]> Cypherpunks repositories - gostls13.git/log
gostls13.git
4 years agoruntime/cgo: use correct lean and mean macro
Jason A. Donenfeld [Sun, 14 Feb 2021 18:43:31 +0000 (19:43 +0100)]
runtime/cgo: use correct lean and mean macro

WIN64_LEAN_AND_MEAN is not the correct macro to use and doesn't ever
exist.

Change-Id: I32a5523cc0f7cc3f3a4d022071cf81f88db39aa9
Reviewed-on: https://go-review.googlesource.com/c/go/+/291634
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>
4 years agocmd/link: set .ctors COFF section to writable and aligned
Jason A. Donenfeld [Sun, 14 Feb 2021 18:01:39 +0000 (19:01 +0100)]
cmd/link: set .ctors COFF section to writable and aligned

Without setting these flags, LLVM's LLD ignores the .ctors section when
merging objects.

Updates #44250.
Updates #39326.
Updates #38755.
Updates #36439.
Updates #43800.

Change-Id: I8766104508f7acd832088a590ee7d68afa0d6065
Reviewed-on: https://go-review.googlesource.com/c/go/+/291633
Trust: Jason A. Donenfeld <Jason@zx2c4.com>
Run-TryBot: Jason A. Donenfeld <Jason@zx2c4.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Than McIntosh <thanm@google.com>
4 years agocmd/link: do not pass -Bsymbolic for PE DLLs
Jason A. Donenfeld [Sun, 14 Feb 2021 13:44:14 +0000 (14:44 +0100)]
cmd/link: do not pass -Bsymbolic for PE DLLs

This is only a valid option on ELF. Binutils accepts it, but LLVM
rejects it, so for Windows, it's best to just omit it.

Updates #44250.
Updates #39326.
Updates #38755.
Updates #36439.
Updates #43800.

Change-Id: Iffd2345d757f23dd737e63bd464cd412527077c4
Reviewed-on: https://go-review.googlesource.com/c/go/+/291632
Trust: Jason A. Donenfeld <Jason@zx2c4.com>
Run-TryBot: Jason A. Donenfeld <Jason@zx2c4.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Than McIntosh <thanm@google.com>
4 years agocmd/link: set SizeOfRawData rather than VirtualSize in COFF files for .bss section
Jason A. Donenfeld [Sat, 13 Feb 2021 23:04:48 +0000 (00:04 +0100)]
cmd/link: set SizeOfRawData rather than VirtualSize in COFF files for .bss section

GCC and Clang both set the SizeOfRawData field rather than the
VirtualSize field for communicating the size of the .bss section. As a
consequence, LLD does not look at VirtualSize and collapses the .bss
section into whatever is around it, resulting in runtime crashes. This
commit changes the logic so that if the requested "file size" is 0, then
the SizeOfRawData field is set rather than the VirtualSize field as the
sole length marker.

Fixes #44250.
Fixes #39326.
Updates #38755.
Updates #36439.
Updates #43800.

Change-Id: Ied89ddaa0a717fed840238244c6e4848845aeeb6
Reviewed-on: https://go-review.googlesource.com/c/go/+/291630
Trust: Jason A. Donenfeld <Jason@zx2c4.com>
Run-TryBot: Jason A. Donenfeld <Jason@zx2c4.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Than McIntosh <thanm@google.com>
4 years agocmd/internal/dwarf: minor cleanups
Than McIntosh [Wed, 6 Jan 2021 18:27:17 +0000 (13:27 -0500)]
cmd/internal/dwarf: minor cleanups

Remove a stale comment, demote PutInlinedFunc from public to private,
and remove an unused interface originally used for sorting vars.
No change in functionality.

Change-Id: I5ee1ad2b10b78b158e2223c6979bab830202db95
Reviewed-on: https://go-review.googlesource.com/c/go/+/295009
Trust: Than McIntosh <thanm@google.com>
Run-TryBot: Than McIntosh <thanm@google.com>
Reviewed-by: Jeremy Faller <jeremy@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Go Bot <gobot@golang.org>

4 years agoruntime: enable race detector on openbsd/amd64
Joel Sing [Mon, 7 Dec 2020 17:14:34 +0000 (04:14 +1100)]
runtime: enable race detector on openbsd/amd64

Now that this commit[1] has landed in LLVM the .syso file can be generated for
OpenBSD.

With the changes to src/runtime running the sample race[2] detects the data
race as expected.

Based on golang/go#39464 (https://go-review.googlesource.com/c/go/+/237057) from
Aaron Bieber <deftly@gmail.com>, however the race_openbsd_amd64.syso file has
been built on OpenBSD 6.4 and necessary changes added to race.bash.

