]> Cypherpunks repositories - gostls13.git/log
gostls13.git
3 years ago[dev.typeparams] cmd/compile: simplify openDeferSave
Cherry Mui [Tue, 8 Jun 2021 20:55:36 +0000 (16:55 -0400)]
[dev.typeparams] cmd/compile: simplify openDeferSave

Now it is only used to save the deferred the function (closure),
which must be a function type. Simplify the code.

Change-Id: Id4b8f2760fbf39a95883df2327f97378e7edab88
Reviewed-on: https://go-review.googlesource.com/c/go/+/326060
Trust: Cherry Mui <cherryyz@google.com>
Run-TryBot: Cherry Mui <cherryyz@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Than McIntosh <thanm@google.com>
3 years ago[dev.typeparams] cmd/compile, runtime: remove _defer.siz field
Cherry Mui [Mon, 7 Jun 2021 22:18:00 +0000 (18:18 -0400)]
[dev.typeparams] cmd/compile, runtime: remove _defer.siz field

As deferred function now always has zero arguments, _defer.siz is
always 0 and can be removed.

Change-Id: Ibb89f65b2f9d2ba4aeabe50438cc3d4b6a88320b
Reviewed-on: https://go-review.googlesource.com/c/go/+/325921
Trust: Cherry Mui <cherryyz@google.com>
Run-TryBot: Cherry Mui <cherryyz@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Michael Knyszek <mknyszek@google.com>
3 years ago[dev.typeparams] runtime: allow deferproc split stack
Cherry Mui [Sat, 5 Jun 2021 00:07:50 +0000 (20:07 -0400)]
[dev.typeparams] runtime: allow deferproc split stack

deferproc was not allowed to split stack because it had a special
stack layout, where the go'd function's arguments were passed on
stack but not included in the signature (therefore the stack map).
Now it no longer has argument, so it does not need to be nosplit.

Change-Id: I6d4b5302bd6fea6642bb4202984d86e3ebbc9054
Reviewed-on: https://go-review.googlesource.com/c/go/+/325920
Trust: Cherry Mui <cherryyz@google.com>
Run-TryBot: Cherry Mui <cherryyz@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Michael Knyszek <mknyszek@google.com>
3 years ago[dev.typeparams] runtime: make deferproc take a func() argument
Cherry Mui [Fri, 4 Jun 2021 22:30:51 +0000 (18:30 -0400)]
[dev.typeparams] runtime: make deferproc take a func() argument

Previously it takes a *funcval, as it can be any function types.
Now it must be func(). Make it so.

Change-Id: I04273047b024386f55dbbd5fbda4767cbee7ac93
Reviewed-on: https://go-review.googlesource.com/c/go/+/325918
Trust: Cherry Mui <cherryyz@google.com>
Run-TryBot: Cherry Mui <cherryyz@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Michael Knyszek <mknyszek@google.com>
3 years ago[dev.typeparams] cmd/compile, runtime: remove the siz argument of newproc/deferproc
Cherry Mui [Fri, 4 Jun 2021 22:11:59 +0000 (18:11 -0400)]
[dev.typeparams] cmd/compile, runtime: remove the siz argument of newproc/deferproc

newproc/deferproc takes a siz argument for the go'd/deferred
function's argument size. Now it is always zero. Remove the
argument.

Change-Id: If1bb8d427e34015ccec0ba10dbccaae96757fa8c
Reviewed-on: https://go-review.googlesource.com/c/go/+/325917
Trust: Cherry Mui <cherryyz@google.com>
Run-TryBot: Cherry Mui <cherryyz@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Michael Knyszek <mknyszek@google.com>
3 years ago[dev.typeparams] runtime: remove tracebackdefers
Cherry Mui [Fri, 4 Jun 2021 21:18:09 +0000 (17:18 -0400)]
[dev.typeparams] runtime: remove tracebackdefers

tracebackdefers is used for scanning/copying deferred functions'
arguments. Now that deferred functions are always argumentless,
it does nothing. Remove.

Change-Id: I55bedabe5584ea41a12cdb03d55ec9692a5aacd9
Reviewed-on: https://go-review.googlesource.com/c/go/+/325916
Trust: Cherry Mui <cherryyz@google.com>
Run-TryBot: Cherry Mui <cherryyz@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Michael Knyszek <mknyszek@google.com>
3 years ago[dev.typeparams] runtime: remove variadic defer/go calls
Cherry Mui [Fri, 4 Jun 2021 21:04:46 +0000 (17:04 -0400)]
[dev.typeparams] runtime: remove variadic defer/go calls

Now that defer/go wrapping is used, deferred/go'd functions are
always argumentless. Remove the code handling arguments.

This CL is mostly removing the fallback code path. There are more
cleanups to be done, in later CLs.

Change-Id: I87bfd3fb2d759fbeb6487b8125c0f6992863d6e5
Reviewed-on: https://go-review.googlesource.com/c/go/+/325915
Trust: Cherry Mui <cherryyz@google.com>
Run-TryBot: Cherry Mui <cherryyz@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Michael Knyszek <mknyszek@google.com>
3 years ago[dev.typeparams] cmd/compile: remove variadic defer calls
Cherry Mui [Fri, 4 Jun 2021 20:32:03 +0000 (16:32 -0400)]
[dev.typeparams] cmd/compile: remove variadic defer calls

Now that defer wrapping is used, deferred function is always
argumentless. Remove the code handling arguments.

This CL is mostly removing the fallback code path. There are more
cleanups to be done, in later CLs.

Change-Id: If6c729d3055c7a507cb1f1a000f5bbd3ad7ff235
Reviewed-on: https://go-review.googlesource.com/c/go/+/325914
Trust: Cherry Mui <cherryyz@google.com>
Reviewed-by: Than McIntosh <thanm@google.com>
3 years ago[dev.typeparams] cmd/compile, runtime: always enable defer/go wrapping
Cherry Mui [Fri, 4 Jun 2021 17:53:18 +0000 (13:53 -0400)]
[dev.typeparams] cmd/compile, runtime: always enable defer/go wrapping

Hardwire regabidefers to true. Remove it from GOEXPERIMENTs.

Fallback paths are not cleaned up in this CL. That will be done
in later CLs.

Change-Id: Iec1112a1e55d5f6ef70232a5ff6e702f649071c4
Reviewed-on: https://go-review.googlesource.com/c/go/+/325913
Trust: Cherry Mui <cherryyz@google.com>
Run-TryBot: Cherry Mui <cherryyz@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Michael Knyszek <mknyszek@google.com>
Reviewed-by: Than McIntosh <thanm@google.com>
3 years ago[dev.typeparams] internal/goexperiment: regenerate generated files
Cherry Mui [Mon, 7 Jun 2021 22:22:14 +0000 (18:22 -0400)]
[dev.typeparams] internal/goexperiment: regenerate generated files

Rerun the generator. exp_regabi_{on,off}.go are gone, as "regabi"
itself is not a goexperiment that we test at run time (the
sub-experiments are).

Change-Id: Ic1f31b4ef2769a143f768e1b3dc7221041aafca9
Reviewed-on: https://go-review.googlesource.com/c/go/+/325912
Trust: Cherry Mui <cherryyz@google.com>
Run-TryBot: Cherry Mui <cherryyz@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Michael Knyszek <mknyszek@google.com>
3 years ago[dev.typeparams] cmd/compile/internal/types2: provide valid signature in errors invol...
Robert Griesemer [Sat, 5 Jun 2021 04:38:40 +0000 (21:38 -0700)]
[dev.typeparams] cmd/compile/internal/types2: provide valid signature in errors involving method expressions

This is an adjusted port of a similar fix in https://golang.org/cl/324733.

Fixes #46583.

Change-Id: Ica1410e4de561e64e58b753e3da04b32156cbaf6
Reviewed-on: https://go-review.googlesource.com/c/go/+/325369
Trust: Robert Griesemer <gri@golang.org>
Reviewed-by: Robert Findley <rfindley@google.com>
3 years ago[dev.typeparams] cmd/compile: do extra markObjects during iexport to deal with generics
Dan Scales [Fri, 4 Jun 2021 22:22:55 +0000 (15:22 -0700)]
[dev.typeparams] cmd/compile: do extra markObjects during iexport to deal with generics

markInlBody/markObject/markType don't fully work as they stand for
generic functions/methods, since markInlBody can't understand method
calls on generic types. Those method calls will be resolved to concrete
methods in a full instantiation, but markInlBody on a generic
function/method can't understand those method calls. So, we won't
necessarily cause export of the appropriate extra method/function bodies
needed for inlining in an instantiated function.

One way to do this is just to make sure that we call markType
on all generic types that are exported (whether explicitly exported via
a capitalized name or unexported types that are referenced by a generic
function body). That way, we will call markInlBody on all possible
generic methods that might be called.

Fixes the current problem for i386-softfloat builds on dev.typeparams.

Change-Id: I2d3625d26042296731bd3c44ba1938aa194d527e
Reviewed-on: https://go-review.googlesource.com/c/go/+/325329
Run-TryBot: Dan Scales <danscales@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
Trust: Dan Scales <danscales@google.com>

3 years ago[dev.typeparams] cmd/compile: convert generic values to interface type using dictionary
Keith Randall [Sat, 5 Jun 2021 00:19:09 +0000 (17:19 -0700)]
[dev.typeparams] cmd/compile: convert generic values to interface type using dictionary

When converting a variable of generic type to an interface, use the
entry in the dictionary for the type field instead of using the
compile-time type (which we only have when fully stenciling).

Note: this isn't all the conversions. Conversions often get processed
in the ir.OCALL case. Those aren't handled yet.

Change-Id: I9a6a4c572e3c54a8e8efad98365184dbb94c4487
Reviewed-on: https://go-review.googlesource.com/c/go/+/325330
Trust: Keith Randall <khr@golang.org>
Trust: Dan Scales <danscales@google.com>
Reviewed-by: Dan Scales <danscales@google.com>
3 years ago[dev.typeparams] cmd/compile: allow conversions from type parameter to interface
Keith Randall [Sat, 5 Jun 2021 05:54:08 +0000 (22:54 -0700)]
[dev.typeparams] cmd/compile: allow conversions from type parameter to interface

When converting from a type param to an interface, allow it if
the type bound implements that interface.

Query: some conversions go through this path, some use another path?
The test does

   var i interface{foo()int} = x

but

   i := (interface{foo()int})(x)

works at tip.

