <gri@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>

cmd/go/internal/modfetch: consolidate global vars

This commit conslidates global variables represeting the current
modfetch state into a single variable in preparation for injection via
the rsc.io/rf tool.

This commit is part of the larger effort to eliminate global module
loader state.

[git-generate]
cd src/cmd/go/internal/modfetch
rf '
  add State func NewState() *State {\
    s := new(State)\
    s.lookupCache = new(par.Cache[lookupCacheKey, Repo])\
    s.downloadCache = new(par.ErrCache[module.Version, string])\
    return s\
  }

  add State var ModuleFetchState *State = NewState()

  mv State.goSumFile State.GoSumFile
  add State:/^[[:space:]]*GoSumFile /-0 // path to go.sum; set by package modload
  ex {
    GoSumFile -> ModuleFetchState.GoSumFile
  }

  mv State.workspaceGoSumFiles State.WorkspaceGoSumFiles
  add State:/^[[:space:]]*WorkspaceGoSumFiles /-0 // path to module go.sums in workspace; set by package modload
  ex {
    WorkspaceGoSumFiles -> ModuleFetchState.WorkspaceGoSumFiles
  }

  add State:/^[[:space:]]*lookupCache /-0 // The Lookup cache is used cache the work done by Lookup.\
      // It is important that the global functions of this package that access it do not\
      // do so after they return.
  add State:/^[[:space:]]*downloadCache /-0 // The downloadCache is used to cache the operation of downloading a module to disk\
      // (if it'\\\''s not already downloaded) and getting the directory it was downloaded to.\
      // It is important that downloadCache must not be accessed by any of the exported\
      // functions of this package after they return, because it can be modified by the\
      // non-thread-safe SetState function.
  add State:/^[[:space:]]*downloadCache.*$/ // version → directory;
  ex {
    lookupCache -> ModuleFetchState.lookupCache
    downloadCache -> ModuleFetchState.downloadCache
  }
  rm 'lookupCache'
  rm 'downloadCache'
'

for dir in modload ; do
  cd ../${dir}
  rf '
    ex {
      import "cmd/go/internal/modfetch"
      modfetch.GoSumFile -> modfetch.ModuleFetchState.GoSumFile
      modfetch.WorkspaceGoSumFiles -> modfetch.ModuleFetchState.WorkspaceGoSumFiles
    }
  '
done

cd ../modfetch
rf '
  rm GoSumFile
  rm WorkspaceGoSumFiles
'

Change-Id: Iaa0f8a40192127457a539120eb94337940cb8d4a
Reviewed-on: https://go-review.googlesource.com/c/go/+/719700
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/go/internal/modfetch/codehost: fix flaky TestReadZip

In normal use by the go command, ReadZip always happens after Stat,
which makes sure that the relevant revision has been fetched to
local disk. But TestReadZip calls ReadZip directly, and ReadZip was not
ensuring that fetch had happened.

This made 'go test' pass (because other earlier tests had done Stat of
the hg test repo) but 'go test -run=ReadZip' fail by itself.
And on big enough machines, the tests that were doing the Stat
were running in parallel with ReadZip, causing non-deterministic
failures depending on whether the Stat completed before ReadZip started.

Fix the race by calling Stat directly in ReadZip, like we already do in ReadFile.

Fixes longtest builder flake.

Change-Id: Ib42f64876b7ef51d8148c616539b412b42f8b24e
Reviewed-on: https://go-review.googlesource.com/c/go/+/720280
Reviewed-by: David Chase <drchase@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>

runtime: prefer to restart Ps on the same M after STW

Today, Ps jump around arbitrarily across STW. Instead, try to keep the P
on the previous M it ran on. In the future, we'll likely want to try to
expand this beyond STW to create a more general affinity for specific
Ms.

For this to be useful, the Ps need to have runnable Gs. Today, STW
preemption goes through goschedImpl, which places the G on the global
run queue. If that was the only G then the P won't have runnable
goroutines anymore.

It makes more sense to keep the G with its P across STW anyway, so add a
special case to goschedImpl for that.

On my machine, this CL reduces the error rate in TestTraceSTW from 99.8%
to 1.9%.

As a nearly 2% error rate shows, there are still cases where this best
effort scheduling doesn't work. The most obvious is that while
procresize assigns Ps back to their original M, startTheWorldWithSema
calls wakep to start a spinning M. The spinning M may steal a goroutine
from another P if that P is too slow to start.

For #65694.

Change-Id: I6a6a636c0969c587d039b68bc68ea16c74ff1fc9
Reviewed-on: https://go-review.googlesource.com/c/go/+/714801
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>

runtime/pprof: fix goroutine leak profile tests for noopt

The goroutine leak profile tests currently rely on a function being
inlined, which results in a slightly different representation in the
pprof proto. This function is not inlined on the noopt builder.

Disable inlining of the one function which could be inlined to align but
the regular and noopt builder versions of this test. We're not
interested in testing the inlining functionality of profiles with this
test, we care about certain stack frames appearing in a certain order.

This also simplifies a bunch of the checking in the test.

Cq-Include-Trybots: luci.golang.try:gotip-linux-amd64-noopt
Change-Id: I28902cc4c9fae32d1e3fa41b93b00c3be4d6074a
Reviewed-on: https://go-review.googlesource.com/c/go/+/720100
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>

spec: remove cycle restriction for type parameters

Fixes #75883.

Change-Id: I708c0594ef3182d3aca37a6358aa0a0ef89809b6
Reviewed-on: https://go-review.googlesource.com/c/go/+/711422
Reviewed-by: Robert Griesemer <gri@google.com>
Reviewed-by: Alan Donovan <adonovan@google.com>
Auto-Submit: Robert Griesemer <gri@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>

cmd/asm,cmd/internal/obj/riscv: add support for riscv compressed instructions

Add support for compressed instructions in the RISC-V assembler. This
implements instruction validation and encoding for all instructions in
the "C" extension.

It is worth noting that the validation and encoding of these instructions
is far more convoluted then the typical instruction validation and
encoding. While the current model has been followed for now, it would be
worth revisiting this in the future and potentially switching to a table
based or even per-instruction implementation.

Additionally, the current instruction encoding is lacking some of the bits
needed for compressed instructions - this is solved by compressedEncoding,
which provides the missing information. This will also be addressed in the
future, likely by changing the instruction encoding format.

Updates #71105

Change-Id: I0f9359d63f93ebbdc6e708e79429b2d61eae220d
Reviewed-on: https://go-review.googlesource.com/c/go/+/713020
Reviewed-by: Mark Ryan <markdryan@rivosinc.com>
Reviewed-by: Junyang Shao <shaojunyang@google.com>
Reviewed-by: Cherry Mui <cherryyz@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Meng Zhuo <mengzhuo1203@gmail.com>

cmd/internal/obj/riscv: implement better bit pattern encoding

Replace the extractBitAndShift function with an encodeBitPattern function.
This allows the caller to specify a slice of bits that are to be extracted
and encoded, rather than making multiple function calls and combining the
results.

Change-Id: I3d51caa10ecf714f2ad2fb66d38376202c4e0628
Reviewed-on: https://go-review.googlesource.com/c/go/+/702397
Reviewed-by: Junyang Shao <shaojunyang@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Mark Ryan <markdryan@rivosinc.com>
Reviewed-by: Meng Zhuo <mengzhuo1203@gmail.com>
Reviewed-by: Cherry Mui <cherryyz@google.com>

cmd/internal/obj/loong64: add {,X}VS{ADD,SUB}.{B/H/W/V}{,U} instructions support

Go asm syntax:
 VSADD{B,H,W,V}     V1, V2, V3
 VSADD{B,H,W,V}U    V1, V2, V3
 VSSUB{B,H,W,V}     V1, V2, V3
 VSSUB{B,H,W,V}U    V1, V2, V3
XVSADD{B,H,W,V}     X1, X2, X3
XVSADD{B,H,W,V}U    X1, X2, X3
XVSSUB{B,H,W,V}     X1, X2, X3
XVSSUB{B,H,W,V}U    X1, X2, X3

Equivalent platform assembler syntax:
 vsadd.{b,h,w,d}    v3, v2, v1
 vsadd.{b,h,w,d}u   v3, v2, v1
 vssub.{b,h,w,d}    v3, v2, v1
 vssub.{b,h,w,d}u   v3, v2, v1
xvsadd.{b,h,w,d}    x3, x2, x1
xvsadd.{b,h,w,d}u   x3, x2, x1
xvssub.{b,h,w,d}    x3, x2, x1
xvssub.{b,h,w,d}u   x3, x2, x1

Change-Id: Iab8c1a9bdc4940598936f3beac846466e913ffa2
Reviewed-on: https://go-review.googlesource.com/c/go/+/719200
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Meidan Li <limeidan@loongson.cn>
Reviewed-by: Junyang Shao <shaojunyang@google.com>
Reviewed-by: sophie zhao <zhaoxiaolin@loongson.cn>
Reviewed-by: Michael Pratt <mpratt@google.com>

cmd/go: keep objects alive while stopping cleanups

There are two places in cmd/go where cleanups are stopped before they
fire, and where the objects the cleanups are attached to may be dead
while we call Stop. This is essentially a race between Stop and the
cleanup being called. This can be fine, but these cleanups are used as
a way to check some invariants, so just panic if they're executed. As a
result, if they fire erroneously, they'll take down the whole process,
even if no invariant was actually violated.

The runtime.Cleanup.Stop documentation explains that users of Stop need
to hold the object alive across the call to Stop if they want to be sure
that Stop succeeds, so do that here by adding an explicit
runtime.KeepAlive call.

Kudos to Michael Pratt for finding the issue.

Fixes #74780.

Change-Id: I22e6f4642ac68f727ca3781f5d39a85015047925
Reviewed-on: https://go-review.googlesource.com/c/go/+/719961
Auto-Submit: Michael Knyszek <mknyszek@google.com>
Reviewed-by: Michael Pratt <mpratt@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Carlos Amedee <carlos@golang.org>

runtime: fix list test memory management for mayMoreStack

The NIH tests simply failed to allocate the nodes outside the heap at
all. The Manual tests failed to keep the nodes alive, allowing the GC to
collect them.

Cq-Include-Trybots: luci.golang.try:gotip-linux-amd64-longtest
Change-Id: I6a6a636c434bb703d6888383d32dbf95fb0a15ea
Reviewed-on: https://go-review.googlesource.com/c/go/+/719962
TryBot-Bypass: Michael Pratt <mpratt@google.com>
Auto-Submit: Michael Pratt <mpratt@google.com>
Reviewed-by: Michael Knyszek <mknyszek@google.com>

crypto/internal/fips140/sha3: remove outdated TODO

Change-Id: I6a6a6964725fed48b55ce72a24a8f9bd3044ae16
Reviewed-on: https://go-review.googlesource.com/c/go/+/719940
Reviewed-by: Roland Shoemaker <roland@golang.org>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Auto-Submit: Filippo Valsorda <filippo@golang.org>
Reviewed-by: Daniel McCarney <daniel@binaryparadox.net>
Reviewed-by: Junyang Shao <shaojunyang@google.com>

sync/atomic: amend comments for Value.{Swap,CompareAndSwap}

Related to CL 241661, CL 403094.

Change-Id: I86877d9a013b05b1a97f9aa1333cd96ce98469cb
GitHub-Last-Rev: 7fedca5848cb17730d4ccb586dc03cfcb1b83259
GitHub-Pull-Request: golang/go#76259
Reviewed-on: https://go-review.googlesource.com/c/go/+/719660
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Auto-Submit: Keith Randall <khr@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
Reviewed-by: Keith Randall <khr@google.com>
Reviewed-by: Junyang Shao <shaojunyang@google.com>

cmd/compile: don't clear ptrmask in fillptrmask

It is only ever called with a newly allocated slice.

This clearing code dates back to the C version of the compiler,
in which the function started like this:

static void
gengcmask(Type *t, uint8 gcmask[16])
{
...

memset(gcmask, 0, 16);
if(!haspointers(t))
return;

That memset was required for C, but not for Go.