[1] https://github.com/llvm/llvm-project/commit/fcf6ae2f070eba73074b6ec8d8281e54d29dbeeb
[2] https://golang.org/doc/articles/race_detector.html

Change-Id: Ic2479ccfa91d6b2cb4585346a11d813d96450f68
Reviewed-on: https://go-review.googlesource.com/c/go/+/275892
Trust: Joel Sing <joel@sing.id.au>
Run-TryBot: Joel Sing <joel@sing.id.au>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
4 years agointernal/poll: fix the verbose condition in splice
Andy Pan [Thu, 19 Nov 2020 09:30:27 +0000 (17:30 +0800)]
internal/poll: fix the verbose condition in splice

Change-Id: I0b433ea1a78632de20ea58c48c9be0f1fb6eb083
Reviewed-on: https://go-review.googlesource.com/c/go/+/271499
Reviewed-by: Tobias Klauser <tobias.klauser@gmail.com>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Trust: Tobias Klauser <tobias.klauser@gmail.com>
Run-TryBot: Tobias Klauser <tobias.klauser@gmail.com>

4 years agoios/fs: mention f.dir in (*subFS).fixErr godoc
Tobias Klauser [Tue, 16 Feb 2021 09:36:10 +0000 (10:36 +0100)]
ios/fs: mention f.dir in (*subFS).fixErr godoc

There is no dir parameter to (f *subFS).fixErr.

Change-Id: I49e42bac5e102cfab0d289658d9871429cfec515
Reviewed-on: https://go-review.googlesource.com/c/go/+/292389
Trust: Tobias Klauser <tobias.klauser@gmail.com>
Run-TryBot: Tobias Klauser <tobias.klauser@gmail.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
4 years agocmd/internal/obj/riscv: prevent constant loads that do not target registers
Joel Sing [Wed, 17 Feb 2021 10:03:59 +0000 (21:03 +1100)]
cmd/internal/obj/riscv: prevent constant loads that do not target registers

Check that the target of a constant load is a register and add test coverage
for this error condition. While here, rename the RISC-V testdata and tests
to be consistent with other platforms.

Change-Id: I7fd0bfcee8cf9df0597d72e65cd74a2d0bfd349a
Reviewed-on: https://go-review.googlesource.com/c/go/+/292895
Trust: Joel Sing <joel@sing.id.au>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Go Bot <gobot@golang.org>

4 years agocmd/internal/obj/riscv: clean up branch tests
Joel Sing [Fri, 15 Jan 2021 16:09:15 +0000 (03:09 +1100)]
cmd/internal/obj/riscv: clean up branch tests

Address review comments from earlier changes, which would have previously
caused unwanted conflicts.

Change-Id: If2c61ffe977d721cccf276f931825b003521fda1
Reviewed-on: https://go-review.googlesource.com/c/go/+/284116
Trust: Joel Sing <joel@sing.id.au>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Run-TryBot: Cherry Zhang <cherryyz@google.com>

4 years agoruntime: fix windows/arm signal handling assembly
Russ Cox [Mon, 22 Feb 2021 17:56:33 +0000 (12:56 -0500)]
runtime: fix windows/arm signal handling assembly

Bug introduced in CL 288799: R12 is used but not set.

Fixes windows/arm builder.

Change-Id: I015a5a83cfa3bdd23da1ffb73713623764f2f817
Reviewed-on: https://go-review.googlesource.com/c/go/+/295109
Trust: Russ Cox <rsc@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
4 years agocmd/compile: remove backend's "scratch mem" support
Cuong Manh Le [Wed, 17 Feb 2021 08:27:42 +0000 (15:27 +0700)]
cmd/compile: remove backend's "scratch mem" support

This CL rebases CL 273987 on top of master with @mdempsky's permission.

The last (only?) use for this feature was 387 support, which was
removed in golang.org/cl/258957.

Change-Id: I4f79fee8d0c336c9b6082bcd5eb6ece52c032dc0
Reviewed-on: https://go-review.googlesource.com/c/go/+/292893
Trust: Cuong Manh Le <cuong.manhle.vn@gmail.com>
Run-TryBot: Cuong Manh Le <cuong.manhle.vn@gmail.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
4 years agocmd/compile: simplify assert{E,I}2I{,2} calling conventions
Cuong Manh Le [Wed, 17 Feb 2021 08:14:21 +0000 (15:14 +0700)]
cmd/compile: simplify assert{E,I}2I{,2} calling conventions

This CL rebases CL 273694 on top of master with @mdempsky's permission.

For assertE2I and assertI2I, there's no need to pass through the
interface's data pointer: it's always going to come back unmodified.

For assertE2I2 and assertI2I2, there's no need for an extra bool
result parameter: it's redundant with testing the returned interface
value for nil.

Change-Id: Ic92d4409ad381952f875d3d74b8cf11c32702fa6
Reviewed-on: https://go-review.googlesource.com/c/go/+/292892
Trust: Cuong Manh Le <cuong.manhle.vn@gmail.com>
Run-TryBot: Cuong Manh Le <cuong.manhle.vn@gmail.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
4 years agocmd/compile: remove selectnbrecv2
Cuong Manh Le [Wed, 17 Feb 2021 04:38:14 +0000 (11:38 +0700)]
cmd/compile: remove selectnbrecv2

Previous CL did remove selectnbrecv2 in runtime, the compiler now only
call selectnbrecv, so remove this. Make this as separated CL because it
adds much of noise to git stat.

Change-Id: I06e89c823c0403e9bd66f2633409c455a46d6e79
Reviewed-on: https://go-review.googlesource.com/c/go/+/292891
Trust: Cuong Manh Le <cuong.manhle.vn@gmail.com>
Run-TryBot: Cuong Manh Le <cuong.manhle.vn@gmail.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
4 years agocmd/compile,runtime: make selectnbrecv return two values
Cuong Manh Le [Wed, 17 Feb 2021 04:36:58 +0000 (11:36 +0700)]
cmd/compile,runtime: make selectnbrecv return two values

The only different between selectnbrecv and selectnbrecv2 is the later
set the input pointer value by second return value from chanrecv.