Change-Id: I84d497e5228c0e1d1c9d76ffebaedce09dc45e8e
Reviewed-on: https://go-review.googlesource.com/c/go/+/325409
Trust: Keith Randall <khr@golang.org>
Trust: Dan Scales <danscales@google.com>
Run-TryBot: Keith Randall <khr@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Dan Scales <danscales@google.com>
3 years ago[dev.typeparams] cmd/compile: introduce IsTypeParam() helper
Keith Randall [Sat, 5 Jun 2021 06:01:13 +0000 (23:01 -0700)]
[dev.typeparams] cmd/compile: introduce IsTypeParam() helper

better than Kind() == types.TTYPEPARAM

Change-Id: I4f35a177cd0cda3be615a92b7b2af1b5a60a3bbc
Reviewed-on: https://go-review.googlesource.com/c/go/+/325410
Trust: Keith Randall <khr@golang.org>
Trust: Dan Scales <danscales@google.com>
Run-TryBot: Keith Randall <khr@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Dan Scales <danscales@google.com>
3 years agoMerge "[dev.typeparams] all: merge master (8212707) into dev.typeparams" into dev...
Gerrit Code Review [Mon, 7 Jun 2021 19:48:09 +0000 (19:48 +0000)]
Merge "[dev.typeparams] all: merge master (8212707) into dev.typeparams" into dev.typeparams

3 years ago[dev.typeparams] cmd/compile: create .dict Param in the package of the instantiated...
Dan Scales [Sat, 5 Jun 2021 01:17:49 +0000 (18:17 -0700)]
[dev.typeparams] cmd/compile: create .dict Param in the package of the instantiated function

The instantiated functions are created in the source package of the
generic function, so all lookups of symbols should be relative to that
package, so all symbols are consistently in the source package.

Fixes #46575

Change-Id: Iba67b2ba8014a630c5d4e032c0f2f2fbaaedce65
Reviewed-on: https://go-review.googlesource.com/c/go/+/325529
Reviewed-by: Keith Randall <khr@golang.org>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
3 years ago[dev.typeparams] all: merge master (8212707) into dev.typeparams
Matthew Dempsky [Mon, 7 Jun 2021 19:27:24 +0000 (12:27 -0700)]
[dev.typeparams] all: merge master (8212707) into dev.typeparams

Conflicts:

- src/go/internal/gcimporter/iimport.go

  CL 325429 and CL 319931 made unrelated changes to adjacent lines

Merge List:

+ 2021-06-07 8212707871 crypto/elliptic: update P-521 docs to say it's constant-time
+ 2021-06-07 7406180012 fmt: split package documentation into more sections
+ 2021-06-07 e3176bbc3e crypto/tls: fix typo in Config.NextProtos docs
+ 2021-06-05 e1fa26026d spec: improve wording consistency by eliminating "specifier"
+ 2021-06-05 f490134126 spec: improve wording by choosing an official term "keyword"
+ 2021-06-05 e3cb381704 go/internal/gcimporter: don't waste CPU copying bytes in `io.ReadAll`
+ 2021-06-05 9d669ed47a misc/cgo/errors: use expected column numbers
+ 2021-06-04 95939e8de7 cmd/compile/internal/abi: fix typo in comment
+ 2021-06-04 831f9376d8 net/http: fix ResponseWriter.ReadFrom with short reads
+ 2021-06-04 3a9d906edc os: avoid finalizer race in windows process object
+ 2021-06-04 105c5b50e0 os: terminate windows processes via handle directly
+ 2021-06-04 79cd407f88 syscall: regenerate zsyscall_windows.go
+ 2021-06-04 c6b6211229 doc/go1.17: document testing changes for Go 1.17
+ 2021-06-04 0214440075 syscall: do not pass console handles to PROC_THREAD_ATTRIBUTE_HANDLE_LIST on Windows 7
+ 2021-06-04 962d5c997a cmd/compile,go/types: restrict use of unsafe.{Add,Slice} to go1.17 or newer
+ 2021-06-04 b29b123e07 cmd/compile: remove spurious ir.Dump
+ 2021-06-03 6d98301114 cmd/link: use correct alignment in PE DWARF sections
+ 2021-06-03 e0d029f758 runtime: avoid gp.lockedm race in exitsyscall0
+ 2021-06-02 dd7ba3ba2c net: don't rely on system hosts in TestCVE202133195
+ 2021-06-02 4f572d7076 io/fs: minor corrections to Sub docs
+ 2021-06-02 e11d14225c doc/go1.17: remove runtime section
+ 2021-06-02 6e189afd3e doc/go1.17: mention SYS_WAIT6/WEXITED on NetBSD
+ 2021-06-02 ff9f5fb859 cmd/link: recognize clang linker error in testCGOLTO
+ 2021-06-02 1c6a2ea2ea doc/go1.17: document time changes for Go1.17
+ 2021-06-02 d743e67e06 doc/go1.17: document flag changes for Go 1.17
+ 2021-06-02 dc8f87b749 runtime/internal/sys: generate //go:build lines in gengoos.go
+ 2021-06-02 84c0e5d47f cmd/link: move issue 43830 tests out of TestScript
+ 2021-06-02 cae68700cc runtime: fix formatting
+ 2021-06-01 567ee865f6 cmd/go: add declaration to cgo_lto_issue43830 test
+ 2021-06-01 24e9707cbf cmd/link, cmd/cgo: support -flto in CFLAGS

Change-Id: I9ef88e7de0f8b1841ed9604b613b41672df67e71

3 years ago[dev.typeparams] go/types: factor out constraint satisfaction check
Rob Findley [Thu, 3 Jun 2021 16:14:16 +0000 (12:14 -0400)]
[dev.typeparams] go/types: factor out constraint satisfaction check

This is a port of CL 322070 to go/types, adjusted for the different
error reporting API.

Change-Id: I75eafe015b5b00554116527ea021e7a5f9e0343b
Reviewed-on: https://go-review.googlesource.com/c/go/+/324759
Trust: Robert Findley <rfindley@google.com>
Run-TryBot: Robert Findley <rfindley@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Robert Griesemer <gri@golang.org>
3 years ago[dev.typeparams] go/types: simplify Interface accessors
Rob Findley [Thu, 3 Jun 2021 15:55:11 +0000 (11:55 -0400)]
[dev.typeparams] go/types: simplify Interface accessors

This is a straightforward port of CL 321850 to go/types.

Change-Id: I719c19b8839390fdfa961255c6f1e79561cda6e6
Reviewed-on: https://go-review.googlesource.com/c/go/+/324757
Trust: Robert Findley <rfindley@google.com>
Run-TryBot: Robert Findley <rfindley@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Robert Griesemer <gri@golang.org>
3 years ago[dev.typeparams] go/types: re-use existing code for Interface.Complete
Rob Findley [Thu, 3 Jun 2021 15:49:52 +0000 (11:49 -0400)]
[dev.typeparams] go/types: re-use existing code for Interface.Complete

This is a port of CL 321751 to go/types, adjusted to use token.Pos, and
to exclude a missing position from a panic message (an unresolved
comment on the original CL).

Change-Id: I5814067aecb67aca9d73f2093fb6004b769924f3
Reviewed-on: https://go-review.googlesource.com/c/go/+/324756
Trust: Robert Findley <rfindley@google.com>
Run-TryBot: Robert Findley <rfindley@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Robert Griesemer <gri@golang.org>
3 years ago[dev.typeparams] go/types: add Named.SetTParams and Named.Orig methods
Rob Findley [Thu, 3 Jun 2021 15:37:26 +0000 (11:37 -0400)]
[dev.typeparams] go/types: add Named.SetTParams and Named.Orig methods

This is a port of CL 309832 to go/types, adjusted to not export the new
API and to amend TestSizeof.

Change-Id: I67efd3ba9b921c8431528eba1cd88ec1f41898bb
Reviewed-on: https://go-review.googlesource.com/c/go/+/324755
Trust: Robert Findley <rfindley@google.com>
Run-TryBot: Robert Findley <rfindley@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Robert Griesemer <gri@golang.org>
3 years ago[dev.typeparams] go/types: move signature checking into separate file
Rob Findley [Thu, 3 Jun 2021 15:13:38 +0000 (11:13 -0400)]
[dev.typeparams] go/types: move signature checking into separate file

This is a port of CL 321590 to go/types. Specifically, the same checker
methods were moved.

Change-Id: If4522d316f29c6b6f887580aa037e6b6dedbb6ac
Reviewed-on: https://go-review.googlesource.com/c/go/+/324754
Trust: Robert Findley <rfindley@google.com>
Run-TryBot: Robert Findley <rfindley@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Robert Griesemer <gri@golang.org>
3 years agocrypto/elliptic: update P-521 docs to say it's constant-time
Filippo Valsorda [Mon, 7 Jun 2021 14:22:05 +0000 (10:22 -0400)]
crypto/elliptic: update P-521 docs to say it's constant-time

This is true since CL 315274.

Also adjust the P-256 note, since Add, Double, and IsOnCurve use the
generic, non-constant-time implementation.

Change-Id: I4b3b340f65bce91dcca30bcf86456cc8ce4dd4bb
Reviewed-on: https://go-review.googlesource.com/c/go/+/325650
Trust: Filippo Valsorda <filippo@golang.org>
Trust: Katie Hockman <katie@golang.org>
Run-TryBot: Filippo Valsorda <filippo@golang.org>
Reviewed-by: Katie Hockman <katie@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>

3 years agofmt: split package documentation into more sections
Branden J Brown [Wed, 2 Jun 2021 18:55:34 +0000 (14:55 -0400)]
fmt: split package documentation into more sections

The package-level documentation on fmt previously had only two formal
sections, for printing and scanning. Because of this, the section on
printing was very long, including some pseudo-sections describing
particular features. This feature makes those pseudo-sections into
proper sections, both to improve readability and so that those sections
have hyperlinks on documentation sites.

Fixes #46522

Change-Id: I38b7bc3447610faca446051da235edcbbd063f61
Reviewed-on: https://go-review.googlesource.com/c/go/+/324349
Reviewed-by: Rob Pike <r@golang.org>
Reviewed-by: Dmitri Shuralyov <dmitshur@golang.org>
Run-TryBot: Dmitri Shuralyov <dmitshur@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>
Trust: Tobias Klauser <tobias.klauser@gmail.com>

3 years agocrypto/tls: fix typo in Config.NextProtos docs
Filippo Valsorda [Mon, 7 Jun 2021 14:24:11 +0000 (10:24 -0400)]
crypto/tls: fix typo in Config.NextProtos docs

Change-Id: I916df584859595067e5e86c35607869397dbbd8c
Reviewed-on: https://go-review.googlesource.com/c/go/+/325651
Trust: Filippo Valsorda <filippo@golang.org>
Run-TryBot: Filippo Valsorda <filippo@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Dmitri Shuralyov <dmitshur@golang.org>
3 years ago[dev.typeparams] cmd/compile/internal/types2: return Universe for ((*Package)(nil...
Robert Griesemer [Sun, 6 Jun 2021 04:23:48 +0000 (21:23 -0700)]
[dev.typeparams] cmd/compile/internal/types2: return Universe for ((*Package)(nil)).Scope()

For #46594.

Change-Id: I53776cbdc1b8f6da511abe2c6659a7313d7a163a
Reviewed-on: https://go-review.googlesource.com/c/go/+/325469
Trust: Robert Griesemer <gri@golang.org>
Run-TryBot: Robert Griesemer <gri@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
3 years ago[dev.typeparams] cmd/compile: refactor export writing
Matthew Dempsky [Fri, 4 Jun 2021 20:14:32 +0000 (13:14 -0700)]
[dev.typeparams] cmd/compile: refactor export writing

This CL reorganizes export writing in preparation for unified IR:

1. It moves dumpexport into noder as noder.WriteExports so that it can
be extended to include unified IR's export data.

2. Adds an "extensions" flag to typecheck.WriteExports to control
whether the compiler-only extension data (e.g., function bodies and
linker symbol info) is included in the exports.