Change-Id: I6fceb99b2dc8682685dca2e4289fcd58e2e5a0e0
Reviewed-on: https://go-review.googlesource.com/c/go/+/718340
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Keith Randall <khr@golang.org>
Auto-Submit: Ian Lance Taylor <iant@golang.org>
Reviewed-by: Keith Randall <khr@google.com>
Reviewed-by: Junyang Shao <shaojunyang@google.com>

internal/abi: correctly describe result in Name.Name doc comment

Change-Id: I3d0b138a4b5fbdd6a6c7a26984f8ce5d3ea4c7ec
Reviewed-on: https://go-review.googlesource.com/c/go/+/718320
Reviewed-by: David Chase <drchase@google.com>
Reviewed-by: Junyang Shao <shaojunyang@google.com>
Auto-Submit: Ian Lance Taylor <iant@golang.org>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>

cmd/compile: use FCLASSD for subnormal checks on riscv64

Only implemented for 64 bit floating point operations for now.

goos: linux
goarch: riscv64
pkg: math
cpu: Spacemit(R) X60
                    │       sec/op        │   sec/op     vs base                │
Acos                          154.1n ± 0%   154.1n ± 0%        ~ (p=0.303 n=10)
Acosh                         215.8n ± 6%   226.7n ± 0%        ~ (p=0.439 n=10)
Asin                          149.2n ± 1%   149.2n ± 0%        ~ (p=0.700 n=10)
Asinh                         262.1n ± 0%   258.5n ± 0%   -1.37% (p=0.000 n=10)
Atan                          99.48n ± 0%   99.49n ± 0%        ~ (p=0.836 n=10)
Atanh                         244.9n ± 0%   243.8n ± 0%   -0.43% (p=0.002 n=10)
Atan2                         158.2n ± 1%   153.3n ± 0%   -3.10% (p=0.000 n=10)
Cbrt                          186.8n ± 0%   181.1n ± 0%   -3.03% (p=0.000 n=10)
Ceil                          36.71n ± 1%   36.71n ± 0%        ~ (p=0.434 n=10)
Copysign                      6.531n ± 1%   6.526n ± 0%        ~ (p=0.268 n=10)
Cos                           98.19n ± 0%   95.40n ± 0%   -2.84% (p=0.000 n=10)
Cosh                          233.1n ± 0%   222.6n ± 0%   -4.50% (p=0.000 n=10)
Erf                           122.5n ± 0%   114.2n ± 0%   -6.78% (p=0.000 n=10)
Erfc                          126.0n ± 1%   116.6n ± 0%   -7.46% (p=0.000 n=10)
Erfinv                        138.8n ± 0%   138.6n ± 0%        ~ (p=0.082 n=10)
Erfcinv                       140.0n ± 0%   139.7n ± 0%        ~ (p=0.359 n=10)
Exp                           193.3n ± 0%   184.2n ± 0%   -4.68% (p=0.000 n=10)
ExpGo                         204.8n ± 0%   194.5n ± 0%   -5.03% (p=0.000 n=10)
Expm1                         152.5n ± 1%   145.0n ± 0%   -4.92% (p=0.000 n=10)
Exp2                          174.5n ± 0%   164.2n ± 0%   -5.85% (p=0.000 n=10)
Exp2Go                        184.4n ± 1%   175.4n ± 0%   -4.88% (p=0.000 n=10)
Abs                           4.912n ± 0%   4.914n ± 0%        ~ (p=0.283 n=10)
Dim                           15.50n ± 1%   15.52n ± 1%        ~ (p=0.331 n=10)
Floor                         36.89n ± 1%   36.76n ± 1%        ~ (p=0.325 n=10)
Max                           31.05n ± 1%   31.17n ± 1%        ~ (p=0.628 n=10)
Min                           31.01n ± 0%   31.06n ± 0%        ~ (p=0.767 n=10)
Mod                           294.1n ± 0%   245.6n ± 0%  -16.52% (p=0.000 n=10)
Frexp                         44.86n ± 1%   35.20n ± 0%  -21.53% (p=0.000 n=10)
Gamma                         195.8n ± 0%   185.4n ± 1%   -5.29% (p=0.000 n=10)
Hypot                         84.91n ± 0%   84.54n ± 1%   -0.43% (p=0.006 n=10)
HypotGo                       96.70n ± 0%   95.42n ± 1%   -1.32% (p=0.000 n=10)
Ilogb                         45.03n ± 0%   35.07n ± 1%  -22.10% (p=0.000 n=10)
J0                            634.5n ± 0%   627.2n ± 0%   -1.16% (p=0.000 n=10)
J1                            644.5n ± 0%   636.9n ± 0%   -1.18% (p=0.000 n=10)
Jn                            1.357µ ± 0%   1.344µ ± 0%   -0.92% (p=0.000 n=10)
Ldexp                         49.89n ± 0%   39.96n ± 0%  -19.90% (p=0.000 n=10)
Lgamma                        186.6n ± 0%   184.3n ± 0%   -1.21% (p=0.000 n=10)
Log                           150.4n ± 0%   141.1n ± 0%   -6.15% (p=0.000 n=10)
Logb                          46.70n ± 0%   35.89n ± 0%  -23.15% (p=0.000 n=10)
Log1p                         164.1n ± 0%   163.9n ± 0%        ~ (p=0.122 n=10)
Log10                         153.1n ± 0%   143.5n ± 0%   -6.24% (p=0.000 n=10)
Log2                          58.83n ± 0%   49.75n ± 0%  -15.43% (p=0.000 n=10)
Modf                          40.82n ± 1%   40.78n ± 0%        ~ (p=0.239 n=10)
Nextafter32                   49.15n ± 0%   48.93n ± 0%   -0.44% (p=0.011 n=10)
Nextafter64                   43.33n ± 0%   43.23n ± 0%        ~ (p=0.228 n=10)
PowInt                        269.4n ± 0%   243.8n ± 0%   -9.49% (p=0.000 n=10)
PowFrac                       618.0n ± 0%   571.7n ± 0%   -7.48% (p=0.000 n=10)
Pow10Pos                      13.09n ± 0%   13.05n ± 0%   -0.31% (p=0.003 n=10)
Pow10Neg                      30.99n ± 1%   30.99n ± 0%        ~ (p=0.173 n=10)
Round                         23.73n ± 0%   23.65n ± 0%   -0.36% (p=0.011 n=10)
RoundToEven                   27.87n ± 0%   27.73n ± 0%   -0.48% (p=0.003 n=10)
Remainder                     282.1n ± 0%   249.6n ± 0%  -11.52% (p=0.000 n=10)
Signbit                       11.46n ± 0%   11.42n ± 0%   -0.39% (p=0.003 n=10)
Sin                           115.2n ± 0%   113.2n ± 0%   -1.74% (p=0.000 n=10)
Sincos                        140.6n ± 0%   138.6n ± 0%   -1.39% (p=0.000 n=10)
Sinh                          252.0n ± 0%   241.4n ± 0%   -4.21% (p=0.000 n=10)
SqrtIndirect                  4.909n ± 0%   4.893n ± 0%   -0.34% (p=0.021 n=10)
SqrtLatency                   19.57n ± 1%   19.57n ± 0%        ~ (p=0.087 n=10)
SqrtIndirectLatency           19.64n ± 0%   19.57n ± 0%   -0.36% (p=0.025 n=10)
SqrtGoLatency                 198.1n ± 0%   197.4n ± 0%   -0.35% (p=0.014 n=10)
SqrtPrime                     5.733µ ± 0%   5.725µ ± 0%        ~ (p=0.116 n=10)
Tan                           149.1n ± 0%   146.8n ± 0%   -1.54% (p=0.000 n=10)
Tanh                          248.2n ± 1%   238.1n ± 0%   -4.05% (p=0.000 n=10)
Trunc                         36.86n ± 0%   36.70n ± 0%   -0.43% (p=0.029 n=10)
Y0                            638.2n ± 0%   633.6n ± 0%   -0.71% (p=0.000 n=10)
Y1                            641.8n ± 0%   636.1n ± 0%   -0.87% (p=0.000 n=10)
Yn                            1.358µ ± 0%   1.345µ ± 0%   -0.92% (p=0.000 n=10)
Float64bits                   5.721n ± 0%   5.709n ± 0%   -0.22% (p=0.044 n=10)
Float64frombits               4.905n ± 0%   4.893n ± 0%        ~ (p=0.266 n=10)
Float32bits                   12.27n ± 0%   12.23n ± 0%        ~ (p=0.122 n=10)
Float32frombits               4.909n ± 0%   4.893n ± 0%   -0.32% (p=0.024 n=10)
FMA                           6.556n ± 0%   6.526n ± 0%        ~ (p=0.283 n=10)
geomean                       86.82n        83.75n        -3.54%

Change-Id: I522297a79646d76543d516accce291f5a3cea337
Reviewed-on: https://go-review.googlesource.com/c/go/+/717560
Reviewed-by: Keith Randall <khr@golang.org>
Reviewed-by: Keith Randall <khr@google.com>
Auto-Submit: Keith Randall <khr@golang.org>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Junyang Shao <shaojunyang@google.com>

net/url: disallow raw IPv6 addresses in host

RFC 3986 requires square brackets around IPv6 addresses.
Parse's acceptance of raw IPv6 addresses is non compliant,
and complicates splitting out a port.

This is a resubmission of CL 710176 after the revert in CL 711800,
this time with a new urlstrictipv6 godebug to control the behavior.

Fixes #31024
Fixes #75223

Change-Id: I4cbe5bb84266b3efe9c98cf4300421ddf1df7291
Reviewed-on: https://go-review.googlesource.com/c/go/+/712840
Reviewed-by: Junyang Shao <shaojunyang@google.com>
Reviewed-by: Damien Neil <dneil@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>

cmd/compile: optimize liveness in stackalloc

The stackalloc code needs to run a liveness pass to build the
interference graph between stack slots. Because the values that we need
liveness over is so sparse, we can optimize the analysis by using a path
exploration algorithm rather than a iterative dataflow one

In local testing, this cuts 74.05 ms of CPU time off a build of cmd/compile.

Change-Id: I765ace87d5e8aae177e65eb63da482e3d698bea7
Reviewed-on: https://go-review.googlesource.com/c/go/+/718540
Reviewed-by: Keith Randall <khr@golang.org>
Auto-Submit: Keith Randall <khr@golang.org>
Reviewed-by: Junyang Shao <shaojunyang@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Keith Randall <khr@google.com>

crypto/x509: move BetterTLS suite from crypto/tls

Move the BetterTLS test suite from crypto/tls to crypto/x509. Despite
the name, the test suites we care about are actually related to X.509
path building and name constraint checking. As such it makes more sense
to include these in the crypto/x509 package, so we are more likely to
catch breaking behaviors during local testing.

Change-Id: I5237903dcc9d9f60d6c7070db3c996ceb643b04c
Reviewed-on: https://go-review.googlesource.com/c/go/+/719120
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Daniel McCarney <daniel@binaryparadox.net>
Reviewed-by: Junyang Shao <shaojunyang@google.com>
Auto-Submit: Roland Shoemaker <roland@golang.org>

cmd/link: change shdr and phdr from arrays to slices

Removes an arbitrary and unnecessary limit.

Change-Id: Iba04568ed5e6b1a8f8f23369f51f068e830f1059
Reviewed-on: https://go-review.googlesource.com/c/go/+/718600
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Cherry Mui <cherryyz@google.com>
Auto-Submit: Ian Lance Taylor <iant@golang.org>
Reviewed-by: Junyang Shao <shaojunyang@google.com>

runtime: switch p.gcFractionalMarkTime to atomic.Int64

atomic.Int64 automatically maintains proper alignment, avoiding the need
to manually adjust alignment back and forth as fields above change.

Change-Id: I6a6a636c4c3c366353f6dc8ecac473c075dd5cd9
Reviewed-on: https://go-review.googlesource.com/c/go/+/716700
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>

runtime,runtime/pprof: clean up goroutine leak profile writing

Cleaned up goroutine leak profile extraction:
- removed the acquisition of goroutineProfile semaphore
- inlined the call to saveg when recording stacks instead of using
doRecordGoroutineProfile, which had side-effects over
goroutineProfile fields.

Added regression tests for goroutine leak profiling frontend for binary
and debug=1 profile formats.
Added stress tests for concurrent goroutine and goroutine leak profile requests.

Change-Id: I55c1bcef11e9a7fb7699b4c5a2353e594d3e7173
GitHub-Last-Rev: 5e9eb3b1d80c4d2d9b668a01f6b39a7b42f7bb45
GitHub-Pull-Request: golang/go#76045
Reviewed-on: https://go-review.googlesource.com/c/go/+/714580
Reviewed-by: Michael Pratt <mpratt@google.com>
Reviewed-by: Michael Knyszek <mknyszek@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Junyang Shao <shaojunyang@google.com>

cmd/go: clarify the -o testflag is only for copying the binary

Fixes #74769

Change-Id: Iebbaea8fb1a25e30a541a827815def9e269a8135
Reviewed-on: https://go-review.googlesource.com/c/go/+/709255
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Michael Matloob <matloob@golang.org>
Reviewed-by: Ian Alexander <jitsu@google.com>
Reviewed-by: Michael Matloob <matloob@google.com>

mime: parse media types that contain braces

This CL fixes a bug introduced by CL 666655: isTokenChar would no longer
(but should) report true for '{' and '}'.

Fixes #76236

Change-Id: Ifc0953c30d7cae7bfba9bc4b6bb6951a83c52576
GitHub-Last-Rev: c91a75c2c8778a9a8343c6bb4fa89eb1f978059f
GitHub-Pull-Request: golang/go#76243
Reviewed-on: https://go-review.googlesource.com/c/go/+/719380
Reviewed-by: Sean Liao <sean@liao.dev>
Reviewed-by: Jorropo <jorropo.pgm@gmail.com>
Reviewed-by: Michael Knyszek <mknyszek@google.com>
Reviewed-by: Damien Neil <dneil@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>

os/exec: include Cmd.Start in the list of methods that run Cmd

Fixes #76265

Change-Id: I451271c5662dd3bcdeec07b55761b15f64c00dcd
Reviewed-on: https://go-review.googlesource.com/c/go/+/719860
Reviewed-by: Keith Randall <khr@golang.org>
Reviewed-by: Alan Donovan <adonovan@google.com>
Auto-Submit: 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>

std,cmd: go fix -any std cmd

This change mechanically replaces all occurrences of interface{}
by 'any' (where deemed safe by the 'any' modernizer) throughout
std and cmd, minus their vendor trees.