So by making selectnbrecv return two values from chanrecv, we can get
rid of selectnbrecv2, the compiler can now call only selectnbrecv and
generate simpler code.

Change-Id: Ifaf6cf1314c4f47b06ed9606b1578319be808507
Reviewed-on: https://go-review.googlesource.com/c/go/+/292890
Trust: Cuong Manh Le <cuong.manhle.vn@gmail.com>
Run-TryBot: Cuong Manh Le <cuong.manhle.vn@gmail.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
4 years agocmd/compile: use transitive relations for slice len/cap in poset
Cuong Manh Le [Fri, 27 Nov 2020 19:46:00 +0000 (02:46 +0700)]
cmd/compile: use transitive relations for slice len/cap in poset

Currently, we keep track of slice len by mapping from slice ID to
len/cap SSA value. However, slice len/cap can have multiple SSA values,
so when updating fact table for slice len/cap, we only update in one
place.

Instead, we can take advantage of the transitive relations provided by
poset. So all duplicated slice lens are set as equal to one another.
When updating fact table for one, that fact will be reflected to all
others. The same mechanism is applied for slice cap.

Removes 15 bounds checks from std/cmd.

Fixes #42603

Change-Id: I32c07968824cc33765b1e441b3ae2c4b5f5997c3
Reviewed-on: https://go-review.googlesource.com/c/go/+/273670
Trust: Cuong Manh Le <cuong.manhle.vn@gmail.com>
Run-TryBot: Cuong Manh Le <cuong.manhle.vn@gmail.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
4 years agocmd/compile/internal/types2: review of sanitize.go
Robert Griesemer [Sat, 20 Feb 2021 04:06:24 +0000 (20:06 -0800)]
cmd/compile/internal/types2: review of sanitize.go

Remove the "// UNREVIEWED" marker and add guards (as in go/types)
to prevent data races. To see the added guards, see compare patch
sets 3 and 4. The equivalent changes for go/types were done in
https://golang.org/cl/294411.

Change-Id: Ibef07eaae400bd32bff32b102cc743580290d135
Reviewed-on: https://go-review.googlesource.com/c/go/+/294510
Trust: Robert Griesemer <gri@golang.org>
Reviewed-by: Robert Findley <rfindley@google.com>
4 years agocmd/compile: mark OpSB, OpSP as poor statement Op
Cuong Manh Le [Fri, 27 Nov 2020 04:00:29 +0000 (11:00 +0700)]
cmd/compile: mark OpSB, OpSP as poor statement Op

So that would make them last choice for a statement boundary.

This is follow up of CL 273506.

Change-Id: I0203aa0e0d95d538064c2113143c85c4fbb1e65e
Reviewed-on: https://go-review.googlesource.com/c/go/+/273666
TryBot-Result: Go Bot <gobot@golang.org>
Trust: Cuong Manh Le <cuong.manhle.vn@gmail.com>
Run-TryBot: Cuong Manh Le <cuong.manhle.vn@gmail.com>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
4 years agogo/types, cmd/compile/internal/types2: use regular type printing for unsafe.Pointer
Robert Griesemer [Tue, 23 Feb 2021 01:02:46 +0000 (17:02 -0800)]
go/types, cmd/compile/internal/types2: use regular type printing for unsafe.Pointer

Type string printing special-cased printing of unsafe.Pointer because
it's a built-in type; yet it's declared in a package like any other
imported or used-defined type (unlike built-in types such as int).

Use the same mechanism for printing unsafe.Pointer like any other
(non-basic) type. This will make it possible to use the package
Qualifier if so desired.

Fixes #44515.

Change-Id: I0dd1026f850737ecfc4bb99135cfb8e3c18be9e7
Reviewed-on: https://go-review.googlesource.com/c/go/+/295271
Trust: Robert Griesemer <gri@golang.org>
Reviewed-by: Robert Findley <rfindley@google.com>
4 years agocmd/compile/internal/types2: review of conversions.go
Robert Griesemer [Sat, 20 Feb 2021 04:00:33 +0000 (20:00 -0800)]
cmd/compile/internal/types2: review of conversions.go

The changes between (equivalent, and reviewed) go/types/conversions.go
and conversions.go can be seen by comparing patchset 1 and 2. The actual
change is just removing the "// UNREVIEWED" marker.

Change-Id: I86d20d8100ec29fe3be996b975c9b4aff01be85e
Reviewed-on: https://go-review.googlesource.com/c/go/+/294509
Trust: Robert Griesemer <gri@golang.org>
Reviewed-by: Robert Findley <rfindley@google.com>
4 years agocmd/compile/internal/types2: review of issues_test.go
Robert Griesemer [Sat, 20 Feb 2021 00:58:24 +0000 (16:58 -0800)]
cmd/compile/internal/types2: review of issues_test.go

The changes between (equivalent, and reviewed) go/types/issues_test.go
and issues_test.go can be seen by comparing patchset 1 and 3. The actual
change is just removing the "// UNREVIEWED" marker and making making
some minor code adjustments to match go/types's version more closely.