3. It moves the gc.exporter type into typecheck and renames it to
"crawler". The type originated as the implementation of
the (pre-iexport) binary exporter, but since the removal of bexport
it's been relegated to simply crawling the exported functions/bodies
graph to identify which inline bodies need to be included.

4. It changes inline.Inline_Flood into the method crawler.markInlBody.
Inline_Flood doesn't actually have anything to do with the rest of
inlining; its current name and location are just historical quirks.

Passes toolstash -cmp.

Change-Id: I6445e2de9d3ce500a3aded5a8e20b09f46d23dbc
Reviewed-on: https://go-review.googlesource.com/c/go/+/325212
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
Trust: Matthew Dempsky <mdempsky@google.com>
Trust: Robert Griesemer <gri@golang.org>
Reviewed-by: Robert Griesemer <gri@golang.org>
3 years ago[dev.typeparams] cmd/compile: refactor import reading
Matthew Dempsky [Fri, 4 Jun 2021 17:26:40 +0000 (10:26 -0700)]
[dev.typeparams] cmd/compile: refactor import reading

This CL restructures the gcimports importer to mmap the export data
into memory as a string, and then pass that same string to both the
typecheck and types2 importers.

This is primarily motivated by preparation for unified IR; but it
should also improve performance (fewer string copies) and reduces
divergance between the two importers.

Passes toolstash -cmp.

Change-Id: I397f720693e9e6360bfcb5acb12609ab339d251f
Reviewed-on: https://go-review.googlesource.com/c/go/+/325210
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
Trust: Matthew Dempsky <mdempsky@google.com>
Trust: Robert Griesemer <gri@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Robert Griesemer <gri@golang.org>
3 years agospec: improve wording consistency by eliminating "specifier"
DQNEO [Mon, 31 May 2021 03:33:28 +0000 (12:33 +0900)]
spec: improve wording consistency by eliminating "specifier"

The word "specifier" is used once only here and technically not defined.

Change-Id: Ifc9f0582f4eb3c3011ba60d8008234de511d4be6
Reviewed-on: https://go-review.googlesource.com/c/go/+/323730
Reviewed-by: Robert Griesemer <gri@golang.org>
Trust: Ian Lance Taylor <iant@golang.org>
Trust: Robert Griesemer <gri@golang.org>

3 years agospec: improve wording by choosing an official term "keyword"
DQNEO [Mon, 31 May 2021 03:15:12 +0000 (12:15 +0900)]
spec: improve wording by choosing an official term "keyword"

Replace "reserved word" by "keyword" as the latter is the official term.

Change-Id: I9f269759b872026034a9f47e4a761cff2d348ca0
Reviewed-on: https://go-review.googlesource.com/c/go/+/323729
Reviewed-by: Robert Griesemer <gri@golang.org>
Trust: Ian Lance Taylor <iant@golang.org>
Trust: Robert Griesemer <gri@golang.org>

3 years agogo/internal/gcimporter: don't waste CPU copying bytes in `io.ReadAll`
Sergey Zagursky [Fri, 4 Jun 2021 09:25:51 +0000 (12:25 +0300)]
go/internal/gcimporter: don't waste CPU copying bytes in `io.ReadAll`

`io.ReadAll` dynamically reallocates byte slice because it doesn't know
its size in advance. We don't need to read an entire file into memory
and therefore may use `bufio.Reader` to read its contents.

Fixes #46564

Change-Id: Id504b1512662b6dea4775d523455896fa4162ab3
Reviewed-on: https://go-review.googlesource.com/c/go/+/325429
Reviewed-by: Dominik Honnef <dominik@honnef.co>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
Trust: Matthew Dempsky <mdempsky@google.com>
Trust: Dominik Honnef <dominik@honnef.co>
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Go Bot <gobot@golang.org>

3 years agomisc/cgo/errors: use expected column numbers
Ian Lance Taylor [Fri, 4 Jun 2021 15:16:58 +0000 (08:16 -0700)]
misc/cgo/errors: use expected column numbers

The test was using the wrong column numbers, and was erroneously
passing because there happened to be line numbers that matched those
column numbers. Change the test harness to require the expected line
number for the ERROR HERE regexp case, so that this doesn't happen again.

Also rename a couple of variables in the test to avoid useless
redeclaration errors.

Fixes #46534

Change-Id: I2fcbf5e379c346de5346035c73d174a3980c0927
Reviewed-on: https://go-review.googlesource.com/c/go/+/324970
Trust: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
3 years ago[dev.typeparams] runtime/race: make test compatible with types2
Matthew Dempsky [Fri, 4 Jun 2021 06:52:34 +0000 (23:52 -0700)]
[dev.typeparams] runtime/race: make test compatible with types2

types2 correctly distinguishes variable assignment from use even
within function literals. Whatever the outcome of #3059, the test
cases in runtime/race need to be fixed to accomodate that.

Change-Id: Ibe3547f07b681ff41225caabaf050872a48c98d1
Reviewed-on: https://go-review.googlesource.com/c/go/+/325030
Trust: Matthew Dempsky <mdempsky@google.com>
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Robert Griesemer <gri@golang.org>
3 years ago[dev.typeparams] cmd/compile: rename (types2.Inferred.)Targs to TArgs
Robert Griesemer [Fri, 4 Jun 2021 04:58:32 +0000 (21:58 -0700)]
[dev.typeparams] cmd/compile: rename (types2.Inferred.)Targs to TArgs

This is consistent with Named.TArgs.

This is a straight-forward port of https://golang.org/cl/321289
plus the necessary compiler noder changes.

Change-Id: I50791e5abe0d7f294293bed65cebc8dde8bf8c06
Reviewed-on: https://go-review.googlesource.com/c/go/+/325010
Trust: Robert Griesemer <gri@golang.org>
Reviewed-by: Robert Findley <rfindley@google.com>
3 years ago[dev.typeparams] cmd/compile/internal/syntax: not all index expressions can be instan...
Robert Griesemer [Fri, 4 Jun 2021 03:53:08 +0000 (20:53 -0700)]
[dev.typeparams] cmd/compile/internal/syntax: not all index expressions can be instantiated types

An index expression followed by an opening "{" may indicate
a composite literal but only if the index expression can be
a type. Exclude cases where the index expression cannot be
a type (e.g. s[0], a[i+j], etc.).

This leads to a better error message in code that is erroneous.

Fixes #46558.

Change-Id: Ida9291ca30683c211812dfb95abe4969f44c474f
Reviewed-on: https://go-review.googlesource.com/c/go/+/325009
Trust: Robert Griesemer <gri@golang.org>
Reviewed-by: Robert Findley <rfindley@google.com>
3 years ago[dev.typeparams] cmd/compile: point StructKeyExpr at the types.Field
Matthew Dempsky [Fri, 4 Jun 2021 07:01:22 +0000 (00:01 -0700)]
[dev.typeparams] cmd/compile: point StructKeyExpr at the types.Field

When constructing struct literals, importers need a way to specify
precisely which field to initialize without worrying about visibility
or those fields being blank. (A blank field doesn't actually need to
be initialized, but the expression needs to be evaluated still, and
with the right order-of-operations.)

This CL changes StructKeyExpr's Field field to point directly to the
corresponding types.Field, rather than merely holding a copy of its
Sym and Offset. This is akin to past changes to add
SelectorExpr.Selection.

Change-Id: I95b72b1788f73206fcebc22b456cf6b1186db6a7
Reviewed-on: https://go-review.googlesource.com/c/go/+/325031
Trust: Matthew Dempsky <mdempsky@google.com>
Reviewed-by: Cuong Manh Le <cuong.manhle.vn@gmail.com>
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Go Bot <gobot@golang.org>

3 years agocmd/compile/internal/abi: fix typo in comment
sryoya [Fri, 4 Jun 2021 18:12:03 +0000 (03:12 +0900)]
cmd/compile/internal/abi: fix typo in comment

Change-Id: I196045314b2b0e908d7b31ac0cea5b25404f3ee0
Reviewed-on: https://go-review.googlesource.com/c/go/+/325249
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
Trust: Keith Randall <khr@golang.org>

3 years ago[dev.typeparams] cmd/compile: handle dictionaries for top-level instantiations
Keith Randall [Thu, 3 Jun 2021 22:39:23 +0000 (15:39 -0700)]
[dev.typeparams] cmd/compile: handle dictionaries for top-level instantiations

There's no outer function in these cases, so we won't be reading
the dictionary as a subdictionary from the outer scope's dictionary.
It will always be a compile-time constant.

Change-Id: I754b126652a6ffb62255734d53fcec29d77cfa9e
Reviewed-on: https://go-review.googlesource.com/c/go/+/324949
Trust: Keith Randall <khr@golang.org>
Trust: Dan Scales <danscales@google.com>
Reviewed-by: Dan Scales <danscales@google.com>
3 years ago[dev.typeparams] cmd/compile: allow inlining in instantiated functions
Dan Scales [Tue, 1 Jun 2021 17:49:14 +0000 (10:49 -0700)]
[dev.typeparams] cmd/compile: allow inlining in instantiated functions

Change markType to scan generic types and methods, so that inlineable
functions inside generic functions/methods will be properly marked for
export, which means inlining inside instantiated functions will work
correctly.

Also, fix handling of closures for instantiated functions. Some code
needs to be adjusted, since instantiated functions/methods are compiled
as if in the package of the source generic function/type, rather than in
the local package. When we create the closure struct, we want to make
sure that the .F field has the same package as the other fields for the
closure variables. Also, we need to disable a check in tcCompLit() when
being done for an instantiated function, since fields of the closure
struct will be from the source package, not the local package.

Re-enabled part of the orderedmapsimp test that was disabled because of
these issues.

Change-Id: Ic4dba8917da0a36b17c0bdb69d6d6edfdf14104a
Reviewed-on: https://go-review.googlesource.com/c/go/+/324331
Trust: Dan Scales <danscales@google.com>
Run-TryBot: Dan Scales <danscales@google.com>
Reviewed-by: Keith Randall <khr@golang.org>
3 years agonet/http: fix ResponseWriter.ReadFrom with short reads
Damien Neil [Fri, 12 Mar 2021 21:53:11 +0000 (13:53 -0800)]
net/http: fix ResponseWriter.ReadFrom with short reads

CL 249238 changes ResponseWriter.ReadFrom to probe the source with
a single read of sniffLen bytes before writing the response header.
If the source returns less than sniffLen bytes without reaching
EOF, this can cause Content-Type and Content-Length detection to
fail.

Fix ResponseWrite.ReadFrom to copy a full sniffLen bytes from
the source as a probe.

Drop the explicit call to w.WriteHeader; writing the probe will
trigger a WriteHeader call.

Consistently use io.CopyBuffer; ReadFrom has already acquired a
copy buffer, so it may as well use it.

Fixes #44953.

Change-Id: Ic49305fb827a2bd7da4764b68d64b797b5157dc0
Reviewed-on: https://go-review.googlesource.com/c/go/+/301449
Trust: Damien Neil <dneil@google.com>
Run-TryBot: Damien Neil <dneil@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
3 years agoos: avoid finalizer race in windows process object
Jason A. Donenfeld [Tue, 25 May 2021 14:24:41 +0000 (16:24 +0200)]
os: avoid finalizer race in windows process object

If proc.Release is called concurrently, a handle will be double-freed.