Since this fix is relatively numerous, it gets its own CL.

Also, 'go generate go/types'.

Change-Id: I14a6b52856c3291c1d27935409bca8d5fd4242a2
Reviewed-on: https://go-review.googlesource.com/c/go/+/719702
Commit-Queue: Alan Donovan <adonovan@google.com>
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>
Auto-Submit: Alan Donovan <adonovan@google.com>

runtime: doubly-linked sched.midle list

This will be used by CL 714801 to remove Ms from the middle of the list.

We could simply convert schedlink to the doubly-linked list, bringing
along all other uses of schedlink.

However, CL 714801 removes Ms from the middle of the midle list. It
would be an easy mistake to make to accidentally remove an M from
schedlink, assuming that it is on the midle list when it is actually on
a completely different list. Using separate a list node makes this
impossible.

For #65694.

Change-Id: I6a6a636c223d925fdc30c0c648460cbf5c2af4d6
Reviewed-on: https://go-review.googlesource.com/c/go/+/714800
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>

runtime: use new list type for spanSPMCs

Now that we have a reusable list type, use it to replace the custom
linked list code for spanSPMCs.

Change-Id: I6a6a636ca54f2ba4b5c7dddba607c94ebf3c3ac8
Reviewed-on: https://go-review.googlesource.com/c/go/+/714021
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>

runtime: reusable intrusive doubly-linked list

Unfortunately we have two nearly identical types. One for standard types
and one for sys.NotInHeap types or cases that must avoid write barriers.
The latter must use uintptr fields, as assignment to unsafe.Pointer
fields generates a write barrier.

Change-Id: I6a6a636c62d83fa93b991033c7108d3b934412ac
Reviewed-on: https://go-review.googlesource.com/c/go/+/714020
Reviewed-by: Michael Knyszek <mknyszek@google.com>
Commit-Queue: Michael Pratt <mpratt@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Auto-Submit: Michael Pratt <mpratt@google.com>

internal/trace/testtrace: fix flag name typos

Change-Id: I6a6a636cd82a3e22a482ea2b2ab1004c45e2c304
Reviewed-on: https://go-review.googlesource.com/c/go/+/719400
Auto-Submit: Michael Pratt <mpratt@google.com>
Reviewed-by: Michael Knyszek <mknyszek@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>

cmd/go: implement accurate pseudo-versions for Mercurial

This is CL 124515 (which only did Git) but for Mercurial,
quite a few years late. I'm not sure why we didn't do it
at the time - probably Mercurial blindness. Do it now.

Change-Id: I28f448a19143f7ce3de337cd1891bae86023b499
Reviewed-on: https://go-review.googlesource.com/c/go/+/718502
Reviewed-by: Michael Matloob <matloob@golang.org>
Reviewed-by: Michael Matloob <matloob@google.com>
Auto-Submit: Russ Cox <rsc@golang.org>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>

cmd/go/internal/vcweb: cache hg servers

Cuts TestScript/reuse_hg from 73s to 47s.
(Python startup is slow! What's left is all Python too!)

Change-Id: Ia7124d4819286b3820355e4f427ffcfdc125491b
Reviewed-on: https://go-review.googlesource.com/c/go/+/718501
Reviewed-by: Michael Matloob <matloob@golang.org>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Michael Matloob <matloob@google.com>
Auto-Submit: Russ Cox <rsc@golang.org>

cmd/go: implement -reuse for Mercurial repos

When we added -reuse in CL 411398, we only handled Git repos.
This was partly because we were focused on Git traffic,
partly because Git is the dominant module VCS, and
partly because I couldn't see how to retrieve the metadata needed
in other version control systems.

This CL adds -reuse support for Mercurial, the second
most popular VCS for modules, now that I see how to
implement it. Although the Mercurial command line does
not have sufficient information, the Mercurial Python API does,
so we ship and invoke a Mercurial extension written in Python
that can compute a hash of the remote repo without downloading
it entirely, as well as resolve a remote name to a hash or check
the continued existence of a hash. Then we can avoid
downloading the repo at all if it hasn't changed since the last check
or if the specific reference we need still resolves or exists.

Fixes #75119.

Change-Id: Ia47d89b15c1091c44efef9d325270fc400a412c4
Reviewed-on: https://go-review.googlesource.com/c/go/+/718382
Auto-Submit: Russ Cox <rsc@golang.org>
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>

spec: more precise prose for special case of append

As written, the special case for append may be interpreted
such that any first argument that is assignable to []byte
is permissible, including nil. This change makes makes it
clear that a slice argument is required.

Not a language change. The compiler always expected a slice
argument.

Fixes #76226.

Change-Id: I73a1725d10096690335d6edf5793beb5de73578f
Reviewed-on: https://go-review.googlesource.com/c/go/+/718880
Reviewed-by: Junyang Shao <shaojunyang@google.com>
Reviewed-by: Alan Donovan <adonovan@google.com>
Reviewed-by: Robert Griesemer <gri@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Auto-Submit: Robert Griesemer <gri@google.com>
Reviewed-by: Ian Lance Taylor <iant@golang.org>

go/types, types2: first argument to append must never be be nil

The current implementation followed the spec faithfully for
the special case for append. Per the spec:

As a special case, append also accepts a first argument assignable to
type []byte with a second argument of string type followed by ... .

As it happens, nil is assignable to []byte, so append(nil, ""...)
didn't get an error message but a subsequent assertion failed.

This CL ensures that the first argument to append is never nil and
always a slice. We should make the spec more precise (separate CL).

Fixes #76220.

Change-Id: I581d11827a75afbb257077814beea813d4fe2441
Reviewed-on: https://go-review.googlesource.com/c/go/+/718860
Auto-Submit: Robert Griesemer <gri@google.com>
Reviewed-by: Alan Donovan <adonovan@google.com>
Reviewed-by: Robert Griesemer <gri@google.com>
Reviewed-by: Brett Howell <devbrett90@gmail.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>

.gitignore: ignore go test artifacts

go test -artifacts writes artifacts to _artifacts. These are outputs
from a specific test and are never intended to be commited.

Change-Id: I6a6a636cd7689e5e664b190c83ccb93060f26d4e
Reviewed-on: https://go-review.googlesource.com/c/go/+/717521
Reviewed-by: Michael Knyszek <mknyszek@google.com>
Reviewed-by: Damien Neil <dneil@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Auto-Submit: Michael Pratt <mpratt@google.com>

internal/trace: add "command" to convert text traces to raw

This is primarily helpful for parsing traces dumped via CI.

cmd/dist doesn't like commands in std which are not actually part of the
Go distribution. So rather than using a real command, this is actually a
test which does the conversion.

Change-Id: I6a6a636c829a4acc0bce8cf7548105ad59d83c67
Reviewed-on: https://go-review.googlesource.com/c/go/+/716882
Reviewed-by: Michael Knyszek <mknyszek@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>

runtime: fix lock rank for work.spanSPMCs.lock

Currently this lock is treated like a leaf lock, but it's not one. It
can acquire the globalAlloc lock via fixalloc and persistentalloc.

This means we need to figure out where it can be acquired. In general,
it can be acquired by any write barrier, which is already incredibly
broad. AFAICT, it can't be acquired directly otherwise, except when
destroying a spanSPMC during procresize, in which case we'll be holding
sched.lock.

Fixes #75916.

Change-Id: I2da1f5b82c750bf4e8d6a8a562046b9a17fd44be
Reviewed-on: https://go-review.googlesource.com/c/go/+/717500
Reviewed-by: Michael Pratt <mpratt@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>

internal/runtime/sys,math/bits: eliminate bounds checks on len8tab

The compiler cannot currently determine that the accesses to len8tab
are within bounds. Cast to uint8 to avoid unnecessary bounds checks.

Fixes #76166

Change-Id: I1fd930bba2b20d3998252c476308642e08ce00b4
Reviewed-on: https://go-review.googlesource.com/c/go/+/718040
Reviewed-by: Meng Zhuo <mengzhuo1203@gmail.com>
Reviewed-by: Keith Randall <khr@google.com>
Reviewed-by: Keith Randall <khr@golang.org>
Auto-Submit: Keith Randall <khr@golang.org>
Reviewed-by: Junyang Shao <shaojunyang@google.com>
Auto-Submit: Joel Sing <joel@sing.id.au>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>

runtime: document that tracefpunwindoff applies to some profilers

The tracefpunwindoff GODEBUG applies to all the places where we use
frame pointer unwinding. Originally that was just the execution tracer,
but now we also use it for the block and mutex profilers.

Change-Id: I6a6a69641fede668d96335582dba6d43e62f6a51
Reviewed-on: https://go-review.googlesource.com/c/go/+/718522
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Michael Knyszek <mknyszek@google.com>
Auto-Submit: Nick Ripley <nick.ripley@datadoghq.com>
Reviewed-by: Michael Pratt <mpratt@google.com>

cmd/go: remove redundant AVX regex in security flag checks

Remove "-m(no-)?avx[0-9a-z]*" from validCompilerFlags in security.go,
because "-m(no-)?avx[0-9a-z.]*" (see line 108) already covers its
matching range.

Change-Id: Ic86a45eefa7639ed3a8cdee95f08021d9515678e
Reviewed-on: https://go-review.googlesource.com/c/go/+/717740
Reviewed-by: Junyang Shao <shaojunyang@google.com>
Auto-Submit: Michael Pratt <mpratt@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>

cmd/internal/obj/arm64: shorten constant integer loads

Large integer constants can take up to 4 instructions to encode.

We can encode some large constants with a single instruction, namely
those which are bit patterns (repetitions of certain runs of 0s and 1s).

Often the constants we want to encode are *close* to those bit patterns,
but don't exactly match. For those, we can use 2 instructions, one to
load the close-by bit pattern and one to fix up any mismatches.

The constants we use to strength reduce divides often fit this pattern.
For unsigned divides by 1 through 15, this CL applies to the constant
for N=3,5,6,10,12,15.

Triggers 17 times in hello world.

Change-Id: I623abf32961fb3e74d0a163f6822f0647cd94499
Reviewed-on: https://go-review.googlesource.com/c/go/+/717900
Auto-Submit: 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: Cherry Mui <cherryyz@google.com>

runtime/msan: use different msan routine for copying

__msan_memmove records the fact that we're copying memory, and
actually does the copy. Use instead __msan_copy_shadow, which
records the fact that we're copying memory, but doesn't actually
do the copy itself.

We're doing the copy ourselves, so we don't need msan to do it also.

More importantly, msan doing the copy clobbers the target before
we issue the write barrier, which causes pointers to get lost.

Fixes #76138

Change-Id: I17aea739f9444de21fac2bbfd81e48534a39481d
Reviewed-on: https://go-review.googlesource.com/c/go/+/719020
Reviewed-by: Cherry Mui <cherryyz@google.com>
Reviewed-by: t hepudds <thepudds1460@gmail.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Keith Randall <khr@google.com>
Auto-Submit: Keith Randall <khr@golang.org>
Reviewed-by: Radu Berinde <radu@cockroachlabs.com>
Reviewed-by: Ian Lance Taylor <iant@golang.org>

runtime: tweak wording for comment of mcache.flushGen

Change-Id: I5f59b2a2d18a7657892ae4c042a1b013cf8f9736
GitHub-Last-Rev: d5b5bb3e19a131df64aa5df8a3d131fbc61fd151
GitHub-Pull-Request: golang/go#76215
Reviewed-on: https://go-review.googlesource.com/c/go/+/718461
Reviewed-by: Michael Pratt <mpratt@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Carlos Amedee <carlos@golang.org>

sync: don't call Done when f() panics in WaitGroup.Go

This change is based on
https://github.com/golang/go/issues/76126#issuecomment-3473417226
by adonovan.

Fixes #76126
Fixes #74702

Change-Id: Ie404d8204be8917fa8a7b414bb6d319238267c83
GitHub-Last-Rev: b1beddcd725e9168d4d544a9d0322a5a6d8d65b2
GitHub-Pull-Request: golang/go#76172
Reviewed-on: https://go-review.googlesource.com/c/go/+/717760
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Mateusz Poliwczak <mpoliwczak34@gmail.com>
Auto-Submit: Michael Pratt <mpratt@google.com>
Reviewed-by: Alan Donovan <adonovan@google.com>
Reviewed-by: Michael Pratt <mpratt@google.com>

cmd/go: update goSum if necessary

During investigation into #74691, we found a case where the hash was
missing from the goSum global var.  This change updates DownloadZip to
update the goSum hash if necessary when both the zipfile and
ziphashfile are already present.