Change-Id: I26f3f700d12db69fc68161a6b0dc081a0e9cd0d4
Reviewed-on: https://go-review.googlesource.com/c/go/+/294473
Trust: Robert Griesemer <gri@golang.org>
Reviewed-by: Robert Findley <rfindley@google.com>
4 years agocmd/compile/internal/types2: enable TestIssue25627
Robert Griesemer [Sat, 20 Feb 2021 00:47:04 +0000 (16:47 -0800)]
cmd/compile/internal/types2: enable TestIssue25627

Since we have syntax.Walk, we can make this test work again.

Change-Id: I55cbde7303e5bcbe1123b6679f2ce859d377fd86
Reviewed-on: https://go-review.googlesource.com/c/go/+/294472
Trust: Robert Griesemer <gri@golang.org>
Reviewed-by: Robert Findley <rfindley@google.com>
4 years agotest: add test for findTypeLoop with symbols from other packages
Cuong Manh Le [Tue, 16 Feb 2021 18:14:19 +0000 (01:14 +0700)]
test: add test for findTypeLoop with symbols from other packages

CL 274294 improved findTypeLoop but also fixed a new found bug on master
branch. This Cl adds test cases for this.

Updates #44266

Change-Id: Ie4a07a3487758a1e4ad2f2847dcde975b10d2a77
Reviewed-on: https://go-review.googlesource.com/c/go/+/292889
Run-TryBot: Cuong Manh Le <cuong.manhle.vn@gmail.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
Trust: Cuong Manh Le <cuong.manhle.vn@gmail.com>

4 years agocmd/compile: fix import of functions of multiple nested closure
Dan Scales [Fri, 19 Feb 2021 17:40:35 +0000 (09:40 -0800)]
cmd/compile: fix import of functions of multiple nested closure

For import of functions with closures, the connections among closure
variables are constructed on-the-fly via CaptureName(). For multiple
nested closures, we need to temporarily set r.curfn to each closure we
construct, so that the processing of closure variables will be correct
for any nested closure inside that closure.

Fixes #44335

Change-Id: I34f99e2822250542528ff6b2232bf36756140868
Reviewed-on: https://go-review.googlesource.com/c/go/+/294212
Run-TryBot: Dan Scales <danscales@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Trust: Dan Scales <danscales@google.com>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
4 years agoall: faster midpoint computation in binary search
yangwenmai [Fri, 25 Dec 2020 01:36:41 +0000 (09:36 +0800)]
all: faster midpoint computation in binary search

On my machine (3.1 GHz Quad-Core Intel Core i7, macOS 10.15.7 10.15.7), go 1.15.6

benchstat:
name          old time/op  new time/op  delta
SearchInts-8  20.3ns ± 1%  16.6ns ± 6%  -18.37%  (p=0.000 n=9+10)

Change-Id: I346e5955fd6df6ce10254b22267dbc8d5a2b16c0
Reviewed-on: https://go-review.googlesource.com/c/go/+/279439
Reviewed-by: Ben Shi <powerman1st@163.com>
Reviewed-by: Robert Griesemer <gri@golang.org>
Trust: Robert Griesemer <gri@golang.org>

4 years agocmd/dist: match goexperiment.regabi tag when GOEXPERIMENT is on
Cherry Zhang [Thu, 18 Feb 2021 19:18:53 +0000 (14:18 -0500)]
cmd/dist: match goexperiment.regabi tag when GOEXPERIMENT is on

Change-Id: I5e4347dde6dcb49cd96608e4f67e54c7b3050bc2
Reviewed-on: https://go-review.googlesource.com/c/go/+/293851
Trust: Cherry Zhang <cherryyz@google.com>
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Michael Knyszek <mknyszek@google.com>
4 years agogo/parser: return ast.BadExpr for missing index operands
Rob Findley [Mon, 22 Feb 2021 20:39:00 +0000 (15:39 -0500)]
go/parser: return ast.BadExpr for missing index operands

The parser was returning the indexed operand when a slice or index or
instance expression was missing any index arguments (as in the
expression `a[]`). This can result in returning an *ast.Ident for the
LHS of the (invalid) assignment `a[] = ...` -- in this case parsing the
LHS as just `a`. Unfortunately, as the indexed operand `a` has already
been resolved, this results in a panic for duplicate resolution.

Fix this by instead returning an ast.BadExpr. This can suppress some
subsequent errors from the typechecker, but those errors may or may not
be correct anyway. Other interpretations, such as an *ast.IndexExpr with
bad or missing X, run into potential misinterpretations downstream (both
caused errors in go/types and/or gopls).

Fixes #44504

Change-Id: I5ca8bed4a1861bcc7db8898770b08937110981d4
Reviewed-on: https://go-review.googlesource.com/c/go/+/295151
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>
4 years agocmd/compile: correctly use X15 to zero frame
Cherry Zhang [Mon, 22 Feb 2021 22:43:08 +0000 (17:43 -0500)]
cmd/compile: correctly use X15 to zero frame

In CL 288093 we reserve X15 as the zero register and use that to
zero values. It only covered zeroing generated in SSA but missed
zeroing the frame generated late in the compilation. The latter
still uses X0, but now DUFFZERO expects X15, so it doesn't
actually zero the frame. Change it to use X15.

Should fix #44333.