Change-Id: I0c0c32e312e07bc8615e0bf9e9b691214444d8d5
Reviewed-on: https://go-review.googlesource.com/c/go/+/322510
Trust: Jason A. Donenfeld <Jason@zx2c4.com>
Run-TryBot: Jason A. Donenfeld <Jason@zx2c4.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
3 years agoos: terminate windows processes via handle directly
Jason A. Donenfeld [Tue, 25 May 2021 14:23:16 +0000 (16:23 +0200)]
os: terminate windows processes via handle directly

We already have a handle to the process, so use that for termination,
rather than doing a new lookup based on the PID.

Change-Id: I2958c1817f12f3dd783412baacbf629049f6956a
Reviewed-on: https://go-review.googlesource.com/c/go/+/322509
Trust: Jason A. Donenfeld <Jason@zx2c4.com>
Run-TryBot: Jason A. Donenfeld <Jason@zx2c4.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
3 years ago[dev.typeparams] test: test regabidefers in live.go
Cherry Mui [Fri, 4 Jun 2021 15:09:25 +0000 (11:09 -0400)]
[dev.typeparams] test: test regabidefers in live.go

Previously, live.go is conditioned on not using regabidefers. Now
we have regabidefers enabled by default everywhere, and we may
remove the fallback path in the near future, test that
configuration instead.

Change-Id: Idf910aee323bdd6478bc7a2062b2052d82ce003f
Reviewed-on: https://go-review.googlesource.com/c/go/+/325111
Trust: Cherry Mui <cherryyz@google.com>
Run-TryBot: Cherry Mui <cherryyz@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Than McIntosh <thanm@google.com>
3 years ago[dev.typeparams] runtime: undo go'd closure argument workaround
Cherry Mui [Thu, 3 Jun 2021 22:29:05 +0000 (18:29 -0400)]
[dev.typeparams] runtime: undo go'd closure argument workaround

In CL 298669 we added defer/go wrapping, and, as it is not
allowed for closures to escape when compiling runtime, we worked
around it by rewriting go'd closures to argumentless
non-capturing closures, so it is not a real closure and so not
needed to escape.

Previous CL removes the restriction. Now we can undo the
workaround.

Updates #40724.

Change-Id: Ic7bf129da4aee7b7fdb7157414eca943a6a27264
Reviewed-on: https://go-review.googlesource.com/c/go/+/325110
Trust: Cherry Mui <cherryyz@google.com>
Reviewed-by: Than McIntosh <thanm@google.com>
3 years ago[dev.typeparams] cmd/compile: allow go'd closure to escape when compiling runtime
Cherry Mui [Thu, 3 Jun 2021 22:25:47 +0000 (18:25 -0400)]
[dev.typeparams] cmd/compile: allow go'd closure to escape when compiling runtime

When compiling runtime, we don't allow closures to escape,
because we don't want (implicit) allocations to occur when it is
not okay to allocate (e.g. in the allocator itself). However, for
go statement, it already allocates a new goroutine anyway. It is
okay to allocate the closure. Allow it.

Also include the closure's name when reporting error.

Updates #40724.

Change-Id: Id7574ed17cc27709609a059c4eaa67ba1c4436dc
Reviewed-on: https://go-review.googlesource.com/c/go/+/325109
Trust: Cherry Mui <cherryyz@google.com>
Reviewed-by: Than McIntosh <thanm@google.com>
3 years ago[dev.typeparams] cmd/compile: export/import of recursive generic types.
Dan Scales [Mon, 17 May 2021 22:00:39 +0000 (15:00 -0700)]
[dev.typeparams] cmd/compile: export/import of recursive generic types.

Deal with export/import of recursive generic types. This includes
typeparams which have bounds that reference the typeparam.

There are three main changes:

  - Change export/import of typeparams to have an implicit "declaration"
    (doDecl). We need to do a declaration of typeparams (via the
    typeparam's package and unique name), because it may be referenced
    within its bound during its own definition.

  - We delay most of the processing of the Instantiate call until we
    finish the creation of the top-most type (similar to the way we
    delay CheckSize). This is because we can't do the full instantiation
    properly until the base type is fully defined (with methods). The
    functions delayDoInst() and resumeDoInst() delay and resume the
    processing of the instantiations.

  - To do the full needed type substitutions for type instantiations
    during import, I had to separate out the type subster in stencil.go
    and move it to subr.go in the typecheck package. The subster in
    stencil.go now does node substitution and makes use of the type
    subster to do type substitutions.

Notable other changes:
 - In types/builtins.go, put the newly defined typeparam for a union type
   (related to use of real/imag, etc.) in the current package, rather
   than the builtin package, so exports/imports work properly.

 - In types2, allowed NewTypeParam() to be called with a nil bound, and
   allow setting the bound later. (Needed to import a typeparam whose
   bound refers to the typeparam itself.)

 - During import of typeparams in types2 (importer/import.go), we need
   to keep an index of the typeparams by their package and unique name
   (with id). Use a new map typParamIndex[] for that. Again, this is
   needed to deal with typeparams whose bounds refer to the typeparam
   itself.

 - Added several new tests absdiffimp.go and orderedmapsimp.go. Some of
   the orderemapsimp tests are commented out for now, because there are
   some issues with closures inside instantiations (relating to unexported
   names of closure structs).

 - Renamed some typeparams in test value.go to make them all T (to make
   typeparam uniqueness is working fine).

Change-Id: Ib47ed9471c19ee8e9fbb34e8506907dad3021e5a
Reviewed-on: https://go-review.googlesource.com/c/go/+/323029
Trust: Dan Scales <danscales@google.com>
Trust: Robert Griesemer <gri@golang.org>
Reviewed-by: Robert Griesemer <gri@golang.org>
3 years agosyscall: regenerate zsyscall_windows.go
Bryan C. Mills [Fri, 4 Jun 2021 14:56:06 +0000 (10:56 -0400)]
syscall: regenerate zsyscall_windows.go

The declaration order in CL 319310 does not match what the generator
produces from scratch. That currently causes
cmd/internal/moddeps.TestAllDependencies to fail, since it is
explicitly checking for that kind of skew.

Updates #45914

Change-Id: If2a9cabc3d54e21deba7cb438fa364df205f38ac
Reviewed-on: https://go-review.googlesource.com/c/go/+/325112
Trust: Bryan C. Mills <bcmills@google.com>
Trust: Jason A. Donenfeld <Jason@zx2c4.com>
Reviewed-by: Jason A. Donenfeld <Jason@zx2c4.com>
Run-TryBot: Bryan C. Mills <bcmills@google.com>
TryBot-Result: Go Bot <gobot@golang.org>

3 years agodoc/go1.17: document testing changes for Go 1.17
Aaron Sheah [Thu, 3 Jun 2021 17:43:36 +0000 (17:43 +0000)]
doc/go1.17: document testing changes for Go 1.17

For #44513. Fixes #46024

Change-Id: Icf3877d1fcd67448fbc79a0ce3db3f319ad4a0e9
GitHub-Last-Rev: 8c015935c2e376134d81aa577bffdca7fc03170d
GitHub-Pull-Request: golang/go#46324
Reviewed-on: https://go-review.googlesource.com/c/go/+/322109
Reviewed-by: Heschi Kreinick <heschi@google.com>
Trust: Jeremy Faller <jeremy@golang.org>
Trust: Dmitri Shuralyov <dmitshur@golang.org>

3 years ago[dev.typeparams] go/types: move struct checking into separate file
Rob Findley [Thu, 3 Jun 2021 15:10:08 +0000 (11:10 -0400)]
[dev.typeparams] go/types: move struct checking into separate file

This is a port of CL 321589 to go/types. Specifically, the same checker
methods were moved.

Change-Id: If07d96faa77d2f9409d8895f970149c42cbfe440
Reviewed-on: https://go-review.googlesource.com/c/go/+/324753
Trust: Robert Findley <rfindley@google.com>
Run-TryBot: Robert Findley <rfindley@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Robert Griesemer <gri@golang.org>
3 years ago[dev.typeparams] go/types: move interface checking into separate file
Rob Findley [Thu, 3 Jun 2021 15:04:50 +0000 (11:04 -0400)]
[dev.typeparams] go/types: move interface checking into separate file

This is a port of CL 321549 to go/types. Specifically, the same checker
methods were moved.

Change-Id: I491a8c5a985d71ebb23e4b34541a557da0af0cfc
Reviewed-on: https://go-review.googlesource.com/c/go/+/324752
Trust: Robert Findley <rfindley@google.com>
Run-TryBot: Robert Findley <rfindley@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Robert Griesemer <gri@golang.org>
3 years ago[dev.typeparams] go/types: use correct type parameter list in missingMethod
Rob Findley [Thu, 3 Jun 2021 14:57:09 +0000 (10:57 -0400)]
[dev.typeparams] go/types: use correct type parameter list in missingMethod

This is a port of CL 321232 to go/types, adjusted to add a missing
comment and to remove optional support for method type params.

Fixes #46275

Change-Id: I63fcbb669e7607876a888fca89b9064568805448
Reviewed-on: https://go-review.googlesource.com/c/go/+/324751
Trust: Robert Findley <rfindley@google.com>
Run-TryBot: Robert Findley <rfindley@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Robert Griesemer <gri@golang.org>
3 years ago[dev.typeparams] go/types: better recv Var for method expressions
Rob Findley [Thu, 3 Jun 2021 14:12:37 +0000 (10:12 -0400)]
[dev.typeparams] go/types: better recv Var for method expressions

This is a port of CL 320489 to go/types, adjusted to be consistent about
named/unnamed parameters. TestEvalPos was failing without this addition.

For #46209

Change-Id: Icdf86e84ebce8ccdb7846a63b5605e360e2b8781
Reviewed-on: https://go-review.googlesource.com/c/go/+/324733
Trust: Robert Findley <rfindley@google.com>
Run-TryBot: Robert Findley <rfindley@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Robert Griesemer <gri@golang.org>
3 years ago[dev.typeparams] go/types: fix panic with nil package name
Rob Findley [Thu, 3 Jun 2021 13:52:05 +0000 (09:52 -0400)]
[dev.typeparams] go/types: fix panic with nil package name

This is a straightforward port of CL 320490 to go/types.

Change-Id: I763c806c777f926a563d8f9384764e5b3f7f083c
Reviewed-on: https://go-review.googlesource.com/c/go/+/324732
Trust: Robert Findley <rfindley@google.com>
Run-TryBot: Robert Findley <rfindley@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Robert Griesemer <gri@golang.org>
3 years ago[dev.typeparams] go/types: print "incomplete" for interfaces in debug mode only
Rob Findley [Thu, 3 Jun 2021 13:49:21 +0000 (09:49 -0400)]
[dev.typeparams] go/types: print "incomplete" for interfaces in debug mode only

This is a straightforward port of CL 320150 to go/types.