For #74691

Change-Id: I904b7265f667256d0c60b66503a7954b467c6454
Reviewed-on: https://go-review.googlesource.com/c/go/+/705215
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>

cmd/go: output missing GoDebug entries

The `go help mod edit` command references the GoDebug struct, but `go
mod edit -json` was not displaying them when appropriate.

Fixes #75105

Change-Id: Iec987882941e01b0cf4d4fe31dda9e7a6e2dde87
Reviewed-on: https://go-review.googlesource.com/c/go/+/706757
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Michael Matloob <matloob@google.com>
Reviewed-by: Michael Matloob <matloob@golang.org>

cmd/go: rewrite hgrepo1 test repo to be deterministic

The use of git convert made it non-determinstic.
Build it up by hand using hg commands instead.

Change-Id: Iaa9c0925864a7003ea61e5bf2a9196ff3e0a662b
Reviewed-on: https://go-review.googlesource.com/c/go/+/718520
Auto-Submit: Russ Cox <rsc@golang.org>
Reviewed-by: Michael Matloob <matloob@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Michael Matloob <matloob@golang.org>

cmd/go: copy git reuse and support repos to hg

The reuse_hg.txt is reuse_git.txt with a skip at the top
and a global substitute of git->hg and fetch->pull.

The prefixtagtests.txt and tagtests.txt are straight
copies of the git equivalents.

This is to set up a readable diff in the followup CL
that turns it into a reuse test for hg.

For #75119.

Change-Id: I50c7f5559ee6479f8328ed54c14eafcc7f1ecc1f
Reviewed-on: https://go-review.googlesource.com/c/go/+/718381
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/go: improve TestScript/reuse_git

Most reuse origin info in git should not list RepoSum; check that.

Also add tests of go list -reuse, which were omitted entirely
in CL 411398.

Also be extra sure there is no repo left in module cache during reuse tests.

For #75119.

Change-Id: Ia1436b8d9e17db49664d24f1d43c448ac7cd00d2
Reviewed-on: https://go-review.googlesource.com/c/go/+/718380
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Michael Matloob <matloob@google.com>
Reviewed-by: Michael Matloob <matloob@golang.org>

cmd/link: clean up some comments to Go standards

Also drop a couple of unused names.

Change-Id: I0f09775a276c34ece7809cf5d3f7c6b8cd1fa487
Reviewed-on: https://go-review.googlesource.com/c/go/+/718580
Auto-Submit: Ian Lance Taylor <iant@golang.org>
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>

runtime: correctly print panics before fatal-ing on defer

Fixes #67792

Change-Id: I93d16580cb31e54cee7c8490212404e4d0dec446
Reviewed-on: https://go-review.googlesource.com/c/go/+/613757
Reviewed-by: Keith Randall <khr@golang.org>
Reviewed-by: Keith Randall <khr@google.com>
Reviewed-by: Michael Pratt <mpratt@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Auto-Submit: Keith Randall <khr@golang.org>

runtime/cgo: improve error messages after pointer panic

This CL improves the error messages after panics due to the
sharing of an unpinned Go pointer (or a pointer to an unpinned Go
pointer) between Go and C.

This occurs when it is:
1. returned from Go to C (through cgoCheckResult)
2. passed as argument to a C function (through cgoCheckPointer).

An unpinned Go pointer refers to a memory location that might be moved or
freed by the garbage collector.

Therefore:

- change the signature of cgoCheckArg (it does the real work behind
  cgoCheckResult and cgoCheckPointer)
- change the signature of cgoCheckUnknownPointer (called by cgoCheckArg
  for checking unexpected pointers)
- introduce cgoFormatErr (it is called by cgoCheckArg and
  cgoCheckUnknownPointer to format panic error messages)
- update the cgo pointer tests (add new tests, and a field errTextRegexp
  to the struct ptrTest)
- remove a loop variable in TestPointerChecks (similar to CL 711640).

1. cgoCheckResult

When an unpinned Go pointer (or a pointer to an unpinned Go pointer) is
returned from Go to C,

     1  package main
     2
     3  /*
     4     #include <stdio.h>
     5
     6     extern void* GoFoo();
     7
     8     static void CFoo() { GoFoo();}
     9   */
    10  import (
    11          "C"
    12  )
    13
    14  //export GoFoo
    15  func GoFoo() map[int]int {
    16          return map[int]int{0: 1,}
    17  }
    18
    19  func main() {
    20          C.CFoo();
    21  }

This error shows up at runtime:

    panic: runtime error: cgo result is unpinned Go pointer or points to unpinned Go pointer

    goroutine 1 [running]:
    main._Cfunc_CFoo()
            _cgo_gotypes.go:46 +0x3a
    main.main()
            /mnt/tmp/parse.go:20 +0xf
    exit status 2

GoFoo is the faulty Go function; it is not mentioned in the error message.

Moreover the error does not say which kind of pointer caused the panic; for
instance, a Go map.

Retrieve name and file/line of the Go function, as well as the kind of
pointer; use them in the error message:

    panic: runtime error: /mnt/tmp/parse.go:15: result of Go function GoFoo called from cgo is unpinned Go map or points to unpinned Go map

    goroutine 1 [running]:
    main._Cfunc_CFoo()
            _cgo_gotypes.go:46 +0x3a
    main.main()
            /mnt/tmp/parse.go:20 +0xf
    exit status 2

2. cgoCheckPointer

When a pointer to an unpinned Go pointer is passed to a C function,

     1  package main
     2
     3  /*
     4  #include <stdio.h>
     5  void foo(void *bar) {}
     6  */
     7  import "C"
     8  import "unsafe"
     9
    10  func main() {
    11          m := map[int]int{0: 1,}
    12          C.foo(unsafe.Pointer(&m))
    13  }

This error shows up at runtime:

    panic: runtime error: cgo argument has Go pointer to unpinned Go pointer

    goroutine 1 [running]:
    main.main.func1(...)
            /mnt/tmp/cgomap.go:12
    main.main()
            /mnt/tmp/cgomap.go:12 +0x91
    exit status 2

Retrieve kind of pointer; use it in the error message.

    panic: runtime error: argument of cgo function has Go pointer to unpinned Go map

    goroutine 1 [running]:
    main.main.func1(...)
            /mnt/tmp/cgomap.go:12
    main.main()
            /mnt/tmp/cgomap.go:12 +0x9b
    exit status 2

Link: https://pkg.go.dev/cmd/cgo#hdr-Passing_pointers
Suggested-by: Ian Lance Taylor <iant@golang.org>
Fixes #75856

Change-Id: Ia72f01df016feeae0cddb2558ced51a1b07e4486
GitHub-Last-Rev: 76257c7dd7c52a3d88234d43ec7f22bda81edcdd
GitHub-Pull-Request: golang/go#75894
Reviewed-on: https://go-review.googlesource.com/c/go/+/711801
Reviewed-by: Funda Secgin <fundasecgin73@gmail.com>
Reviewed-by: 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>
Reviewed-by: Keith Randall <khr@google.com>

runtime: tweak wording for doc

Change-Id: Id0be4b8744ded98cd28f94d89898d8cbac0f543e
GitHub-Last-Rev: e8a610e10a6403ad2e6daf99d461a3e539a3630d
GitHub-Pull-Request: golang/go#76197
Reviewed-on: https://go-review.googlesource.com/c/go/+/718420
Reviewed-by: Keith Randall <khr@google.com>
Auto-Submit: Keith Randall <khr@golang.org>
Reviewed-by: Michael Pratt <mpratt@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Keith Randall <khr@golang.org>

cmd/link: move pclntab out of relro section

The .gopclntab section should not have any relocations.
Move it out of relro to regular rodata.

Note that this is tested by tests like TestNoTextrel in
cmd/cgo/internal/testshared.

The existing test TestMachoSectionsReadOnly looks for sections in a
Mach-O file that are read-only after relocations are applied
(this is marked by a segment with a flags field set to 0x10).
We remove the __gopclntab section, as that section is now read-only
at all times, not only after relocating.

For #76038

Change-Id: I7f837e423bf1e802509277f5dc7fdd1ed0228e32
Reviewed-on: https://go-review.googlesource.com/c/go/+/718065
Reviewed-by: Russ Cox <rsc@golang.org>
Reviewed-by: Cherry Mui <cherryyz@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Auto-Submit: Ian Lance Taylor <iant@golang.org>

iter: fix a tiny doc comment bug

Use the right name for method Pos.Value in a sample doc comment in the package overview.

Fixes #76187.

Change-Id: Id1f5b0ca4ea39493b10140bc304c57c081c805ee
GitHub-Last-Rev: 7bf5d0774749111cd13e40700b31a5a83c288db9
GitHub-Pull-Request: golang/go#76195
Reviewed-on: https://go-review.googlesource.com/c/go/+/718321
Reviewed-by: Dmitri Shuralyov <dmitshur@google.com>
Reviewed-by: Alan Donovan <adonovan@google.com>
Auto-Submit: Alan Donovan <adonovan@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>

Revert "runtime: remove the pc field of _defer struct"

This reverts commit 361d51a6b58bccaab0559e06737c918018a7a5fa.

Reason for revert: Breaks some tests inside Google (on arm64?)

Change-Id: Iaea45fdcf9b4f9d36553687ca7f479750fe559da
Reviewed-on: https://go-review.googlesource.com/c/go/+/718066
Auto-Submit: Keith Randall <khr@golang.org>
Reviewed-by: Cuong Manh Le <cuong.manhle.vn@gmail.com>
Reviewed-by: Youlin Feng <fengyoulin@live.com>
Reviewed-by: Keith Randall <khr@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: David Chase <drchase@google.com>

cmd/internal/obj/loong64: add {,X}VSHUF.{B/H/W/V} instructions support

Go asm syntax:
         VSHUFB             V1, V2, V3, V4
        XVSHUFB             X1, X2, X3, X4
         VSHUF{H/W/V}       V1, V2, V3
        XVSHUF{H/W/V}       X1, X2, X3

Equivalent platform assembler syntax:
         vshuf.b            v4, v3, v2, v1
        xvshuf.b            x4, x3, x2, x1
         vshuf.{h/w/d}      v3, v2, v1
        xvshuf.{h/w/d}      x3, x2, x1

Change-Id: I8983467495f587cf46083fd81cb024400c7dc2a7
Reviewed-on: https://go-review.googlesource.com/c/go/+/716804
Reviewed-by: Meidan Li <limeidan@loongson.cn>
Reviewed-by: Michael Knyszek <mknyszek@google.com>
Reviewed-by: Michael Pratt <mpratt@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: sophie zhao <zhaoxiaolin@loongson.cn>

cmd/internal/obj/loong64: add {,X}VEXTRINS.{B,H,W,V} instruction support

Go asm syntax:
 VEXTRINS{B,H,W,V} $0x1b, vj,vd
XVEXTRINS{B,H,W,V} $0x1b, vj,vd

Equivalent platform assembler syntax:
         vextrins.{b,h,w,d}     vd, vj, $0x1b
        xvextrins.{b,h,w,d}     xd, xj, $0x1b

Change-Id: Ibc0bf926befaa2f810cfedd9a40f7ad9a6a9d7fc
Reviewed-on: https://go-review.googlesource.com/c/go/+/716803
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: sophie zhao <zhaoxiaolin@loongson.cn>
Reviewed-by: Michael Pratt <mpratt@google.com>
Reviewed-by: Meidan Li <limeidan@loongson.cn>
Reviewed-by: Michael Knyszek <mknyszek@google.com>

internal/chacha8rand: replace VORV with instruction VMOVQ on loong64

Change-Id: Id67623f403abfca54a04fc4c47792cd5b6d5ab73
Reviewed-on: https://go-review.googlesource.com/c/go/+/716802
Reviewed-by: Michael Pratt <mpratt@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Michael Knyszek <mknyszek@google.com>
Reviewed-by: sophie zhao <zhaoxiaolin@loongson.cn>
Reviewed-by: Meidan Li <limeidan@loongson.cn>

cmd/compile: fix error message on loong64

Change-Id: I90428330b17ab9f93ae94a77cefc24464e225df5
Reviewed-on: https://go-review.googlesource.com/c/go/+/717700
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Michael Pratt <mpratt@google.com>
Reviewed-by: Meidan Li <limeidan@loongson.cn>
Reviewed-by: sophie zhao <zhaoxiaolin@loongson.cn>
Reviewed-by: Michael Knyszek <mknyszek@google.com>

cmd/link: support weak binding on darwin

Symbols loaded from host files can have the N_WEAK_REF bit set,
which is used to instruct the loader to not fail if that symbol
can't be resolved.

The Go internal linker should honor this information by setting the
BIND_SYMBOL_FLAGS_WEAK_IMPORT flag in the corresponding bind table
entry.