Change-Id: I239d2b78a5f6468bc86b70aecdd294045311759f
Reviewed-on: https://go-review.googlesource.com/c/go/+/295210
Trust: Cherry Zhang <cherryyz@google.com>
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: David Chase <drchase@google.com>
4 years agocmd/compile/internal: loop opt
cui [Thu, 18 Feb 2021 17:51:24 +0000 (17:51 +0000)]
cmd/compile/internal: loop opt

Change-Id: I5fe767237b8046934e9b0f33bd3dafabdb727dd6
GitHub-Last-Rev: 94fea3d57279e8b2d62f7f6be4301698dc8841e3
GitHub-Pull-Request: golang/go#44384
Reviewed-on: https://go-review.googlesource.com/c/go/+/293809
Reviewed-by: Robert Griesemer <gri@golang.org>
Trust: Robert Griesemer <gri@golang.org>
Trust: Matthew Dempsky <mdempsky@google.com>
Run-TryBot: Robert Griesemer <gri@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>

4 years agofix typo in issue16760.go
zhengjianxun [Mon, 22 Feb 2021 13:30:03 +0000 (13:30 +0000)]
fix typo in issue16760.go

fix typo in issue16760.go, unconditinally ->  unconditionally

Change-Id: I3a04fbcb23395c562821b35bc2d81cfaec0bc1ed
GitHub-Last-Rev: 5ce52a3deb52826bc28022776c3fe3ffa7376084
GitHub-Pull-Request: golang/go#44495
Reviewed-on: https://go-review.googlesource.com/c/go/+/294969
Trust: Alberto Donizetti <alb.donizetti@gmail.com>
Reviewed-by: zhengjianxun <zhuimengshaonian04@gmail.com>
Reviewed-by: Robert Griesemer <gri@golang.org>
4 years agocmd/compile: reject some rare looping CFGs in shortcircuit
Josh Bleecher Snyder [Mon, 22 Feb 2021 19:55:14 +0000 (11:55 -0800)]
cmd/compile: reject some rare looping CFGs in shortcircuit

One CFGs that shortcircuit looks for is:

p   q
 \ /
  b
 / \
t   u

The test case creates a CFG like that in which p == t.
That caused the compiler to generate a (short-lived) invalid phi value.

Fix this with a relatively big hammer: Disallow single-length loops entirely.
This is probably overkill, but it such loops are very rare.
This doesn't change the generated code for anything in std.

It generates worse code for the test case:
It no longer compiles the entire function away.

Fixes #44465

Change-Id: Ib8cdcd6cc9d7f48b4dab253652038ace24eae152
Reviewed-on: https://go-review.googlesource.com/c/go/+/295130
Trust: Josh Bleecher Snyder <josharian@gmail.com>
Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com>
Reviewed-by: Keith Randall <khr@golang.org>
4 years agotest: add test for issue 38698
Josh Bleecher Snyder [Mon, 22 Feb 2021 19:11:51 +0000 (11:11 -0800)]
test: add test for issue 38698

It was fixed by CL 294289, for #44378.
This is a different style of test that uses
line directives instead of extremely long lines.

Fixes #38698.

Change-Id: I50a1585030978b35fffa9981d6ed96b99216dc3e
Reviewed-on: https://go-review.googlesource.com/c/go/+/295129
Trust: Josh Bleecher Snyder <josharian@gmail.com>
Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com>
Reviewed-by: Than McIntosh <thanm@google.com>
TryBot-Result: Go Bot <gobot@golang.org>

4 years agocmd/internal: cleanup ppc64 optab structure
Paul E. Murphy [Mon, 2 Nov 2020 21:05:40 +0000 (15:05 -0600)]
cmd/internal: cleanup ppc64 optab structure

This is no-functionality change to begin the process of supporting
more than 6 operands.

This rewrites the table to use named arguments, and removes default
initialized argument values.  The following sed regexes rewrote the table:

  s/{\([^,]*\),\([^,]*\),\([^,]*\),\([^,]*\),\([^,]*\),\([^,]*\),\([^,]*\),\([^,]*\)}/{as:\1,a1:\2,a2:\3,a3:\4,a4:\5,type_:\6,size:\7,param:\8}
  s/a[1-4]: C_NONE, //g
  s/, param: 0//

Change-Id: I5f4de9da75f2fb3964d625d6b4e2f1ce1e29cc47
Reviewed-on: https://go-review.googlesource.com/c/go/+/294189
Trust: Lynn Boger <laboger@linux.vnet.ibm.com>
Run-TryBot: Lynn Boger <laboger@linux.vnet.ibm.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Carlos Eduardo Seo <carlos.seo@linaro.org>
4 years agogo/internal/gccgoimporter: fix up gccgo installation test
Than McIntosh [Fri, 19 Feb 2021 15:34:55 +0000 (10:34 -0500)]
go/internal/gccgoimporter: fix up gccgo installation test

Change the TestInstallationImporter testpoint to query type
information for sort.Search instead of sort.Ints. The latter function
changed recently (1.16 timeframe), parameter "a" is now "x". A better
candidate for this sort of query is sort.Search, which has been stable
for a while.

Fixes #44425.