Fixes #46167

Change-Id: Id1845046f598ac4fefd68cda6a5a03b7a5fc5a4a
Reviewed-on: https://go-review.googlesource.com/c/go/+/324731
Trust: Robert Findley <rfindley@google.com>
Run-TryBot: Robert Findley <rfindley@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Robert Griesemer <gri@golang.org>
3 years ago[dev.typeparams] go/types: make TestManual work for directories
Rob Findley [Thu, 3 Jun 2021 13:45:01 +0000 (09:45 -0400)]
[dev.typeparams] go/types: make TestManual work for directories

This is a port of CL 315769 to go/types, adjusted for the additional
testPkg indirection in go/types on top of testFiles, and to remove the
colDelta argument.

Change-Id: Ieb722d77866313a01645aeec49912c16cb475462
Reviewed-on: https://go-review.googlesource.com/c/go/+/324730
Trust: Robert Findley <rfindley@google.com>
Trust: Robert Griesemer <gri@golang.org>
Run-TryBot: Robert Findley <rfindley@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Robert Griesemer <gri@golang.org>
3 years ago[dev.typeparams] go/types: implement types.Instantiate
Rob Findley [Thu, 3 Jun 2021 13:29:35 +0000 (09:29 -0400)]
[dev.typeparams] go/types: implement types.Instantiate

This is a straightforward port of CL 314773 to go/types.

Change-Id: If9e2d6d99790d694615389acbe6ccb3c8c0bd1da
Reviewed-on: https://go-review.googlesource.com/c/go/+/324729
Trust: Robert Findley <rfindley@google.com>
Run-TryBot: Robert Findley <rfindley@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Robert Griesemer <gri@golang.org>
3 years ago[dev.typeparams] go/types: rename Inferred.Targs to TArgs
Rob Findley [Wed, 19 May 2021 19:51:45 +0000 (15:51 -0400)]
[dev.typeparams] go/types: rename Inferred.Targs to TArgs

This is consistent with Named.TArgs.

Change-Id: Ib25f7ac5b7242e169c8c1701dfa407f763f26125
Reviewed-on: https://go-review.googlesource.com/c/go/+/321289
Trust: Robert Findley <rfindley@google.com>
Run-TryBot: Robert Findley <rfindley@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Robert Griesemer <gri@golang.org>
3 years ago[dev.typeparams] go/types: use Checker-provided type parameter IDs when possible
Rob Findley [Thu, 6 May 2021 13:08:32 +0000 (09:08 -0400)]
[dev.typeparams] go/types: use Checker-provided type parameter IDs when possible

Incrementing type parameter subscripts for each type checking pass is
distracting for an interactive program where packages are type-checked
on each keystroke.

We should perhaps hide the type parameter ID altogether, but for now at
least add a layer of indirection so that type parameters for a single
type-checked package can be stabilized.

This change should have no effect on non-generic type checking.

For #46003

Change-Id: I60d747e0a2bfb68e7d64e897eac23f609a2a4429
Reviewed-on: https://go-review.googlesource.com/c/go/+/321269
Trust: Robert Findley <rfindley@google.com>
Run-TryBot: Robert Findley <rfindley@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Robert Griesemer <gri@golang.org>
3 years agosyscall: do not pass console handles to PROC_THREAD_ATTRIBUTE_HANDLE_LIST on Windows 7
Jason A. Donenfeld [Thu, 13 May 2021 11:19:14 +0000 (13:19 +0200)]
syscall: do not pass console handles to PROC_THREAD_ATTRIBUTE_HANDLE_LIST on Windows 7

On Windows 7 (and below), console handles are not real kernel handles
but are rather userspace objects, with information passed via special
bits in the handle itself. That means they can't be passed in
PROC_THREAD_ATTRIBUTE_HANDLE_LIST, even though they can be inherited.
So, we filter the list passed to PROC_THREAD_ATTRIBUTE_HANDLE_LIST to
not have any console handles on Windows 7. At the same time, it turns
out that the presence of a NULL handle in the list is enough to render
PROC_THREAD_ATTRIBUTE_HANDLE_LIST completely useless, so filter these
out too. Console handles also can't be duplicated into parent processes,
as inhertance always happens from the present process, so duplicate
always into the present process even when a parent process is specified.

Fixes #45914.

Change-Id: I70b4ff4874dbf0507d9ec9278f63b9b4dd4f1999
Reviewed-on: https://go-review.googlesource.com/c/go/+/319310
Trust: Jason A. Donenfeld <Jason@zx2c4.com>
Trust: Alex Brainman <alex.brainman@gmail.com>
Run-TryBot: Jason A. Donenfeld <Jason@zx2c4.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Alex Brainman <alex.brainman@gmail.com>
3 years ago[dev.typeparams] cmd/compile: lazy import resolution for types2
Matthew Dempsky [Sat, 29 May 2021 22:35:18 +0000 (15:35 -0700)]
[dev.typeparams] cmd/compile: lazy import resolution for types2

This CL adds three new functions to the types2 API to support lazy
import resolution:

1. A new Scope.InsertLazy method to allow recording that Objects exist
in a particular Scope (in particular, package scopes) without having
to yet fully construct those objects. Instead, types2 will call the
provided `resolve` function if/when the object is actually needed.

2. Similarly, a new NewTypeNameLazy function to create TypeName
objects without yet instantiating their underlying Named
instance.

3. Finally, an InstantiateLazy method, that allows creating type
instances without requiring any of the types to be expanded right
away. Importantly, this requires providing a types2.Checker argument
to handle recursive types correctly.

The APIs as-is are a bit clumsy (esp. NewTypeNameLazy), but seem to
work well for cmd/compile's needs. In particular, they simplify some
of the complexities of handling recursive type definitions within the
importer.

Also, the current prototype is a bit fragile. It uses sync.Once to
manage concurrent lazy resolution, which is frustrating to debug in
the presence of reentrancy issues. It also means the importer needs to
deal with concurrency as well. These aren't issues for types2 though
as cmd/compile only walks the type-checked AST sequentially.

Finally, it looks like some of the details of lazy type names are
similar to the lazy "instance" stuff used for generics, so maybe
there's opportunity for unifying them under a more general (but still
internal) lazy type mechanism.

I had originally intended for this CL to also update the types2
importer, but (1) it doesn't have access to the types2.Checker
instance needed to call InstantiateLazy, and (2) it creates a new
TypeName/TypeParam at each use rather than reusing them, which
evidently works with types2.Instantiate but not
types2.(*Checker).instantiate (i.e., InstantiateLazy). I spent a while
trying to fix these issues, but kept running into more subtle
issues. Instead, I've included my WIP "unified IR" CL as a followup CL
that demonstrates these Lazy methods (see noder/reader2.go).

Updates #46449.

Change-Id: I4d1e8e649f6325a11790d25fd90c39fa07c8d41d
Reviewed-on: https://go-review.googlesource.com/c/go/+/323569
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
Trust: Matthew Dempsky <mdempsky@google.com>
Trust: Robert Griesemer <gri@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Robert Griesemer <gri@golang.org>
3 years agocmd/compile,go/types: restrict use of unsafe.{Add,Slice} to go1.17 or newer
Matthew Dempsky [Wed, 2 Jun 2021 19:22:50 +0000 (12:22 -0700)]
cmd/compile,go/types: restrict use of unsafe.{Add,Slice} to go1.17 or newer

This CL updates cmd/compile (including types2) and go/types to report
errors about using unsafe.Add and unsafe.Slice when language
compatibility is set to Go 1.16 or older.

Fixes #46525.

Change-Id: I1bfe025a672d9f4b929f443064ad1effd38d0363
Reviewed-on: https://go-review.googlesource.com/c/go/+/324369
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
Reviewed-by: Robert Findley <rfindley@google.com>
Reviewed-by: Robert Griesemer <gri@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>
Trust: Matthew Dempsky <mdempsky@google.com>

3 years agocmd/compile: remove spurious ir.Dump
Matthew Dempsky [Fri, 4 Jun 2021 01:00:53 +0000 (18:00 -0700)]
cmd/compile: remove spurious ir.Dump

This ir.Dump call is a debugging artifact introduced in
golang.org/cl/274103, which should never be printed for valid,
non-generic code, but evidently can now sometimes appear due to how
the parser handles invalid syntax.

The parser should probably not recognize "x[2]" as a type expression
in non-generics mode, but also probably we shouldn't try noding after
reporting syntax errors. Either way, this diagnostic has outlived its
usefulness, and noder's days are numbered anyway, so we might as well
just remove it to save end users any confusion.

Updates #46558.

Change-Id: Ib68502ef834d610b883c2f2bb11d9b385bc66e37
Reviewed-on: https://go-review.googlesource.com/c/go/+/324991
Trust: Matthew Dempsky <mdempsky@google.com>
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
Reviewed-by: Robert Griesemer <gri@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>

3 years ago[dev.typeparams] internal/buildcfg: turn on register ABI by default on ARM64
Cherry Mui [Thu, 3 Jun 2021 14:30:47 +0000 (10:30 -0400)]
[dev.typeparams] internal/buildcfg: turn on register ABI by default on ARM64

This CL enables all regabi experiments on ARM64 by default.
regabiwrappers and regabireflect are enabled in the previous CLs.
regabidefer is already enabled everywhere. regabig is no-op on
ARM64 as it already has a G register. regabiargs is enabled in
this CL.

Go1 benchmarks results (GOEXPERIMENT=regabi vs. none, on macOS/ARM64):