Fixes #76023

Cq-Include-Trybots: luci.golang.try:gotip-darwin-amd64-longtest,gotip-darwin-amd64_12,gotip-darwin-arm64_12,gotip-darwin-arm64_15,gotip-darwin-arm64-longtest,gotip-darwin-amd64_14
Change-Id: Id2cef247ec7a9cb08455844f3c30ff874772bb7b
Reviewed-on: https://go-review.googlesource.com/c/go/+/713760
Reviewed-by: Michael Knyszek <mknyszek@google.com>
Reviewed-by: Michael Pratt <mpratt@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>

cmd/go: use go.dev instead of golang.org in flag errors

Also add the URL to the one possible error that didn't have it.
It looks like CL 93836 just missed the third error case when
adding the URL.

Change-Id: I837f8a730b25adb42909c9dfbde0dad2f664fec5
Reviewed-on: https://go-review.googlesource.com/c/go/+/718220
Reviewed-by: Michael Matloob <matloob@google.com>
Auto-Submit: Russ Cox <rsc@golang.org>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Michael Matloob <matloob@golang.org>

cmd/go: fix TestScript/govcs

On my Mac, TestScript/govcs was failing because hg prints a URL
using 1.0.0.127.in-addr.arpa instead of 127.0.0.1, and my Mac
cannot resolve that name back into an IP address.

% host 127.0.0.1
1.0.0.127.in-addr.arpa domain name pointer localhost.
% host 1.0.0.127.in-addr.arpa
%

Change-Id: Ia0762342d5926d13d786fe66de40590dc8977ff5
Reviewed-on: https://go-review.googlesource.com/c/go/+/718184
Reviewed-by: Michael Matloob <matloob@golang.org>
Reviewed-by: Michael Matloob <matloob@google.com>
Auto-Submit: Russ Cox <rsc@golang.org>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>

cmd/go: silence TLS handshake errors during test

We don't need to see messages like "TLS handshake error: EOF"
during go test.

Change-Id: If6bf51e655119914f337b9e61448c99485af34f2
Reviewed-on: https://go-review.googlesource.com/c/go/+/718183
Reviewed-by: Michael Matloob <matloob@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Michael Matloob <matloob@golang.org>
Auto-Submit: Russ Cox <rsc@golang.org>

cmd/go: fix TestCgoPkgConfig on darwin with pkg-config installed

Most darwin systems don't have pkg-config installed and skip this test.
(And it doesn't run in all.bash because it is skipped during -short.)

But for those systems that have pkg-config and run the non-short tests,
it fails because the code was not writing out the bar.pc on darwin and
then expecting pkg-config to be able to tell us about the bar package.
Change the code to write out the bar entry always.

Fixes one failing case in 'go test cmd/go' on my Mac.

Change-Id: Ibc4e9826a652ce2e7c609b905b159ccf2d5a6444
Reviewed-on: https://go-review.googlesource.com/c/go/+/718182
Reviewed-by: Michael Matloob <matloob@golang.org>
Reviewed-by: Michael Matloob <matloob@google.com>
Auto-Submit: Russ Cox <rsc@golang.org>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>

cmd/go: fix TestScript/vet_flags

Test caching can cause an incorrect test failure when the vet step result
is reused, leading to not printing a vet command line at all.
Avoid that with -a (we are also using -n so no real work is done).

Fixes one failing case in 'go test cmd/go' on my Mac.

Change-Id: I028284436b1ecc77145c886db902845b524f4b97
Reviewed-on: https://go-review.googlesource.com/c/go/+/718181
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>
Auto-Submit: Russ Cox <rsc@golang.org>

cmd/go: fix TestScript/tool_build_as_needed

This test assumes that changing GOOS/GOARCH results in an
unrunnable binary, but that's not true if the user has
go_GOOS_GOARCH_exec programs in their path (like I do).
This test was timing out waiting to create a gomote to run
a windows/amd64 binary.

Rewrite the test not to assume that alternate GOOS/GOARCH
binaries are unrunnable.

Fixes one failing case in 'go test cmd/go' on my Mac.

Change-Id: Ib5f721f91e10d285820efb5995a3a9bc29833214
Reviewed-on: https://go-review.googlesource.com/c/go/+/718180
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>
Auto-Submit: Russ Cox <rsc@golang.org>

cmd/cgo: skip escape checks if call site has no argument

This CL improves on CL 579955.

When the call site has no argument,

     1  package main
     2
     3  /*
     4     #include <stdio.h>
     5     void foo() {printf("Hello from C\n");}
     6  */
     7  import "C"
     8
     9  func main() {
    10          C.foo()
    11  }

Escape checks are not needed,

    $ go tool cgo -objdir dir cgonoargs.go
    $ cat -n dir/_cgo_gotypes.go | sed -n '43,$p'
        43  func _Cfunc_foo() (r1 _Ctype_void) {
        44          _cgo_runtime_cgocall(_cgo_c8ba2f813f11_Cfunc_foo, uintptr(unsafe.Pointer(&r1)))
        45          if _Cgo_always_false {
        46          }
        47          return
        48  }

Skip escape checks if call site has no argument.

    $ cat -n dir/_cgo_gotypes.go | sed -n '43,$p'
        43  func _Cfunc_foo() (r1 _Ctype_void) {
        44          _cgo_runtime_cgocall(_cgo_c8ba2f813f11_Cfunc_foo, uintptr(unsafe.Pointer(&r1)))
        45          return
        46  }

For #75856

Change-Id: I9aac0b6fb2985f6833976099e7eead1f28971bee
GitHub-Last-Rev: 1aacde448c922903980420cf8a38a4827d76ad28
GitHub-Pull-Request: golang/go#76186
Reviewed-on: https://go-review.googlesource.com/c/go/+/718060
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Keith Randall <khr@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Reviewed-by: Michael Pratt <mpratt@google.com>
Reviewed-by: Keith Randall <khr@google.com>
Auto-Submit: Keith Randall <khr@golang.org>

os: ignore O_TRUNC errors on named pipes and terminal devices on Windows

Prior to Go 1.24, os.OpenFile used to support O_TRUNC on named pipes and
terminal devices, even when the truncation was really ignored. This
behavior was consistent with Unix semantics.

CL 618836 changed the implementation of os.OpenFile on Windows and
unintentionally started returning an error when O_TRUNC was used on such
files.

Fixes #76071

Change-Id: Id10d3d8120ae9aa0548ef05423a172ff4e502ff9
Reviewed-on: https://go-review.googlesource.com/c/go/+/716420
Reviewed-by: Michael Knyszek <mknyszek@google.com>
Reviewed-by: Alex Brainman <alex.brainman@gmail.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Damien Neil <dneil@google.com>
Reviewed-by: Michael Pratt <mpratt@google.com>

cmd/link, runtime: don't store text start in pcHeader

The textStart field requires a relocation, the only relocation in pclntab.
And nothing uses it. So remove it. Replace it with a zero,
which can itself be removed at some point in coordination with Delve.

For #76038

Change-Id: I35675c0868c5d957bb375e40b804c516ae0300ca
Reviewed-on: https://go-review.googlesource.com/c/go/+/717240
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>
Auto-Submit: Ian Lance Taylor <iant@golang.org>

cmd/link: don't generate .gosymtab section

Since Go 1.2 the section is always empty.

Also remove the code looking for .gosymtab in cmd/internal/objfile.

For #76038

Change-Id: Icd34c870ed0c6da8001e8d32305f79905ee2b066
Reviewed-on: https://go-review.googlesource.com/c/go/+/717200
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Cherry Mui <cherryyz@google.com>
Auto-Submit: Ian Lance Taylor <iant@golang.org>
Commit-Queue: Ian Lance Taylor <iant@golang.org>
Reviewed-by: Michael Pratt <mpratt@google.com>

cmd/link: add and use new SymKind SFirstUnallocated

The linker sources in several places used SXREF to mark the first
SymKind which is not allocated in memory. This is cryptic.
Instead use SFirstUnallocated, following the example of the
existing SFirstWritable.

Change-Id: If326ad63027402699094bcc49ef860db3772f82a
Reviewed-on: https://go-review.googlesource.com/c/go/+/715623
Reviewed-by: Than McIntosh <thanm@golang.org>
Auto-Submit: Ian Lance Taylor <iant@golang.org>
Reviewed-by: Cherry Mui <cherryyz@google.com>
Reviewed-by: Michael Pratt <mpratt@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>

cmd/link: remove misleading comment

The comment suggests that the text section is briefly writable.
That is not the case. As the earlier part of the comment explains,
part of the text section is mapped twice, once r-x and once rw-.
It is never the case that there is writable executable memory.

Change-Id: I56841e19a8a08f2515f29752536a5c8f180ac8c9
Reviewed-on: https://go-review.googlesource.com/c/go/+/715622
Auto-Submit: Ian Lance Taylor <iant@golang.org>
Reviewed-by: Cherry Mui <cherryyz@google.com>
Reviewed-by: Than McIntosh <thanm@golang.org>
Reviewed-by: Michael Pratt <mpratt@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>

cmd/link: remove unused SFILEPATH symbol kind

Last reference appears to have been removed in CL 227759.

Change-Id: Ieb9da0a69a8beb96dcb5309ca43cf1df61d39bce
Reviewed-on: https://go-review.googlesource.com/c/go/+/715541
Auto-Submit: Ian Lance Taylor <iant@golang.org>
Reviewed-by: Michael Pratt <mpratt@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Than McIntosh <thanm@golang.org>
Reviewed-by: Cherry Mui <cherryyz@google.com>

cmd/link: add comments for SymKind values

Change-Id: Ie297a19a59362e0f32eae20e511e298a0a87ab6b
Reviewed-on: https://go-review.googlesource.com/c/go/+/715540
Reviewed-by: Than McIntosh <thanm@golang.org>
Auto-Submit: Ian Lance Taylor <iant@golang.org>
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>

cmd/compile: faster liveness analysis in regalloc

Instead of iterating until dataflow stabilization, fill in values known to be
live at the beginning of loop headers. This reduces the number of passes
over the CFG from the depth of the loopnest to just 2.

In a test instrumented version of this change, run against
cmd/compile/internal/ssa, it brought the time spent in liveness analysis
down to 150.52ms from 225.49ms on my machine.

Change-Id: Ic72762eedfd1f10b1ba74c430ed62ab4ebd3ec5c
Reviewed-on: https://go-review.googlesource.com/c/go/+/695255
Reviewed-by: Keith Randall <khr@google.com>
Auto-Submit: Keith Randall <khr@golang.org>
Reviewed-by: Michael Pratt <mpratt@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Keith Randall <khr@golang.org>

internal/syscall/windows: fix ReOpenFile sentinel error value

ReOpenFile is documented to return INVALID_HANDLE_VALUE on error,
but the previous definition was checking for 0 instead.

Change-Id: Idec5e75e40b9f6c409e068d63a9b606781e80a46
Reviewed-on: https://go-review.googlesource.com/c/go/+/717320
Auto-Submit: Quim Muntal <quimmuntal@gmail.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Damien Neil <dneil@google.com>
Reviewed-by: Alex Brainman <alex.brainman@gmail.com>
Reviewed-by: Michael Pratt <mpratt@google.com>

cmd/compile/internal/ssa: simplify riscv64 FCLASSD rewrite rules

We don't need to check that the bit patterns of the constants
match, it is sufficient to just check the constant is equal to the
given value.

While we're here also change the FCLASSD rules to use a bit pattern
for the mask. I think this improves readability, particularly as
more uses of FCLASSD get added (e.g. CL 717560).

These changes should not affect codegen.

Change-Id: I92a6338dc71e6a71e04306f67d7d86016c6e9c47
Reviewed-on: https://go-review.googlesource.com/c/go/+/717580
Reviewed-by: Michael Pratt <mpratt@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Auto-Submit: Keith Randall <khr@golang.org>
Reviewed-by: Keith Randall <khr@google.com>
Reviewed-by: Keith Randall <khr@golang.org>

runtime: amend doc for setPinned

Change-Id: I9c5a8f8a031e368bda312c830dc266f5986e8b1a
GitHub-Last-Rev: 23145e8fdbc020ae17503e3abe4fe230ba1082ef
GitHub-Pull-Request: golang/go#76160
Reviewed-on: https://go-review.googlesource.com/c/go/+/717341
Reviewed-by: Keith Randall <khr@golang.org>
Auto-Submit: 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: Michael Pratt <mpratt@google.com>

cmd/compile/internal/ssa: more aggressive on dead auto elim

Propagate "unread" across OpMoves. If the addr of this auto is only used
by an OpMove as its source arg, and the OpMove's target arg is the addr
of another auto. If the 2nd auto can be eliminated, this one can also be
eliminated.