Change-Id: I314476eac0b0802f86f5cbce32195cab2926db83
Reviewed-on: https://go-review.googlesource.com/c/go/+/294290
Trust: Than McIntosh <thanm@google.com>
Run-TryBot: Than McIntosh <thanm@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
4 years agocmd/compile: fix panic in DWARF-gen handling obfuscated code
Than McIntosh [Fri, 19 Feb 2021 15:09:15 +0000 (10:09 -0500)]
cmd/compile: fix panic in DWARF-gen handling obfuscated code

DWARF generation uses variable source positions (file/line/col) as a
way to uniquely identify locals and parameters, as part of the process
of matching up post-optimization variables with the corresponding
pre-optimization versions (since the DWARF needs to be in terms of the
original source constructs).

This strategy can run into problems when compiling obfuscated or
machine-generated code, where you can in some circumstances wind up
with two local variables that appear to have the same name, file,
line, and column. This patch changes DWARF generation to skip over
such duplicates as opposed to issuing a fatal error (if an
obfuscation tool is in use, it is unlikely that a human being will be
able to make much sense of DWARF info in any case).

Fixes #44378.

Change-Id: I198022d184701aa9ec3dce42c005d29b72d2e321
Reviewed-on: https://go-review.googlesource.com/c/go/+/294289
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: David Chase <drchase@google.com>
Reviewed-by: Jeremy Faller <jeremy@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Trust: Than McIntosh <thanm@google.com>
Run-TryBot: Than McIntosh <thanm@google.com>

4 years agosyscall: add explicit ios build tag
Ian Lance Taylor [Sat, 20 Feb 2021 19:25:15 +0000 (11:25 -0800)]
syscall: add explicit ios build tag

This permits analysis of the syscall package by tools built with
older versions of Go that do not recognize ios as a GOOS.

Fixes #44459

Change-Id: I79cec2ffe0dbcbc2dc45a385e556dc9e62033125
Reviewed-on: https://go-review.googlesource.com/c/go/+/294634
Trust: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>

4 years agodoc/1.16: fix link to fs.FileInfo
Alberto Donizetti [Sat, 20 Feb 2021 16:53:43 +0000 (17:53 +0100)]
doc/1.16: fix link to fs.FileInfo

Updates #40700

Change-Id: Ifff622ccadaa31c0c0684c3c695aadcaa2305623
Reviewed-on: https://go-review.googlesource.com/c/go/+/294669
Trust: Alberto Donizetti <alb.donizetti@gmail.com>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
4 years agoall: go fmt std cmd (but revert vendor)
Russ Cox [Fri, 19 Feb 2021 23:35:10 +0000 (18:35 -0500)]
all: go fmt std cmd (but revert vendor)

Make all our package sources use Go 1.17 gofmt format
(adding //go:build lines).

Part of //go:build change (#41184).
See https://golang.org/design/draft-gobuild

Change-Id: Ia0534360e4957e58cd9a18429c39d0e32a6addb4
Reviewed-on: https://go-review.googlesource.com/c/go/+/294430
Trust: Russ Cox <rsc@golang.org>
Run-TryBot: Russ Cox <rsc@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Jason A. Donenfeld <Jason@zx2c4.com>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
4 years agocmd/vet: update buildtag check for //go:build lines
Russ Cox [Fri, 19 Feb 2021 14:01:32 +0000 (09:01 -0500)]
cmd/vet: update buildtag check for //go:build lines

Brings in golang.org/x/tools@2363391a
and adjusts, adds cmd/vet tests accordingly.

Part of //go:build change (#41184).
See https://golang.org/design/draft-gobuild

This brings in the new //go:build checks in cmd/vet.

Change-Id: I8a9735cc014171691012b307ec30e94c81aadfe1
Reviewed-on: https://go-review.googlesource.com/c/go/+/240609
Trust: Russ Cox <rsc@golang.org>
Run-TryBot: Russ Cox <rsc@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Jay Conrod <jayconrod@google.com>
4 years agogo/printer: canonicalize //go:build and // +build lines while formatting
Russ Cox [Fri, 26 Jun 2020 20:22:35 +0000 (16:22 -0400)]
go/printer: canonicalize //go:build and // +build lines while formatting

Part of //go:build change (#41184).
See https://golang.org/design/draft-gobuild

Gofmt and any other go/printer-using program will now:

 - move //go:build and //+build lines to the appropriate file location
 - if there's no //go:build line, add one derived from the // +build lines
 - if there is a //go:build line, recompute and replace any // +build lines
   to match what the //go:build line says

For Go 1.17.

Change-Id: Ide5cc3b4a07507ba9ed6f8b0de846e840876f49f
Reviewed-on: https://go-review.googlesource.com/c/go/+/240608
Trust: Russ Cox <rsc@golang.org>
Trust: Jay Conrod <jayconrod@google.com>
Run-TryBot: Russ Cox <rsc@golang.org>
Reviewed-by: Jay Conrod <jayconrod@google.com>
4 years agogo/build: prefer //go:build over // +build lines
Russ Cox [Mon, 22 Jun 2020 17:27:10 +0000 (13:27 -0400)]
go/build: prefer //go:build over // +build lines

Part of //go:build change (#41184).
See https://golang.org/design/draft-gobuild

- Reject files with multiple //go:build lines.
- If a file has both //go:build and // +build lines, only use the //go:build line.
- Otherwise fall back to // +build lines
- Use go/build/constraint for parsing both //go:build and // +build lines.

For Go 1.17.