name                     old time/op    new time/op     delta
BinaryTree17-8              1.20s ± 1%      1.02s ± 0%  -15.08%  (p=0.000 n=9+9)
Fannkuch11-8                1.55s ± 0%      1.57s ± 0%   +1.53%  (p=0.000 n=9+8)
FmtFprintfEmpty-8          22.5ns ± 3%     14.7ns ± 1%  -34.47%  (p=0.000 n=10+8)
FmtFprintfString-8         38.4ns ± 0%     28.8ns ± 0%  -24.99%  (p=0.000 n=9+9)
FmtFprintfInt-8            38.7ns ± 2%     34.5ns ± 0%  -10.79%  (p=0.000 n=10+7)
FmtFprintfIntInt-8         61.1ns ± 1%     57.9ns ± 0%   -5.23%  (p=0.000 n=10+8)
FmtFprintfPrefixedInt-8    69.9ns ± 0%     64.4ns ± 0%   -7.78%  (p=0.000 n=8+8)
FmtFprintfFloat-8           106ns ± 0%       76ns ± 0%  -28.12%  (p=0.000 n=7+10)
FmtManyArgs-8               273ns ± 0%      236ns ± 1%  -13.57%  (p=0.000 n=9+10)
GobDecode-8                3.09ms ± 1%     2.02ms ± 0%  -34.70%  (p=0.000 n=9+10)
GobEncode-8                2.45ms ± 1%     1.44ms ± 1%  -41.26%  (p=0.000 n=10+10)
Gzip-8                      128ms ± 0%      124ms ± 0%   -2.89%  (p=0.000 n=7+8)
Gunzip-8                   23.6ms ± 1%     19.8ms ± 0%  -16.15%  (p=0.000 n=10+9)
HTTPClientServer-8         27.4µs ± 1%     26.3µs ± 0%   -4.05%  (p=0.000 n=10+10)
JSONEncode-8               4.47ms ± 1%     3.45ms ± 1%  -22.73%  (p=0.000 n=10+9)
JSONDecode-8               21.5ms ± 0%     17.2ms ± 0%  -19.78%  (p=0.000 n=9+9)
Mandelbrot200-8            2.33ms ± 1%     2.33ms ± 1%     ~     (p=0.842 n=9+10)
GoParse-8                  1.62ms ± 1%     1.32ms ± 1%  -18.67%  (p=0.000 n=10+10)
RegexpMatchEasy0_32-8      33.1ns ± 0%     26.3ns ± 0%  -20.50%  (p=0.000 n=8+10)
RegexpMatchEasy0_1K-8       121ns ± 6%      121ns ± 8%     ~     (p=0.926 n=10+10)
RegexpMatchEasy1_32-8      31.4ns ± 0%     24.7ns ± 0%  -21.50%  (p=0.000 n=9+10)
RegexpMatchEasy1_1K-8       177ns ± 0%      140ns ± 0%  -20.70%  (p=0.000 n=10+9)
RegexpMatchMedium_32-8     3.02ns ± 3%     2.12ns ± 0%  -29.73%  (p=0.000 n=10+10)
RegexpMatchMedium_1K-8     19.8µs ± 2%     17.1µs ± 0%  -13.50%  (p=0.000 n=9+9)
RegexpMatchHard_32-8        940ns ± 0%      872ns ± 0%   -7.20%  (p=0.000 n=9+8)
RegexpMatchHard_1K-8       28.5µs ± 1%     26.5µs ± 0%   -7.06%  (p=0.000 n=10+10)
Revcomp-8                   186ms ± 1%      179ms ± 1%   -3.66%  (p=0.000 n=10+10)
Template-8                 30.3ms ± 0%     22.3ms ± 0%  -26.58%  (p=0.000 n=8+9)
TimeParse-8                 133ns ± 0%      117ns ± 0%  -12.40%  (p=0.000 n=10+10)
TimeFormat-8                176ns ± 0%      141ns ± 0%  -19.92%  (p=0.000 n=8+9)
[Geo mean]                 21.4µs          17.8µs       -16.81%

name                     old speed      new speed       delta
GobDecode-8               249MB/s ± 1%    381MB/s ± 0%  +53.13%  (p=0.000 n=9+10)
GobEncode-8               314MB/s ± 1%    534MB/s ± 1%  +70.25%  (p=0.000 n=10+10)
Gzip-8                    152MB/s ± 0%    156MB/s ± 0%   +2.97%  (p=0.000 n=7+8)
Gunzip-8                  822MB/s ± 1%    981MB/s ± 0%  +19.26%  (p=0.000 n=10+9)
JSONEncode-8              434MB/s ± 1%    562MB/s ± 1%  +29.41%  (p=0.000 n=10+9)
JSONDecode-8             90.3MB/s ± 0%  112.5MB/s ± 0%  +24.66%  (p=0.000 n=9+9)
GoParse-8                35.7MB/s ± 1%   43.9MB/s ± 1%  +22.96%  (p=0.000 n=10+10)
RegexpMatchEasy0_32-8     967MB/s ± 0%   1216MB/s ± 0%  +25.78%  (p=0.000 n=8+10)
RegexpMatchEasy0_1K-8    8.46GB/s ± 6%   8.45GB/s ± 7%     ~     (p=0.912 n=10+10)
RegexpMatchEasy1_32-8    1.02GB/s ± 0%   1.30GB/s ± 0%  +27.40%  (p=0.000 n=9+10)
RegexpMatchEasy1_1K-8    5.78GB/s ± 0%   7.29GB/s ± 0%  +26.10%  (p=0.000 n=10+9)
RegexpMatchMedium_32-8    331MB/s ± 2%    471MB/s ± 0%  +42.29%  (p=0.000 n=10+10)
RegexpMatchMedium_1K-8   51.7MB/s ± 2%   59.8MB/s ± 0%  +15.60%  (p=0.000 n=9+9)
RegexpMatchHard_32-8     34.0MB/s ± 0%   36.7MB/s ± 0%   +7.75%  (p=0.000 n=9+8)
RegexpMatchHard_1K-8     35.9MB/s ± 1%   38.6MB/s ± 0%   +7.59%  (p=0.000 n=10+10)
Revcomp-8                1.37GB/s ± 1%   1.42GB/s ± 1%   +3.79%  (p=0.000 n=10+10)
Template-8               64.0MB/s ± 0%   87.1MB/s ± 0%  +36.20%  (p=0.000 n=8+9)
[Geo mean]                299MB/s         368MB/s       +23.16%

Binary sizes:
                old            new
hello           1180994        1162626      -1.6%
cmd/compile     23455858       22833970     -2.7%
cmd/link        6425010        6332978      -1.4%

Text sizes:
                old            new
hello           458752         425984       -7.1%
cmd/compile     10190848       9355264      -8.2%
cmd/link        2621440        2441216      -6.9%

Change-Id: I52c10c11bb8fe5952b7043f9dbf09573ef71d2b0
Reviewed-on: https://go-review.googlesource.com/c/go/+/324890
Trust: Cherry Mui <cherryyz@google.com>
Run-TryBot: Cherry Mui <cherryyz@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Michael Knyszek <mknyszek@google.com>
Reviewed-by: Than McIntosh <thanm@google.com>
Reviewed-by: Jeremy Faller <jeremy@golang.org>
3 years ago[dev.typeparams] internal/buildcfg: turn on regabireflect by default on ARM64
Cherry Mui [Thu, 3 Jun 2021 19:18:36 +0000 (15:18 -0400)]
[dev.typeparams] internal/buildcfg: turn on regabireflect by default on ARM64

Change-Id: I4a0a093b07a287cc3a3e0ee939e7ee82d8e9b1aa
Reviewed-on: https://go-review.googlesource.com/c/go/+/324889
Trust: Cherry Mui <cherryyz@google.com>
Run-TryBot: Cherry Mui <cherryyz@google.com>
Reviewed-by: Michael Knyszek <mknyszek@google.com>
Reviewed-by: Than McIntosh <thanm@google.com>
TryBot-Result: Go Bot <gobot@golang.org>

3 years ago[dev.typeparams] cmd/compile: allow nil Syms in Sym.Less
Matthew Dempsky [Thu, 3 Jun 2021 20:05:22 +0000 (13:05 -0700)]
[dev.typeparams] cmd/compile: allow nil Syms in Sym.Less

Allows sorting interfaces that contain embedded anonymous types.

Fixes #46556.

Change-Id: If19afa1d62432323b2e98957087867afbf3f9097
Reviewed-on: https://go-review.googlesource.com/c/go/+/324812
Trust: Matthew Dempsky <mdempsky@google.com>
Trust: Dan Scales <danscales@google.com>
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Dan Scales <danscales@google.com>
3 years ago[dev.typeparams] internal/buildcfg: turn on regabiwrappers by default on ARM64
Cherry Mui [Thu, 3 Jun 2021 19:17:51 +0000 (15:17 -0400)]
[dev.typeparams] internal/buildcfg: turn on regabiwrappers by default on ARM64

Change-Id: I8db0a797a745630ec35af3e56406fcb250ea59fe
Reviewed-on: https://go-review.googlesource.com/c/go/+/324768
Trust: Cherry Mui <cherryyz@google.com>
Run-TryBot: Cherry Mui <cherryyz@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Michael Knyszek <mknyszek@google.com>
Reviewed-by: Than McIntosh <thanm@google.com>
3 years ago[dev.typeparams] runtime: crash the GC at clobberdead pointer on ARM64
Cherry Mui [Thu, 3 Jun 2021 16:26:18 +0000 (12:26 -0400)]
[dev.typeparams] runtime: crash the GC at clobberdead pointer on ARM64

Extend CL 310330 to ARM64, which now has clobberdead mode
implemented in the compiler.

Change-Id: I07f6951d81a0797ef7a74e48b79db5cea2bf876b
Reviewed-on: https://go-review.googlesource.com/c/go/+/324766
Trust: Cherry Mui <cherryyz@google.com>
Run-TryBot: Cherry Mui <cherryyz@google.com>
Reviewed-by: Michael Knyszek <mknyszek@google.com>
TryBot-Result: Go Bot <gobot@golang.org>

3 years ago[dev.typeparams] cmd/compile: implement clobberdead mode on ARM64
Cherry Mui [Thu, 3 Jun 2021 16:23:14 +0000 (12:23 -0400)]
[dev.typeparams] cmd/compile: implement clobberdead mode on ARM64

For debugging.

Change-Id: I5875ccd2413b8ffd2ec97a0ace66b5cae7893b24
Reviewed-on: https://go-review.googlesource.com/c/go/+/324765
Trust: Cherry Mui <cherryyz@google.com>
Run-TryBot: Cherry Mui <cherryyz@google.com>
Reviewed-by: Than McIntosh <thanm@google.com>
TryBot-Result: Go Bot <gobot@golang.org>

3 years ago[dev.typeparams] cmd/compile: properly copy tilde value for unions in types2-to-types...
Dan Scales [Thu, 3 Jun 2021 17:59:35 +0000 (10:59 -0700)]
[dev.typeparams] cmd/compile: properly copy tilde value for unions in types2-to-types1 conversion

Change-Id: I2211020141886b348cddf9e33ab31b71c8478987
Reviewed-on: https://go-review.googlesource.com/c/go/+/324811
Run-TryBot: Dan Scales <danscales@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Robert Griesemer <gri@golang.org>
Trust: Dan Scales <danscales@google.com>

3 years ago[dev.typeparams] reflect: guard abi_test.go with regabiargs build tag
Cherry Mui [Thu, 3 Jun 2021 19:23:22 +0000 (15:23 -0400)]
[dev.typeparams] reflect: guard abi_test.go with regabiargs build tag

The test in abi_test.go relies on the compiler to generate
register-ABI calls using a magic name. As of CL 300150 the name
loses its magic. Guard it with regabiargs for the use of
register-ABI calls.

Change-Id: Ib8b3c24f71ea5161d607c9becfb3027ceee40ac1
Reviewed-on: https://go-review.googlesource.com/c/go/+/324767
Trust: Cherry Mui <cherryyz@google.com>
Run-TryBot: Cherry Mui <cherryyz@google.com>
Reviewed-by: Michael Knyszek <mknyszek@google.com>
TryBot-Result: Go Bot <gobot@golang.org>

3 years agocmd/link: use correct alignment in PE DWARF sections
Than McIntosh [Thu, 3 Jun 2021 18:50:10 +0000 (14:50 -0400)]
cmd/link: use correct alignment in PE DWARF sections

Set the correct section flags to insure that .debug_* sections are
using 1-byte alignment instead of the default. This seems to be
important for later versions of LLVM-mingw on windows (shows up on the
windows/arm64 builder).

Updates #46406.