This CL eliminates unnecessary memory copies and makes the frame smaller
in the following code snippet:

func contains(m map[string][16]int, k string) bool {
        _, ok := m[k]
        return ok
}

These are the benchmark results followed by the benchmark code:

goos: linux
goarch: amd64
cpu: Intel(R) Core(TM) i7-8650U CPU @ 1.90GHz
                │   old.txt   │              new.txt                │
                │   sec/op    │   sec/op     vs base                │
Map1Access2Ok-8   9.582n ± 2%   9.226n ± 0%   -3.72% (p=0.000 n=20)
Map2Access2Ok-8   13.79n ± 1%   10.24n ± 1%  -25.77% (p=0.000 n=20)
Map3Access2Ok-8   68.68n ± 1%   12.65n ± 1%  -81.58% (p=0.000 n=20)

package main_test

import "testing"

var (
        m1 = map[int]int{}
        m2 = map[int][16]int{}
        m3 = map[int][256]int{}
)

func init() {
        for i := range 1000 {
                m1[i] = i
                m2[i] = [16]int{15:i}
                m3[i] = [256]int{255:i}
        }
}

func BenchmarkMap1Access2Ok(b *testing.B) {
        for i := range b.N {
                _, ok := m1[i%1000]
                if !ok {
                        b.Errorf("%d not found", i)
                }
        }
}

func BenchmarkMap2Access2Ok(b *testing.B) {
        for i := range b.N {
                _, ok := m2[i%1000]
                if !ok {
                        b.Errorf("%d not found", i)
                }
        }
}

func BenchmarkMap3Access2Ok(b *testing.B) {
        for i := range b.N {
                _, ok := m3[i%1000]
                if !ok {
                        b.Errorf("%d not found", i)
                }
        }
}

Fixes #75398

Change-Id: If75e9caaa50d460efc31a94565b9ba28c8158771
Reviewed-on: https://go-review.googlesource.com/c/go/+/702875
Reviewed-by: Keith Randall <khr@golang.org>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Auto-Submit: Keith Randall <khr@golang.org>
Reviewed-by: Keith Randall <khr@google.com>
Reviewed-by: Michael Pratt <mpratt@google.com>

cmd/cgo: drop pre-1.18 support

Now that the bootstrap compiler is 1.24, it's no longer needed.

Change-Id: I9b3d6b7176af10fbc580173d50130120b542e7f9
Reviewed-on: https://go-review.googlesource.com/c/go/+/717060
Reviewed-by: David Chase <drchase@google.com>
Reviewed-by: Michael Pratt <mpratt@google.com>
Auto-Submit: Ian Lance Taylor <iant@golang.org>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>

internal/strconv: handle %f with fixedFtoa when possible

Everyone writes papers about fast shortest-output formatting.
Eventually we also sped up fixed-length formatting %e and %g.
But we've neglected %f, which falls back to the slow general code
even for relatively trivial things like %.2f on 1.23.

This CL uses the fast path fixedFtoa for %f when possible by
estimating the number of digits needed.

benchmark \ host                  linux-arm64    local  linux-amd64       s7  linux-386  s7:GOARCH=386
                                      vs base  vs base      vs base  vs base    vs base        vs base
AppendFloat/Decimal                         ~        ~            ~   +0.30%          ~              ~
AppendFloat/Float                      -0.45%        ~       -2.20%        ~     -2.19%              ~
AppendFloat/Exp                        +0.12%        ~       +4.11%        ~          ~              ~
AppendFloat/NegExp                     +0.53%        ~            ~        ~          ~              ~
AppendFloat/LongExp                    +0.41%   -1.42%       +4.50%        ~          ~              ~
AppendFloat/Big                             ~   -1.25%       +3.69%        ~          ~              ~
AppendFloat/BinaryExp                  +0.38%   +1.68%            ~        ~     +2.65%         +0.97%
AppendFloat/32Integer                       ~        ~            ~        ~          ~              ~
AppendFloat/32ExactFraction                 ~        ~       -2.61%        ~          ~              ~
AppendFloat/32Point                    -0.41%        ~       -2.65%        ~          ~              ~
AppendFloat/32Exp                           ~        ~       +5.35%        ~     +1.44%         +0.39%
AppendFloat/32NegExp                   +0.30%        ~       +2.31%        ~          ~         +0.82%
AppendFloat/32Shortest                 +0.28%   -0.85%            ~        ~     -3.20%              ~
AppendFloat/32Fixed8Hard               -0.29%        ~            ~   -1.75%          ~         +4.30%
AppendFloat/32Fixed9Hard                    ~        ~            ~        ~          ~         +1.52%
AppendFloat/64Fixed1                   +0.61%   -2.03%            ~        ~          ~         +4.36%
AppendFloat/64Fixed2                        ~   -3.43%            ~        ~          ~         +1.03%
AppendFloat/64Fixed2.5                 +0.57%   -2.23%            ~        ~          ~         +2.66%
AppendFloat/64Fixed3                        ~   -1.64%            ~   +0.31%     +2.32%         +2.10%
AppendFloat/64Fixed4                   +0.15%   -2.11%            ~        ~     +1.48%         +1.58%
AppendFloat/64Fixed5Hard               +0.45%        ~       +1.58%        ~          ~         +1.73%
AppendFloat/64Fixed12                  -0.16%        ~       +1.63%   -1.23%     +3.93%         +2.42%
AppendFloat/64Fixed16                  -0.33%   -0.49%            ~        ~     +3.67%         +2.33%
AppendFloat/64Fixed12Hard              -0.58%        ~            ~        ~     +4.98%         +0.62%
AppendFloat/64Fixed17Hard              +0.27%   -0.94%            ~        ~     +2.07%         +1.79%
AppendFloat/64Fixed18Hard                   ~        ~            ~        ~          ~              ~
AppendFloat/64FixedF1                 -69.59%  -76.08%      -70.94%  -68.26%    -75.27%        -69.88%
AppendFloat/64FixedF2                 -76.28%  -81.82%      -76.95%  -77.34%    -83.53%        -80.04%
AppendFloat/64FixedF3                 -77.30%  -84.51%      -77.82%  -77.81%    -78.77%        -73.69%
AppendFloat/Slowpath64                      ~   -1.30%       +1.64%        ~     -2.66%         -0.44%
AppendFloat/SlowpathDenormal64         +0.11%   -1.69%            ~        ~     -2.90%              ~

host: linux-arm64
goos: linux
goarch: arm64
pkg: internal/strconv
cpu: unknown
                                 │ 1cc918cc725  │             b66c604f523              │
                                 │    sec/op    │    sec/op     vs base                │
AppendFloat/Decimal-8               60.22n ± 0%    60.21n ± 0%        ~ (p=0.416 n=20)
AppendFloat/Float-8                 88.93n ± 0%    88.53n ± 0%   -0.45% (p=0.000 n=20)
AppendFloat/Exp-8                   93.09n ± 0%    93.20n ± 0%   +0.12% (p=0.000 n=20)
AppendFloat/NegExp-8                93.06n ± 0%    93.56n ± 0%   +0.53% (p=0.000 n=20)
AppendFloat/LongExp-8               99.79n ± 0%   100.20n ± 0%   +0.41% (p=0.000 n=20)
AppendFloat/Big-8                   103.9n ± 0%    104.0n ± 0%        ~ (p=0.004 n=20)
AppendFloat/BinaryExp-8             47.34n ± 0%    47.52n ± 0%   +0.38% (p=0.000 n=20)
AppendFloat/32Integer-8             60.43n ± 0%    60.40n ± 0%        ~ (p=0.006 n=20)
AppendFloat/32ExactFraction-8       86.21n ± 0%    86.24n ± 0%        ~ (p=0.634 n=20)
AppendFloat/32Point-8               83.20n ± 0%    82.87n ± 0%   -0.41% (p=0.000 n=20)
AppendFloat/32Exp-8                 89.43n ± 0%    89.45n ± 0%        ~ (p=0.193 n=20)
AppendFloat/32NegExp-8              87.31n ± 0%    87.58n ± 0%   +0.30% (p=0.000 n=20)
AppendFloat/32Shortest-8            76.28n ± 0%    76.49n ± 0%   +0.28% (p=0.000 n=20)
AppendFloat/32Fixed8Hard-8          52.44n ± 0%    52.29n ± 0%   -0.29% (p=0.000 n=20)
AppendFloat/32Fixed9Hard-8          60.57n ± 0%    60.54n ± 0%        ~ (p=0.285 n=20)
AppendFloat/64Fixed1-8              46.27n ± 0%    46.55n ± 0%   +0.61% (p=0.000 n=20)
AppendFloat/64Fixed2-8              46.77n ± 0%    46.80n ± 0%        ~ (p=0.060 n=20)
AppendFloat/64Fixed2.5-8            43.70n ± 0%    43.95n ± 0%   +0.57% (p=0.000 n=20)
AppendFloat/64Fixed3-8              47.22n ± 0%    47.19n ± 0%        ~ (p=0.008 n=20)
AppendFloat/64Fixed4-8              44.07n ± 0%    44.13n ± 0%   +0.15% (p=0.000 n=20)
AppendFloat/64Fixed5Hard-8          51.81n ± 0%    52.04n ± 0%   +0.45% (p=0.000 n=20)
AppendFloat/64Fixed12-8             78.41n ± 0%    78.29n ± 0%   -0.16% (p=0.000 n=20)
AppendFloat/64Fixed16-8             65.14n ± 0%    64.93n ± 0%   -0.33% (p=0.000 n=20)
AppendFloat/64Fixed12Hard-8         62.12n ± 0%    61.76n ± 0%   -0.58% (p=0.000 n=20)
AppendFloat/64Fixed17Hard-8         73.93n ± 0%    74.13n ± 0%   +0.27% (p=0.000 n=20)
AppendFloat/64Fixed18Hard-8         4.285µ ± 0%    4.283µ ± 0%        ~ (p=0.039 n=20)
AppendFloat/64FixedF1-8            216.10n ± 0%    65.71n ± 0%  -69.59% (p=0.000 n=20)
AppendFloat/64FixedF2-8            227.70n ± 0%    54.02n ± 0%  -76.28% (p=0.000 n=20)
AppendFloat/64FixedF3-8            208.20n ± 1%    47.25n ± 0%  -77.30% (p=0.000 n=20)
AppendFloat/Slowpath64-8            97.40n ± 0%    97.45n ± 0%        ~ (p=0.018 n=20)
AppendFloat/SlowpathDenormal64-8    94.75n ± 0%    94.86n ± 0%   +0.11% (p=0.000 n=20)
geomean                             87.86n         76.99n       -12.37%

host: local
goos: darwin
cpu: Apple M3 Pro
                                  │ 1cc918cc725  │             b66c604f523             │
                                  │    sec/op    │   sec/op     vs base                │
AppendFloat/Decimal-12               21.05n ± 1%   20.91n ± 1%        ~ (p=0.051 n=20)
AppendFloat/Float-12                 32.13n ± 0%   32.04n ± 1%        ~ (p=0.457 n=20)
AppendFloat/Exp-12                   31.84n ± 0%   31.72n ± 0%        ~ (p=0.151 n=20)
AppendFloat/NegExp-12                31.78n ± 1%   31.79n ± 1%        ~ (p=0.867 n=20)
AppendFloat/LongExp-12               33.70n ± 0%   33.22n ± 1%   -1.42% (p=0.000 n=20)
AppendFloat/Big-12                   35.52n ± 1%   35.07n ± 1%   -1.25% (p=0.000 n=20)
AppendFloat/BinaryExp-12             19.32n ± 1%   19.64n ± 0%   +1.68% (p=0.000 n=20)
AppendFloat/32Integer-12             21.32n ± 0%   21.18n ± 1%        ~ (p=0.025 n=20)
AppendFloat/32ExactFraction-12       30.88n ± 0%   31.07n ± 0%        ~ (p=0.087 n=20)
AppendFloat/32Point-12               30.88n ± 0%   30.95n ± 1%        ~ (p=0.250 n=20)
AppendFloat/32Exp-12                 31.57n ± 0%   31.67n ± 2%        ~ (p=0.126 n=20)
AppendFloat/32NegExp-12              30.50n ± 1%   30.76n ± 1%        ~ (p=0.087 n=20)
AppendFloat/32Shortest-12            27.14n ± 0%   26.91n ± 1%   -0.85% (p=0.001 n=20)
AppendFloat/32Fixed8Hard-12          17.11n ± 0%   17.08n ± 0%        ~ (p=0.027 n=20)
AppendFloat/32Fixed9Hard-12          19.16n ± 1%   19.31n ± 1%        ~ (p=0.062 n=20)
AppendFloat/64Fixed1-12              15.50n ± 0%   15.18n ± 1%   -2.03% (p=0.000 n=20)
AppendFloat/64Fixed2-12              15.46n ± 0%   14.93n ± 0%   -3.43% (p=0.000 n=20)
AppendFloat/64Fixed2.5-12            15.28n ± 0%   14.94n ± 1%   -2.23% (p=0.000 n=20)
AppendFloat/64Fixed3-12              15.58n ± 0%   15.32n ± 1%   -1.64% (p=0.000 n=20)
AppendFloat/64Fixed4-12              15.39n ± 0%   15.06n ± 1%   -2.11% (p=0.000 n=20)
AppendFloat/64Fixed5Hard-12          18.00n ± 0%   18.07n ± 1%        ~ (p=0.011 n=20)
AppendFloat/64Fixed12-12             27.97n ± 8%   29.05n ± 3%        ~ (p=0.107 n=20)
AppendFloat/64Fixed16-12             21.48n ± 0%   21.38n ± 0%   -0.49% (p=0.000 n=20)
AppendFloat/64Fixed12Hard-12         20.79n ± 1%   21.05n ± 2%        ~ (p=0.784 n=20)
AppendFloat/64Fixed17Hard-12         27.21n ± 1%   26.95n ± 1%   -0.94% (p=0.000 n=20)
AppendFloat/64Fixed18Hard-12         2.166µ ± 1%   2.182µ ± 1%        ~ (p=0.031 n=20)
AppendFloat/64FixedF1-12            103.35n ± 0%   24.72n ± 0%  -76.08% (p=0.000 n=20)
AppendFloat/64FixedF2-12            114.30n ± 1%   20.78n ± 0%  -81.82% (p=0.000 n=20)
AppendFloat/64FixedF3-12            107.10n ± 0%   16.58n ± 0%  -84.51% (p=0.000 n=20)
AppendFloat/Slowpath64-12            32.01n ± 0%   31.59n ± 0%   -1.30% (p=0.000 n=20)
AppendFloat/SlowpathDenormal64-12    30.21n ± 0%   29.70n ± 0%   -1.69% (p=0.000 n=20)
geomean                              31.84n        27.00n       -15.20%