Change-Id: I023d5208374f867552ba68b45011f7990159868f
Reviewed-on: https://go-review.googlesource.com/c/go/+/324763
Trust: Than McIntosh <thanm@google.com>
Reviewed-by: Cherry Mui <cherryyz@google.com>
Run-TryBot: Than McIntosh <thanm@google.com>
TryBot-Result: Go Bot <gobot@golang.org>

3 years agoruntime: avoid gp.lockedm race in exitsyscall0
Michael Pratt [Wed, 2 Jun 2021 21:44:43 +0000 (17:44 -0400)]
runtime: avoid gp.lockedm race in exitsyscall0

Following https://golang.org/cl/291329, exitsyscall0 accesses gp.lockedm
after releasing gp to the global runq. This creates a race window where
another M may schedule the (unlocked) G, which subsequently calls
LockOSThread, setting gp.lockedm and thus causing exitsyscall0 to think
it should call stoplockedm.

Avoid this race by checking if gp is locked before releasing it to the
global runq.

Fixes #46524

Change-Id: I3acdaf09e7a2178725adbe61e985130e9ebd0680
Reviewed-on: https://go-review.googlesource.com/c/go/+/324350
Trust: Michael Pratt <mpratt@google.com>
Run-TryBot: Michael Pratt <mpratt@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Reviewed-by: Michael Knyszek <mknyszek@google.com>
3 years ago[dev.typeparams] runtime: use ABIInternal callbackWrap in callbackasm1 on ARM64
Cherry Mui [Wed, 2 Jun 2021 23:52:39 +0000 (19:52 -0400)]
[dev.typeparams] runtime: use ABIInternal callbackWrap in callbackasm1 on ARM64

On Windows/ARM64, callbackasm1 calls callbackWrap via cgocallback.
cgocallback uses ABIInternal calling convention to call the
function. Pass the ABIInternal entry point to cgocallback.

Change-Id: I79d21b77525f6ac8dd50d34f4f304749419b2ad4
Reviewed-on: https://go-review.googlesource.com/c/go/+/324735
Trust: Cherry Mui <cherryyz@google.com>
Run-TryBot: Cherry Mui <cherryyz@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Michael Knyszek <mknyszek@google.com>
3 years ago[dev.typeparams] runtime: call cgocallbackg indirectly on ARM64
Cherry Mui [Wed, 2 Jun 2021 23:43:28 +0000 (19:43 -0400)]
[dev.typeparams] runtime: call cgocallbackg indirectly on ARM64

This is CL 312669, for ARM64.

cgocallback calls cgocallbackg after switching the stack. Call it
indirectly to bypass the linker's nosplit check. In particular,
this avoids a nosplit stack overflow on Windows when register ABI
is enabled.

Change-Id: I7054a750fb0ec2579d46004f94b46b6f7b9e3a21
Reviewed-on: https://go-review.googlesource.com/c/go/+/324734
Trust: Cherry Mui <cherryyz@google.com>
Run-TryBot: Cherry Mui <cherryyz@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Michael Knyszek <mknyszek@google.com>
3 years ago[dev.typeparams] runtime, internal/bytealg: port performance-critical functions to...
Cherry Mui [Wed, 2 Jun 2021 21:30:58 +0000 (17:30 -0400)]
[dev.typeparams] runtime, internal/bytealg: port performance-critical functions to register ABI on ARM64

This CL ports a few performance-critical assembly functions to use
register arguments directly. This is similar to CL 308931 and
CL 310184.

Change-Id: I6e30dfff17f76b8578ce8cfd51de21b66610fdb0
Reviewed-on: https://go-review.googlesource.com/c/go/+/324400
Trust: Cherry Mui <cherryyz@google.com>
Run-TryBot: Cherry Mui <cherryyz@google.com>
Reviewed-by: Than McIntosh <thanm@google.com>
Reviewed-by: Michael Knyszek <mknyszek@google.com>
TryBot-Result: Go Bot <gobot@golang.org>

3 years ago[dev.typeparams] test: update all the typeparam tests to use the new union/tilde...
Dan Scales [Thu, 3 Jun 2021 01:12:14 +0000 (18:12 -0700)]
[dev.typeparams] test: update all the typeparam tests to use the new union/tilde syntax

Did a mix of tilde and non-tilde usage. Tilde notation is not quite
fully functional, so no tests are currently trying to distinguish
(fail/not fail) based on tilde usage.

Change-Id: Ib50cec2fc0684f9d9f3561c889fd44c7a7af458c
Reviewed-on: https://go-review.googlesource.com/c/go/+/324572
Trust: Dan Scales <danscales@google.com>
Trust: Robert Griesemer <gri@golang.org>
Run-TryBot: Dan Scales <danscales@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Robert Griesemer <gri@golang.org>
3 years ago[dev.typeparams] internal/bytealg: call memeqbody directly in memequal_varlen on...
Cherry Mui [Wed, 2 Jun 2021 22:00:12 +0000 (18:00 -0400)]
[dev.typeparams] internal/bytealg: call memeqbody directly in memequal_varlen on ARM64

Currently, memequal_varlen opens up a frame and call memequal,
which then tail-calls memeqbody. This CL changes memequal_varlen
tail-calls memeqbody directly.

This makes it simpler to switch to the register ABI in the next
CL.

Change-Id: Ia1367c0abb7f4755fe736c404411793fb9e5c04f
Reviewed-on: https://go-review.googlesource.com/c/go/+/324399
Trust: Cherry Mui <cherryyz@google.com>
Run-TryBot: Cherry Mui <cherryyz@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Michael Knyszek <mknyszek@google.com>
3 years ago[dev.typeparams] test: adjust codegen test for register ABI on ARM64
Cherry Mui [Tue, 1 Jun 2021 23:29:24 +0000 (19:29 -0400)]
[dev.typeparams] test: adjust codegen test for register ABI on ARM64

In codegen/arithmetic.go, previously there are MOVD's that match
for loads of arguments. With register ABI there are no more such
loads. Remove the MOVD matches.

Change-Id: I920ee2629c8c04d454f13a0c08e283d3528d9a64
Reviewed-on: https://go-review.googlesource.com/c/go/+/324251
Trust: Cherry Mui <cherryyz@google.com>
Run-TryBot: Cherry Mui <cherryyz@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Than McIntosh <thanm@google.com>
3 years ago[dev.typeparams] cmd/internal/goobj: add duffzero/duffcopy to builtin list
Cherry Mui [Tue, 1 Jun 2021 23:18:25 +0000 (19:18 -0400)]
[dev.typeparams] cmd/internal/goobj: add duffzero/duffcopy to builtin list

duffzero and duffcopy are commonly referenced functions. Add them
to builtin list, so they are referenced by index, not by name.

Also change gcWriteBarrier to ABIInternal, which is changed in
CL 266638.

Regenerate the file.

Change-Id: If8550d9ed300ac2be930a7c58657a9cf1933ac1d
Reviewed-on: https://go-review.googlesource.com/c/go/+/324250
Trust: Cherry Mui <cherryyz@google.com>
Run-TryBot: Cherry Mui <cherryyz@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Than McIntosh <thanm@google.com>
3 years ago[dev.typeparams] cmd/link: take function address in assembly in TestFuncAlign
Cherry Mui [Tue, 1 Jun 2021 23:16:33 +0000 (19:16 -0400)]
[dev.typeparams] cmd/link: take function address in assembly in TestFuncAlign

In TestFuncAlign we want to get the address of an assembly
function. Take the address in assembly, so we get the actual
function's address, not the wrapper's.

Change-Id: Idc1fe2c8426562c70f8f7d6e489584ef059bc556
Reviewed-on: https://go-review.googlesource.com/c/go/+/324249
Trust: Cherry Mui <cherryyz@google.com>
Run-TryBot: Cherry Mui <cherryyz@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Than McIntosh <thanm@google.com>
3 years ago[dev.typeparams] cmd/compile/internal/types2: add Config.AllowTypeLists to control...
Robert Griesemer [Thu, 3 Jun 2021 00:50:47 +0000 (17:50 -0700)]
[dev.typeparams] cmd/compile/internal/types2: add Config.AllowTypeLists to control type list handling

Eventually the flag should not be set anymore, but we set it in the
compiler until all tests have been converted.

Also, convert some more types2 tests to use the type set notation.

Change-Id: I616599a3473451ab75d67272016b2bd3de6835af
Reviewed-on: https://go-review.googlesource.com/c/go/+/324571
Trust: Robert Griesemer <gri@golang.org>
Reviewed-by: Robert Findley <rfindley@google.com>
3 years ago[dev.typeparams] cmd/compile/internal/types2: disallow ~T where T is a defined type...
Robert Griesemer [Thu, 3 Jun 2021 00:05:37 +0000 (17:05 -0700)]
[dev.typeparams] cmd/compile/internal/types2: disallow ~T where T is a defined type or an interface

Change-Id: I35f6f43db00d56847da48320308f2fcfff924738
Reviewed-on: https://go-review.googlesource.com/c/go/+/324570
Trust: Robert Griesemer <gri@golang.org>
Run-TryBot: Robert Griesemer <gri@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Robert Findley <rfindley@google.com>
3 years ago[dev.typeparams] cmd/compile/internal/types2: convert testdata/check tests to type...
Robert Griesemer [Wed, 2 Jun 2021 23:12:25 +0000 (16:12 -0700)]
[dev.typeparams] cmd/compile/internal/types2: convert testdata/check tests to type set sytax

Change-Id: I0c2dda10ba7cb40330545fd10fbacb8c84f66a2d
Reviewed-on: https://go-review.googlesource.com/c/go/+/324569
Trust: Robert Griesemer <gri@golang.org>
Reviewed-by: Robert Findley <rfindley@google.com>
3 years ago[dev.typeparams] cmd/compile/internal/types2: convert testdata/fixedbugs tests to...
Robert Griesemer [Wed, 2 Jun 2021 22:45:43 +0000 (15:45 -0700)]
[dev.typeparams] cmd/compile/internal/types2: convert testdata/fixedbugs tests to type set sytax

Change-Id: I2ad94c71bebb93e0e3f4eba9d5199a3b3e9fa63d
Reviewed-on: https://go-review.googlesource.com/c/go/+/324530
Trust: Robert Griesemer <gri@golang.org>
Run-TryBot: Robert Griesemer <gri@golang.org>
Reviewed-by: Robert Findley <rfindley@google.com>
3 years ago[dev.typeparams] cmd/compile/internal/types2: convert testdata/examples tests to...
Robert Griesemer [Wed, 2 Jun 2021 22:36:11 +0000 (15:36 -0700)]
[dev.typeparams] cmd/compile/internal/types2: convert testdata/examples tests to type set sytax

Change-Id: Ida3837c9cbb970a2b49cd1598c6e6e9de8aa9690
Reviewed-on: https://go-review.googlesource.com/c/go/+/324529
Trust: Robert Griesemer <gri@golang.org>
Run-TryBot: Robert Griesemer <gri@golang.org>
Reviewed-by: Robert Findley <rfindley@google.com>
TryBot-Result: Go Bot <gobot@golang.org>

3 years ago[dev.typeparams] cmd/compile/internal/importer: review of gcimporter_test.go
Robert Griesemer [Wed, 2 Jun 2021 05:56:01 +0000 (22:56 -0700)]
[dev.typeparams] cmd/compile/internal/importer: review of gcimporter_test.go

This CL removes the // UNREVIEWED disclaimer at the top of the
file. This file is essentially a copy of its reviewed version
at src/go/internal/gcimporter/gcimporter_test.go with adjustments
to make it work for the compiler and types2. To see the changes
made with respect to the original, compare patchset 2 against
patchset 3.

Change-Id: Iaeb9a56a6a56f4c1d93e7bfedc5b1f1968fa6792
Reviewed-on: https://go-review.googlesource.com/c/go/+/324131
Trust: Robert Griesemer <gri@golang.org>
Run-TryBot: Robert Griesemer <gri@golang.org>
Reviewed-by: Robert Findley <rfindley@google.com>
TryBot-Result: Go Bot <gobot@golang.org>

3 years ago[dev.typeparams] cmd/compile/internal/types2: eliminate need for unpack and asUnion...
Robert Griesemer [Fri, 28 May 2021 05:09:58 +0000 (22:09 -0700)]
[dev.typeparams] cmd/compile/internal/types2: eliminate need for unpack and asUnion functions

Change-Id: Iaa75b091d52f44939330e5945305aea323ba58f4
Reviewed-on: https://go-review.googlesource.com/c/go/+/323355
Trust: Robert Griesemer <gri@golang.org>
Run-TryBot: Robert Griesemer <gri@golang.org>
Reviewed-by: Robert Findley <rfindley@google.com>
3 years ago[dev.typeparams] cmd/compile/internal/types2: clean up type set/union intersection
Robert Griesemer [Fri, 28 May 2021 02:03:16 +0000 (19:03 -0700)]
[dev.typeparams] cmd/compile/internal/types2: clean up type set/union intersection

- Eliminate the need for bottom type: This is now represented by
  an empty union (denoting the set of no types).

- Clean up type set intersection and incorporate tilde information
  in intersection operation and satisfaction tests.

- Minor cleanups along the way.

- Note: The intersection algorithm does not always compute the
        largest possible intersection. To be addressed in a follow-up CL.

Change-Id: I7fa19df5996da36a4d8f29300d30a0aa4d8a3e5c
Reviewed-on: https://go-review.googlesource.com/c/go/+/323354
Trust: Robert Griesemer <gri@golang.org>
Run-TryBot: Robert Griesemer <gri@golang.org>
Reviewed-by: Robert Findley <rfindley@google.com>
TryBot-Result: Go Bot <gobot@golang.org>

3 years ago[dev.typeparams] cmd/compile: fix export/import of constants with typeparam type
Dan Scales [Wed, 2 Jun 2021 07:03:25 +0000 (00:03 -0700)]
[dev.typeparams] cmd/compile: fix export/import of constants with typeparam type

A constant will have a TYPEPARAM type if it appears in a place where it
must match that typeparam type (e.g. in a binary operation with a
variable of that typeparam type). If so, then we must write out its
actual constant kind as well, so its constant val can be read in
properly during import.

Fixed some export/import tests which were casting some untyped constants
to avoid this problem.

Change-Id: I285ad8f1c8febbe526769c96e6b27acbd23050f0
Reviewed-on: https://go-review.googlesource.com/c/go/+/324189
Run-TryBot: Dan Scales <danscales@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
Reviewed-by: Cuong Manh Le <cuong.manhle.vn@gmail.com>
3 years ago[dev.typeparams] cmd/link: include "go build" output in test logs
Matthew Dempsky [Wed, 2 Jun 2021 00:00:26 +0000 (17:00 -0700)]
[dev.typeparams] cmd/link: include "go build" output in test logs

If running "go build" outputs anything, write it to the test log even
if the test succeeds. This makes it easier to diagnose errors within
the compiler by adding print statements and finding them in the test
log, even if the compiler exits successfully.

Change-Id: Id04716c4e1dcd9220c35ea0040ea516c1dd5237c
Reviewed-on: https://go-review.googlesource.com/c/go/+/324329
Trust: Matthew Dempsky <mdempsky@google.com>
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Cherry Mui <cherryyz@google.com>
3 years ago[dev.typeparams] cmd/compile/internal/importer: review of gcimporter.go
Robert Griesemer [Wed, 2 Jun 2021 06:01:41 +0000 (23:01 -0700)]
[dev.typeparams] cmd/compile/internal/importer: review of gcimporter.go

This CL removes the // UNREVIEWED disclaimer at the top of the
file. This file is essentially a copy of its reviewed version
at src/go/internal/gcimporter/gcimporter.go with adjustments to
make it work for the compiler and types2. To see the changes
made with respect to the original, compare patchset 1 against
patchset 2.

Change-Id: I0fd635730fb6bdee8cef1b89154f4049a6581751
Reviewed-on: https://go-review.googlesource.com/c/go/+/324132
Trust: Robert Griesemer <gri@golang.org>
Reviewed-by: Robert Findley <rfindley@google.com>
3 years agonet: don't rely on system hosts in TestCVE202133195
Roland Shoemaker [Wed, 2 Jun 2021 16:20:22 +0000 (09:20 -0700)]
net: don't rely on system hosts in TestCVE202133195

Also don't unnecessarily deref the error return.

Fixes #46504

Change-Id: I22d14ac76776f8988fa0774bdcb5fcd801ce0185
Reviewed-on: https://go-review.googlesource.com/c/go/+/324190
Trust: David Chase <drchase@google.com>
Trust: Damien Neil <dneil@google.com>
Run-TryBot: David Chase <drchase@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Damien Neil <dneil@google.com>
3 years ago[dev.typeparams] cmd/compile: sort iface fields before expansion
Matthew Dempsky [Wed, 2 Jun 2021 01:45:40 +0000 (18:45 -0700)]
[dev.typeparams] cmd/compile: sort iface fields before expansion

For toolstash -cmp compatibility with types2, we also need to sort
fields (or at least the embedded types) *before* expanding them. This
is relevant to what position information and parameter names are used
for methods when embedded interfaces have overlapping methods.

This came up in archive/zip, which has:

type fileInfoDirEntry interface {
fs.FileInfo
fs.DirEntry
}

and both of these embedded interfaces in turn have an "IsDir() bool"
method. Traditionally, cmd/compile would keep the method from
fs.FileInfo.IsDir, but with types2 it will now keep fs.DirEntry.IsDir
instead. This doesn't affect correctness at all, but it does end up in
DWARF sometimes.

Change-Id: Iac8d6321894be335466a76b5bf8a0c1b15a3581b
Reviewed-on: https://go-review.googlesource.com/c/go/+/324330
Trust: Matthew Dempsky <mdempsky@google.com>
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Robert Griesemer <gri@golang.org>
3 years ago[dev.typeparams] cmd/compile/internal/importer: review of exportdata.go
Robert Griesemer [Wed, 2 Jun 2021 06:05:25 +0000 (23:05 -0700)]
[dev.typeparams] cmd/compile/internal/importer: review of exportdata.go

This CL removes the // UNREVIEWED disclaimer at the top of the
file. This file is essentially a copy of its reviewed version
at src/go/internal/gcimporter/exportdata.go with adjustments to
make it work for the compiler and types2. To see the changes
made with respect to the original, compare patchset 1 against
patchset 2.

Change-Id: I276d898ef238afa37ec6b9605496407df36cf7d0
Reviewed-on: https://go-review.googlesource.com/c/go/+/324133
Trust: Robert Griesemer <gri@golang.org>
Reviewed-by: Robert Findley <rfindley@google.com>
3 years ago[dev.typeparams] cmd/compile/internal/importer: review of support.go
Robert Griesemer [Wed, 2 Jun 2021 05:48:20 +0000 (22:48 -0700)]
[dev.typeparams] cmd/compile/internal/importer: review of support.go

This CL removes the // UNREVIEWED disclaimer at the top of the
file. This file is essentially a copy of its reviewed version
at src/go/internal/gcimporter/support.go with adjustments to
make it work for the compiler and types2. To see the changes
made with respect to the original, compare patchset 1 against
patchset 2.

Change-Id: Icb8e7e7cac02751265c1020431018293826bad18
Reviewed-on: https://go-review.googlesource.com/c/go/+/324130
Trust: Robert Griesemer <gri@golang.org>
Reviewed-by: Robert Findley <rfindley@google.com>
3 years ago[dev.typeparams] cmd/compile/internal/types2: replace Sum type with Union type
Robert Griesemer [Wed, 26 May 2021 00:49:32 +0000 (17:49 -0700)]
[dev.typeparams] cmd/compile/internal/types2: replace Sum type with Union type

- We still mostly ignore the tilde information.

- More consistent naming: A Union term is the pair (type, tilde).
  Rename Union.terms to Union.types; the Union.types and Union.tilde
  slices make up the Union terms.

- Replace Sum.is with Union.underIs: underIs iterates through all
  union terms and calls its argument function with the underlying
  type of the term (and thus can ignore the tilde information).
  This also eliminates the need to call under in the argument
  function.

- Added Union.is for situations where we need to consider the tilde
  information for each Union term.

Change-Id: I70fcf1813e072651dc0f61d52d5555642ee762fd
Reviewed-on: https://go-review.googlesource.com/c/go/+/323274
Trust: Robert Griesemer <gri@golang.org>
Reviewed-by: Robert Findley <rfindley@google.com>
3 years ago[dev.typeparams] cmd/compile: add dictionary argument to generic functions
Keith Randall [Fri, 16 Apr 2021 21:06:50 +0000 (14:06 -0700)]
[dev.typeparams] cmd/compile: add dictionary argument to generic functions

When converting from a generic function to a concrete implementation,
add a dictionary argument to the generic function (both an actual
argument at each callsite, and a formal argument of each
implementation).

The dictionary argument comes before all other arguments (including
any receiver).

The dictionary argument is checked for validity, but is otherwise unused.
Subsequent CLs will start using the dictionary for, e.g., converting a
value of generic type to interface{}.

Import/export required adding support for LINKSYMOFFSET, which is used
by the dictionary checking code.

Change-Id: I16a7a8d23c7bd6a897e0da87c69f273be9103bd7
Reviewed-on: https://go-review.googlesource.com/c/go/+/323272
Trust: Keith Randall <khr@golang.org>
Trust: Dan Scales <danscales@google.com>
Run-TryBot: Keith Randall <khr@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Dan Scales <danscales@google.com>
3 years agoio/fs: minor corrections to Sub docs
Ian Lance Taylor [Thu, 25 Feb 2021 16:21:52 +0000 (08:21 -0800)]
io/fs: minor corrections to Sub docs

Fixes #44376

Change-Id: I9cd21adb9d4d434c3d8b8eb8af3042b70c763ea1
Reviewed-on: https://go-review.googlesource.com/c/go/+/296389
Trust: Ian Lance Taylor <iant@golang.org>
Reviewed-by: Russ Cox <rsc@golang.org>
3 years agodoc/go1.17: remove runtime section
Jeremy Faller [Wed, 2 Jun 2021 15:20:40 +0000 (11:20 -0400)]
doc/go1.17: remove runtime section

Updates #44513

Change-Id: I359d56fc3eeece3005f092cca2cb485664affc23
Reviewed-on: https://go-review.googlesource.com/c/go/+/324209
Trust: Jeremy Faller <jeremy@golang.org>
Run-TryBot: Jeremy Faller <jeremy@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Dmitri Shuralyov <dmitshur@golang.org>
Reviewed-by: Michael Knyszek <mknyszek@google.com>