host: linux-amd64
goos: linux
goarch: amd64
cpu: Intel(R) Xeon(R) CPU @ 2.30GHz
                                  │ 1cc918cc725  │             b66c604f523              │
                                  │    sec/op    │    sec/op     vs base                │
AppendFloat/Decimal-16               63.62n ± 1%    64.05n ± 1%        ~ (p=0.753 n=20)
AppendFloat/Float-16                 97.12n ± 1%    94.98n ± 1%   -2.20% (p=0.000 n=20)
AppendFloat/Exp-16                   98.12n ± 1%   102.15n ± 1%   +4.11% (p=0.000 n=20)
AppendFloat/NegExp-16                101.1n ± 1%    101.5n ± 1%        ~ (p=0.089 n=20)
AppendFloat/LongExp-16               104.5n ± 1%    109.2n ± 1%   +4.50% (p=0.000 n=20)
AppendFloat/Big-16                   108.5n ± 0%    112.5n ± 1%   +3.69% (p=0.000 n=20)
AppendFloat/BinaryExp-16             47.68n ± 1%    47.44n ± 1%        ~ (p=0.143 n=20)
AppendFloat/32Integer-16             63.77n ± 2%    63.45n ± 1%        ~ (p=0.015 n=20)
AppendFloat/32ExactFraction-16       97.69n ± 1%    95.14n ± 1%   -2.61% (p=0.000 n=20)
AppendFloat/32Point-16               92.17n ± 1%    89.72n ± 1%   -2.65% (p=0.000 n=20)
AppendFloat/32Exp-16                 95.63n ± 1%   100.75n ± 1%   +5.35% (p=0.000 n=20)
AppendFloat/32NegExp-16              94.53n ± 1%    96.72n ± 0%   +2.31% (p=0.000 n=20)
AppendFloat/32Shortest-16            86.43n ± 0%    86.95n ± 0%        ~ (p=0.010 n=20)
AppendFloat/32Fixed8Hard-16          57.75n ± 1%    57.95n ± 1%        ~ (p=0.098 n=20)
AppendFloat/32Fixed9Hard-16          66.56n ± 2%    66.97n ± 1%        ~ (p=0.380 n=20)
AppendFloat/64Fixed1-16              51.02n ± 1%    50.99n ± 1%        ~ (p=0.473 n=20)
AppendFloat/64Fixed2-16              50.94n ± 1%    51.01n ± 1%        ~ (p=0.136 n=20)
AppendFloat/64Fixed2.5-16            49.27n ± 1%    49.37n ± 1%        ~ (p=0.218 n=20)
AppendFloat/64Fixed3-16              51.85n ± 1%    52.55n ± 1%        ~ (p=0.045 n=20)
AppendFloat/64Fixed4-16              50.30n ± 1%    50.43n ± 1%        ~ (p=0.794 n=20)
AppendFloat/64Fixed5Hard-16          57.57n ± 1%    58.48n ± 1%   +1.58% (p=0.000 n=20)
AppendFloat/64Fixed12-16             82.67n ± 1%    84.02n ± 1%   +1.63% (p=0.000 n=20)
AppendFloat/64Fixed16-16             71.10n ± 1%    70.94n ± 1%        ~ (p=0.569 n=20)
AppendFloat/64Fixed12Hard-16         68.36n ± 1%    68.64n ± 1%        ~ (p=0.155 n=20)
AppendFloat/64Fixed17Hard-16         80.16n ± 1%    80.10n ± 1%        ~ (p=0.836 n=20)
AppendFloat/64Fixed18Hard-16         4.916µ ± 1%    4.919µ ± 1%        ~ (p=0.507 n=20)
AppendFloat/64FixedF1-16            239.75n ± 1%    69.67n ± 1%  -70.94% (p=0.000 n=20)
AppendFloat/64FixedF2-16            252.50n ± 1%    58.20n ± 1%  -76.95% (p=0.000 n=20)
AppendFloat/64FixedF3-16            238.00n ± 1%    52.79n ± 1%  -77.82% (p=0.000 n=20)
AppendFloat/Slowpath64-16            100.4n ± 1%    102.0n ± 1%   +1.64% (p=0.000 n=20)
AppendFloat/SlowpathDenormal64-16    97.92n ± 1%    98.01n ± 1%        ~ (p=0.304 n=20)
geomean                              95.58n         84.00n       -12.12%

host: s7
cpu: AMD Ryzen 9 7950X 16-Core Processor
                                  │ 1cc918cc725 │             b66c604f523             │
                                  │   sec/op    │   sec/op     vs base                │
AppendFloat/Decimal-32              22.00n ± 0%   22.06n ± 0%   +0.30% (p=0.001 n=20)
AppendFloat/Float-32                34.83n ± 0%   34.76n ± 0%        ~ (p=0.159 n=20)
AppendFloat/Exp-32                  34.91n ± 0%   34.89n ± 0%        ~ (p=0.188 n=20)
AppendFloat/NegExp-32               35.24n ± 0%   35.32n ± 0%        ~ (p=0.026 n=20)
AppendFloat/LongExp-32              37.02n ± 0%   37.02n ± 0%        ~ (p=0.317 n=20)
AppendFloat/Big-32                  38.51n ± 0%   38.43n ± 0%        ~ (p=0.060 n=20)
AppendFloat/BinaryExp-32            17.57n ± 0%   17.59n ± 0%        ~ (p=0.278 n=20)
AppendFloat/32Integer-32            22.06n ± 0%   22.09n ± 0%        ~ (p=0.762 n=20)
AppendFloat/32ExactFraction-32      32.91n ± 0%   33.00n ± 0%        ~ (p=0.055 n=20)
AppendFloat/32Point-32              33.24n ± 0%   33.18n ± 0%        ~ (p=0.068 n=20)
AppendFloat/32Exp-32                34.50n ± 0%   34.55n ± 0%        ~ (p=0.030 n=20)
AppendFloat/32NegExp-32             33.53n ± 0%   33.61n ± 0%        ~ (p=0.045 n=20)
AppendFloat/32Shortest-32           30.10n ± 0%   30.10n ± 0%        ~ (p=0.931 n=20)
AppendFloat/32Fixed8Hard-32         22.89n ± 0%   22.49n ± 0%   -1.75% (p=0.000 n=20)
AppendFloat/32Fixed9Hard-32         25.82n ± 0%   25.75n ± 1%        ~ (p=0.143 n=20)
AppendFloat/64Fixed1-32             18.80n ± 0%   18.70n ± 0%        ~ (p=0.004 n=20)
AppendFloat/64Fixed2-32             18.64n ± 1%   18.54n ± 0%        ~ (p=0.001 n=20)
AppendFloat/64Fixed2.5-32           17.89n ± 0%   17.81n ± 0%        ~ (p=0.001 n=20)
AppendFloat/64Fixed3-32             19.62n ± 0%   19.68n ± 0%   +0.31% (p=0.000 n=20)
AppendFloat/64Fixed4-32             18.64n ± 0%   18.82n ± 0%        ~ (p=0.010 n=20)
AppendFloat/64Fixed5Hard-32         21.62n ± 0%   21.57n ± 0%        ~ (p=0.058 n=20)
AppendFloat/64Fixed12-32            30.98n ± 1%   30.61n ± 1%   -1.23% (p=0.000 n=20)
AppendFloat/64Fixed16-32            26.89n ± 0%   27.08n ± 1%        ~ (p=0.003 n=20)
AppendFloat/64Fixed12Hard-32        26.03n ± 0%   26.20n ± 1%        ~ (p=0.344 n=20)
AppendFloat/64Fixed17Hard-32        30.03n ± 1%   29.72n ± 1%        ~ (p=0.001 n=20)
AppendFloat/64Fixed18Hard-32        1.824µ ± 0%   1.825µ ± 1%        ~ (p=0.567 n=20)
AppendFloat/64FixedF1-32            83.58n ± 1%   26.52n ± 0%  -68.26% (p=0.000 n=20)
AppendFloat/64FixedF2-32            89.68n ± 1%   20.32n ± 1%  -77.34% (p=0.000 n=20)
AppendFloat/64FixedF3-32            84.84n ± 0%   18.82n ± 0%  -77.81% (p=0.000 n=20)
AppendFloat/Slowpath64-32           35.55n ± 0%   35.61n ± 0%        ~ (p=0.394 n=20)
AppendFloat/SlowpathDenormal64-32   35.03n ± 0%   35.02n ± 0%        ~ (p=0.733 n=20)
geomean                             34.67n        30.31n       -12.56%

host: linux-386
goarch: 386
cpu: Intel(R) Xeon(R) CPU @ 2.30GHz
                                  │ 1cc918cc725 │             b66c604f523             │
                                  │   sec/op    │   sec/op     vs base                │
AppendFloat/Decimal-16              133.6n ± 1%   130.5n ± 1%        ~ (p=0.002 n=20)
AppendFloat/Float-16                242.3n ± 1%   237.0n ± 1%   -2.19% (p=0.000 n=20)
AppendFloat/Exp-16                  249.1n ± 3%   252.5n ± 1%        ~ (p=0.005 n=20)
AppendFloat/NegExp-16               248.7n ± 3%   253.8n ± 2%        ~ (p=0.006 n=20)
AppendFloat/LongExp-16              258.4n ± 2%   253.0n ± 6%        ~ (p=0.185 n=20)
AppendFloat/Big-16                  285.6n ± 1%   279.2n ± 5%        ~ (p=0.012 n=20)
AppendFloat/BinaryExp-16            89.47n ± 1%   91.85n ± 2%   +2.65% (p=0.000 n=20)
AppendFloat/32Integer-16            133.5n ± 1%   129.9n ± 1%        ~ (p=0.004 n=20)
AppendFloat/32ExactFraction-16      213.7n ± 1%   212.2n ± 2%        ~ (p=0.071 n=20)
AppendFloat/32Point-16              202.0n ± 0%   200.4n ± 1%        ~ (p=0.223 n=20)
AppendFloat/32Exp-16                236.4n ± 1%   239.8n ± 1%   +1.44% (p=0.000 n=20)
AppendFloat/32NegExp-16             212.5n ± 1%   211.9n ± 1%        ~ (p=0.995 n=20)
AppendFloat/32Shortest-16           200.3n ± 1%   193.9n ± 1%   -3.20% (p=0.000 n=20)
AppendFloat/32Fixed8Hard-16         136.0n ± 1%   133.2n ± 4%        ~ (p=0.323 n=20)
AppendFloat/32Fixed9Hard-16         155.6n ± 1%   156.7n ± 2%        ~ (p=0.022 n=20)
AppendFloat/64Fixed1-16             132.8n ± 1%   133.0n ± 3%        ~ (p=0.199 n=20)
AppendFloat/64Fixed2-16             128.9n ± 1%   129.7n ± 3%        ~ (p=0.018 n=20)
AppendFloat/64Fixed2.5-16           127.0n ± 1%   126.5n ± 3%        ~ (p=0.825 n=20)
AppendFloat/64Fixed3-16             127.3n ± 1%   130.3n ± 4%   +2.32% (p=0.001 n=20)
AppendFloat/64Fixed4-16             121.4n ± 1%   123.2n ± 2%   +1.48% (p=0.000 n=20)
AppendFloat/64Fixed5Hard-16         136.2n ± 1%   136.2n ± 3%        ~ (p=0.256 n=20)
AppendFloat/64Fixed12-16            159.0n ± 1%   165.2n ± 2%   +3.93% (p=0.000 n=20)
AppendFloat/64Fixed16-16            151.4n ± 0%   156.9n ± 1%   +3.67% (p=0.000 n=20)
AppendFloat/64Fixed12Hard-16        146.5n ± 1%   153.8n ± 1%   +4.98% (p=0.000 n=20)
AppendFloat/64Fixed17Hard-16        166.3n ± 1%   169.8n ± 1%   +2.07% (p=0.001 n=20)
AppendFloat/64Fixed18Hard-16        10.59µ ± 2%   10.60µ ± 0%        ~ (p=0.499 n=20)
AppendFloat/64FixedF1-16            614.4n ± 1%   152.0n ± 1%  -75.27% (p=0.000 n=20)
AppendFloat/64FixedF2-16            845.0n ± 0%   139.1n ± 1%  -83.53% (p=0.000 n=20)
AppendFloat/64FixedF3-16            608.8n ± 1%   129.3n ± 1%  -78.77% (p=0.000 n=20)
AppendFloat/Slowpath64-16           251.7n ± 1%   245.0n ± 1%   -2.66% (p=0.000 n=20)
AppendFloat/SlowpathDenormal64-16   248.4n ± 1%   241.2n ± 1%   -2.90% (p=0.000 n=20)
geomean                             225.7n        193.8n       -14.14%

host: s7:GOARCH=386
cpu: AMD Ryzen 9 7950X 16-Core Processor
                                  │ 1cc918cc725  │             b66c604f523             │
                                  │    sec/op    │   sec/op     vs base                │
AppendFloat/Decimal-32               41.88n ± 0%   42.02n ± 1%        ~ (p=0.004 n=20)
AppendFloat/Float-32                 71.05n ± 0%   71.24n ± 0%        ~ (p=0.044 n=20)
AppendFloat/Exp-32                   74.91n ± 1%   74.80n ± 0%        ~ (p=0.433 n=20)
AppendFloat/NegExp-32                74.10n ± 0%   74.20n ± 0%        ~ (p=0.867 n=20)
AppendFloat/LongExp-32               75.73n ± 0%   75.84n ± 0%        ~ (p=0.147 n=20)
AppendFloat/Big-32                   82.47n ± 0%   82.36n ± 0%        ~ (p=0.490 n=20)
AppendFloat/BinaryExp-32             32.31n ± 1%   32.62n ± 0%   +0.97% (p=0.000 n=20)
AppendFloat/32Integer-32             41.38n ± 1%   41.40n ± 1%        ~ (p=0.106 n=20)
AppendFloat/32ExactFraction-32       62.72n ± 0%   62.92n ± 0%        ~ (p=0.009 n=20)
AppendFloat/32Point-32               60.36n ± 0%   60.33n ± 0%        ~ (p=0.050 n=20)
AppendFloat/32Exp-32                 68.97n ± 0%   69.24n ± 0%   +0.39% (p=0.000 n=20)
AppendFloat/32NegExp-32              62.63n ± 0%   63.15n ± 0%   +0.82% (p=0.000 n=20)
AppendFloat/32Shortest-32            58.76n ± 0%   58.87n ± 0%        ~ (p=0.053 n=20)
AppendFloat/32Fixed8Hard-32          41.67n ± 1%   43.46n ± 1%   +4.30% (p=0.000 n=20)
AppendFloat/32Fixed9Hard-32          49.78n ± 1%   50.53n ± 1%   +1.52% (p=0.000 n=20)
AppendFloat/64Fixed1-32              41.15n ± 0%   42.95n ± 1%   +4.36% (p=0.000 n=20)
AppendFloat/64Fixed2-32              40.83n ± 1%   41.24n ± 1%   +1.03% (p=0.000 n=20)
AppendFloat/64Fixed2.5-32            39.42n ± 0%   40.47n ± 1%   +2.66% (p=0.000 n=20)
AppendFloat/64Fixed3-32              40.73n ± 1%   41.58n ± 1%   +2.10% (p=0.000 n=20)
AppendFloat/64Fixed4-32              38.68n ± 0%   39.29n ± 0%   +1.58% (p=0.000 n=20)
AppendFloat/64Fixed5Hard-32          42.88n ± 1%   43.62n ± 1%   +1.73% (p=0.000 n=20)
AppendFloat/64Fixed12-32             51.67n ± 1%   52.92n ± 1%   +2.42% (p=0.000 n=20)
AppendFloat/64Fixed16-32             49.15n ± 0%   50.30n ± 0%   +2.33% (p=0.000 n=20)
AppendFloat/64Fixed12Hard-32         48.51n ± 0%   48.81n ± 0%   +0.62% (p=0.001 n=20)
AppendFloat/64Fixed17Hard-32         54.62n ± 1%   55.60n ± 1%   +1.79% (p=0.000 n=20)
AppendFloat/64Fixed18Hard-32         3.979µ ± 1%   3.980µ ± 1%        ~ (p=0.569 n=20)
AppendFloat/64FixedF1-32            165.90n ± 1%   49.97n ± 0%  -69.88% (p=0.000 n=20)
AppendFloat/64FixedF2-32            225.50n ± 0%   45.02n ± 1%  -80.04% (p=0.000 n=20)
AppendFloat/64FixedF3-32            160.20n ± 1%   42.16n ± 1%  -73.69% (p=0.000 n=20)
AppendFloat/Slowpath64-32            75.55n ± 0%   75.23n ± 0%   -0.44% (p=0.000 n=20)
AppendFloat/SlowpathDenormal64-32    74.84n ± 0%   75.00n ± 0%        ~ (p=0.268 n=20)
geomean                              69.22n        61.13n       -11.69%

Change-Id: I722d2e2621e74e32cb3fc34a2df5b16cc595715c
Reviewed-on: https://go-review.googlesource.com/c/go/+/717183
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Alan Donovan <adonovan@google.com>

cmd/compile: implement Avg64u, Hmul64, Hmul64u for wasm

This lets us remove useAvg and useHmul from the division rules.
The compiler is simpler and the generated code is faster.

goos: wasip1
goarch: wasm
pkg: internal/strconv
                               │   old.txt   │               new.txt               │
                               │   sec/op    │   sec/op     vs base                │
AppendFloat/Decimal              192.8n ± 1%   194.6n ± 0%   +0.91% (p=0.000 n=10)
AppendFloat/Float                328.6n ± 0%   279.6n ± 0%  -14.93% (p=0.000 n=10)
AppendFloat/Exp                  335.6n ± 1%   289.2n ± 1%  -13.80% (p=0.000 n=10)
AppendFloat/NegExp               336.0n ± 0%   289.1n ± 1%  -13.97% (p=0.000 n=10)
AppendFloat/LongExp              332.4n ± 0%   285.2n ± 1%  -14.20% (p=0.000 n=10)
AppendFloat/Big                  348.2n ± 0%   300.1n ± 0%  -13.83% (p=0.000 n=10)
AppendFloat/BinaryExp            137.4n ± 0%   138.2n ± 0%   +0.55% (p=0.001 n=10)
AppendFloat/32Integer            193.3n ± 1%   196.5n ± 0%   +1.66% (p=0.000 n=10)
AppendFloat/32ExactFraction      283.3n ± 0%   268.9n ± 1%   -5.08% (p=0.000 n=10)
AppendFloat/32Point              279.9n ± 0%   266.5n ± 0%   -4.80% (p=0.000 n=10)
AppendFloat/32Exp                300.1n ± 0%   288.3n ± 1%   -3.90% (p=0.000 n=10)
AppendFloat/32NegExp             288.2n ± 1%   277.9n ± 1%   -3.59% (p=0.000 n=10)
AppendFloat/32Shortest           261.7n ± 0%   250.2n ± 0%   -4.39% (p=0.000 n=10)
AppendFloat/32Fixed8Hard         173.3n ± 1%   158.9n ± 1%   -8.31% (p=0.000 n=10)
AppendFloat/32Fixed9Hard         180.0n ± 0%   167.9n ± 2%   -6.70% (p=0.000 n=10)
AppendFloat/64Fixed1             167.1n ± 0%   149.6n ± 1%  -10.50% (p=0.000 n=10)
AppendFloat/64Fixed2             162.4n ± 1%   146.5n ± 0%   -9.73% (p=0.000 n=10)
AppendFloat/64Fixed2.5           165.5n ± 0%   149.4n ± 1%   -9.70% (p=0.000 n=10)
AppendFloat/64Fixed3             166.4n ± 1%   150.2n ± 0%   -9.74% (p=0.000 n=10)
AppendFloat/64Fixed4             163.7n ± 0%   149.6n ± 1%   -8.62% (p=0.000 n=10)
AppendFloat/64Fixed5Hard         182.8n ± 1%   167.1n ± 1%   -8.61% (p=0.000 n=10)
AppendFloat/64Fixed12            222.2n ± 0%   208.8n ± 0%   -6.05% (p=0.000 n=10)
AppendFloat/64Fixed16            197.6n ± 1%   181.7n ± 0%   -8.02% (p=0.000 n=10)
AppendFloat/64Fixed12Hard        194.5n ± 0%   181.0n ± 0%   -6.99% (p=0.000 n=10)
AppendFloat/64Fixed17Hard        205.1n ± 1%   191.9n ± 0%   -6.44% (p=0.000 n=10)
AppendFloat/64Fixed18Hard        6.269µ ± 0%   6.643µ ± 0%   +5.97% (p=0.000 n=10)
AppendFloat/64FixedF1            211.7n ± 1%   197.0n ± 0%   -6.95% (p=0.000 n=10)
AppendFloat/64FixedF2            189.4n ± 0%   174.2n ± 0%   -8.08% (p=0.000 n=10)
AppendFloat/64FixedF3            169.0n ± 0%   154.9n ± 0%   -8.32% (p=0.000 n=10)
AppendFloat/Slowpath64           321.2n ± 0%   274.2n ± 1%  -14.63% (p=0.000 n=10)
AppendFloat/SlowpathDenormal64   307.4n ± 1%   261.2n ± 0%  -15.03% (p=0.000 n=10)
AppendInt                        3.367µ ± 1%   3.376µ ± 0%        ~ (p=0.517 n=10)
AppendUint                       675.5n ± 0%   676.9n ± 0%        ~ (p=0.196 n=10)
AppendIntSmall                   28.13n ± 1%   28.17n ± 0%   +0.14% (p=0.015 n=10)
AppendUintVarlen/digits=1        20.70n ± 0%   20.51n ± 1%   -0.89% (p=0.018 n=10)
AppendUintVarlen/digits=2        20.43n ± 0%   20.27n ± 0%   -0.81% (p=0.001 n=10)
AppendUintVarlen/digits=3        38.48n ± 0%   37.93n ± 0%   -1.43% (p=0.000 n=10)
AppendUintVarlen/digits=4        41.10n ± 0%   38.78n ± 1%   -5.62% (p=0.000 n=10)
AppendUintVarlen/digits=5        42.25n ± 1%   42.11n ± 0%   -0.32% (p=0.041 n=10)
AppendUintVarlen/digits=6        45.40n ± 1%   43.14n ± 0%   -4.98% (p=0.000 n=10)
AppendUintVarlen/digits=7        46.81n ± 1%   46.03n ± 0%   -1.66% (p=0.000 n=10)
AppendUintVarlen/digits=8        48.88n ± 1%   46.59n ± 1%   -4.68% (p=0.000 n=10)
AppendUintVarlen/digits=9        49.94n ± 2%   49.41n ± 1%   -1.06% (p=0.000 n=10)
AppendUintVarlen/digits=10       57.28n ± 1%   56.92n ± 1%   -0.62% (p=0.045 n=10)
AppendUintVarlen/digits=11       60.09n ± 1%   58.11n ± 2%   -3.30% (p=0.000 n=10)
AppendUintVarlen/digits=12       62.22n ± 0%   61.85n ± 0%   -0.59% (p=0.000 n=10)
AppendUintVarlen/digits=13       64.94n ± 0%   62.92n ± 0%   -3.10% (p=0.000 n=10)
AppendUintVarlen/digits=14       65.42n ± 1%   65.19n ± 1%   -0.34% (p=0.005 n=10)
AppendUintVarlen/digits=15       68.17n ± 0%   66.13n ± 0%   -2.99% (p=0.000 n=10)
AppendUintVarlen/digits=16       70.21n ± 1%   70.09n ± 1%        ~ (p=0.517 n=10)
AppendUintVarlen/digits=17       72.93n ± 0%   70.49n ± 0%   -3.34% (p=0.000 n=10)
AppendUintVarlen/digits=18       73.01n ± 0%   72.75n ± 0%   -0.35% (p=0.000 n=10)
AppendUintVarlen/digits=19       79.27n ± 1%   79.49n ± 1%        ~ (p=0.671 n=10)
AppendUintVarlen/digits=20       82.18n ± 0%   80.43n ± 1%   -2.14% (p=0.000 n=10)
geomean                          143.4n        136.0n        -5.20%

Change-Id: I8245814a0259ad13cf9225f57db8e9fe3d2e4267
Reviewed-on: https://go-review.googlesource.com/c/go/+/717407
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Reviewed-by: Cherry Mui <cherryyz@google.com>

encoding/pem: don't reslice in failure modes

We re-slice the data being processed at the stat of each loop. If the
var that we use to calculate where to re-slice is < 0 or > the length
of the remaining data, return instead of attempting to re-slice.

Change-Id: I1d6c2b6c596feedeea8feeaace370ea73ba02c4c
Reviewed-on: https://go-review.googlesource.com/c/go/+/715260
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
Auto-Submit: Roland Shoemaker <roland@golang.org>
Reviewed-by: Damien Neil <dneil@google.com>