Change-Id: I32e2404d8ce266230f767718dc7cc24e77b425e8
Reviewed-on: https://go-review.googlesource.com/c/go/+/240607
Trust: Russ Cox <rsc@golang.org>
Run-TryBot: Russ Cox <rsc@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Jay Conrod <jayconrod@google.com>
4 years agoruntime/pprof: disable TestMorestack on darwin/arm64
Russ Cox [Fri, 19 Feb 2021 22:01:35 +0000 (17:01 -0500)]
runtime/pprof: disable TestMorestack on darwin/arm64

Something is weird about darwin and TestMorestack,
but it is only manifesting on arm64 and race+amd64.
Disable for now.

Change-Id: I5862372fdd0b5ffae802fdefb65b2aa04e266fcc
Reviewed-on: https://go-review.googlesource.com/c/go/+/294409
Reviewed-by: Dmitri Shuralyov <dmitshur@golang.org>
Trust: Russ Cox <rsc@golang.org>
Run-TryBot: Russ Cox <rsc@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>

4 years agospec: every type has a method set (minor clarification)
Robert Griesemer [Wed, 17 Feb 2021 05:05:08 +0000 (21:05 -0800)]
spec: every type has a method set (minor clarification)

The spec states that a type "may" have a method set associated with it.
Yet every type has a method set, which may be empty. This is clarified
later in the same paragraph. Be clear in the first sentence as well.

Per the suggestion from https://github.com/DQNEO.

Fixes #44318.

Change-Id: I6097b1c7062853e404b7fead56d18a7f9c576fc3
Reviewed-on: https://go-review.googlesource.com/c/go/+/292853
Trust: Robert Griesemer <gri@golang.org>
Run-TryBot: Robert Griesemer <gri@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
4 years agogo/types: don't write during sanitizeInfo if nothing has changed
Rob Findley [Fri, 19 Feb 2021 22:37:17 +0000 (17:37 -0500)]
go/types: don't write during sanitizeInfo if nothing has changed

In its final phase, the typechecker walks the types it produces to
ensure that no unexpanded type instances leak through the API. However,
this also walks shared types (such as those in the universe scope),
resulting in a potential data race during concurrent typechecking
passes.

Fix this by being careful not to write if nothing needs to be changed.
Since any shared types should already be sanitized, this should
eliminate data races.

For #44434

Change-Id: Iadb2e78863efe0e974e69a00e255f26cfaf9386a
Reviewed-on: https://go-review.googlesource.com/c/go/+/294411
Trust: Robert Findley <rfindley@google.com>
Trust: Robert Griesemer <gri@golang.org>
Run-TryBot: Robert Findley <rfindley@google.com>
Reviewed-by: Robert Griesemer <gri@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>

4 years agoall: REVERSE MERGE dev.typeparams (7cdfa49) into master
Robert Griesemer [Fri, 19 Feb 2021 20:23:38 +0000 (12:23 -0800)]
all: REVERSE MERGE dev.typeparams (7cdfa49) into master

This commit is a REVERSE MERGE.
It merges dev.typeparams back into its parent branch, master.
This marks the end of development on dev.typeparams.

Merge List:

+ 2021-02-19 7cdfa4969a [dev.typeparams] all: merge master (06b86e9) into dev.typeparams
+ 2021-02-19 6521c7b378 [dev.typeparams] cmd/compile/internal/types2: resolve decl cycle the same way as in go/types
+ 2021-02-19 dfe0ef961b [dev.typeparams] go/types, types2: revert fancy struct printing (fixes x/tools tests)
+ 2021-02-19 2f37939a21 go/parser: improve error recovery from invalid selector exprs
+ 2021-02-19 8654db4555 [dev.typeparams] go/types: adjust printing of embedded struct fields (fixes x/tools/cmd/guru tests)
+ 2021-02-19 47e4b0739e Merge "[dev.typeparams] all: merge master (eb98272) into dev.typeparams" into dev.typeparams
+ 2021-02-18 a789be7814 [dev.typeparams] cmd/compile: use new converter functions rather than methods (fix build)
+ 2021-02-18 20050a15fe [dev.typeparams] cmd/compile:  support generic types (with stenciling of method calls)
+ 2021-02-18 e7493a9c74 [dev.typeparams] all: merge master (eb98272) into dev.typeparams
+ 2021-02-18 2ff1e05a4c [dev.typeparams] all: update parent repository
+ 2021-02-18 8960ce7735 [dev.typeparams] cmd/compile/internal/types2: minor adjustments to match go/types more closely
+ 2021-02-18 6f3878b942 [dev.typeparams] cmd/compile/internal/types: review of typestring_test.go
+ 2021-02-18 d6bdd1aeef [dev.typeparams] cmd/compile/internal/types: review of typestring.go
+ 2021-02-18 c2314babb8 [dev.typeparams] cmd/compile/internal/types: review of type.go
+ 2021-02-18 099374b55e [dev.typeparams] cmd/compile/internal/types2: remove Type.Under method in favor of function
+ 2021-02-18 653386a89a [dev.typeparams] cmd/compile/internal/types2: replace Named, TypeParam methods with functions
+ 2021-02-18 5e4da8670b [dev.typeparams] cmd/compile/internal/types2: use converter functions rather than methods
+ 2021-02-18 5ecb9a7887 [dev.typeparams] go/types: use a new ast.ListExpr for multi-type instances
+ 2021-02-18 7b679617f3 [dev.typeparams] go/types: conversions to type parameters are not constant
+ 2021-02-17 f5d0c653e6 [dev.typeparams] merge master (2f0da6d) into dev.typeparams
+ 2021-02-17 e196cb8258 [dev.typeparams] cmd/dist:  disable -G=3 on the std go tests for now
+ 2021-02-12 a06bd9fecb [dev.typeparams] cmd/compile/internal/types: review of resolver_test.go
+ 2021-02-12 042f88fe30 [dev.typeparams] cmd/compile/internal/types: review of errors_test.go
+ 2021-02-12 0abd7b768b [dev.typeparams] cmd/compile/internal/types: review of universe.go
+ 2021-02-12 1b6f0bf1b2 [dev.typeparams] cmd/compile/internal/types: review of sizes_test.go
+ 2021-02-12 1758780181 [dev.typeparams] cmd/compile/internal/types: review of sizes.go
+ 2021-02-12 3aee461d5c [dev.typeparams] cmd/compile/internal/types: review of return.go
+ 2021-02-12 7428318af6 [dev.typeparams] cmd/compile/internal/types: review of object_test.go
+ 2021-02-12 b20f9e2da1 [dev.typeparams] cmd/compile/internal/types: review of object.go
+ 2021-02-12 20746b2f37 [dev.typeparams] cmd/compile/internal/types: review of labels.go
+ 2021-02-12 bab3461123 [dev.typeparams] cmd/compile/internal/types: review of infer.go
+ 2021-02-12 9168590977 [dev.typeparams] cmd/compile/internal/types: review of builtin_test.go
+ 2021-02-12 f1777cf84c [dev.typeparams] cmd/compile/internal/types: review of builtin.go
+ 2021-02-12 0f43973b4b [dev.typeparams] go/types: make predeclared "any" alias for interface{}
+ 2021-02-12 58758e0a21 [dev.typeparams] go/types: better error message for invalid ... use
+ 2021-02-11 c0aa7bd760 [dev.typeparams] cmd/compile: small fixes for stenciling
+ 2021-02-10 df23540dde [dev.typeparams] cmd/gofmt: add the -G flag to allow generic code
+ 2021-02-10 ddec18cf82 [dev.typeparams] cmd/compile/internal/types2: overlapping embedded interfaces requires go1.14
+ 2021-02-10 fdf3496fcc [dev.typeparams] cmd/compile:  make type conversions by type parameters work
+ 2021-02-10 12e15d430d [dev.typeparams] cmd/compile: handle calling a method on a type param in stenciling
+ 2021-02-08 ca18c42054 [dev.typeparams] merge dev.regabi (618e3c1) into dev.typeparams
+ 2021-02-08 a360eeb528 [dev.typeparams] cmd/compile/internal/types2: conversions to type parameters are not constant
+ 2021-02-08 0fbde54ea6 [dev.typeparams] cmd/compile: allow generic funcs to call other generic funcs for stenciling
+ 2021-02-05 dcb5e0392e [dev.typeparams] cmd/compile: add stenciling of simple generic functions
+ 2021-02-04 f37b0c6c12 [dev.typeparams] cmd/compile/internal/types2: type alias decl requires go1.9
+ 2021-02-04 721488498a [dev.typeparams] cmd/compile: pass -lang flag value to new type checker
+ 2021-02-04 1ff2fdaaf1 [dev.typeparams] cmd/compile/internal/types2: add support for language version checking
+ 2021-02-04 370e9f5843 [dev.typeparams] cmd/compile/internal/types2: use 512 bits as max. integer precision
+ 2021-02-04 ca2f152893 [dev.typeparams] go/types: add missing test from dev.go2go
+ 2021-02-03 dc122c7a9c [dev.typeparams] test: exclude a failing test again (fix 32bit builds)
+ 2021-02-03 c910fd7b77 [dev.typeparams] cmd/compile: refuse excessively long constants
+ 2021-02-03 3db6e18468 [dev.typeparams] test: enable more errorcheck tests
+ 2021-02-03 bb53a5ad43 [dev.typeparams] cmd/compile/internal/importer: adjust importer to match compiler importer
+ 2021-02-03 3f845b3b45 [dev.typeparams] cmd/compile: deal with inferred type arguments
+ 2021-02-03 e633f343ba [dev.typeparams] cmd/compile:  add OFUNCINST/OTYPEINST nodes for generic func/type instantiation
+ 2021-02-02 0d2d6c7464 [dev.typeparams] all: merge dev.regabi (23b0c1f) into dev.typeparams
+ 2021-02-02 3d5c715bf2 [dev.typeparams] Handling multiple type arguments for call via new node OLIST
+ 2021-02-01 13a7412983 [dev.typeparams] Parse a generic type arg for generic function call
+ 2021-01-30 0aafd69124 [dev.typeparams] cmd/compile: start translating type params in noder2
+ 2021-01-30 a59cb5109d [dev.typeparams] cmd/compile/internal/types2: handle untyped constant arithmetic overflow
+ 2021-01-29 507e641963 [dev.typeparams] cmd/compile/internal/typecheck: declutter generated builtin.go (cleanup)
+ 2021-01-28 2440dd457a [dev.typeparams] cmd/compile:  start adding info needed for typeparams in types & ir
+ 2021-01-28 c0bf904ddf [dev.typeparams] cmd/compile/internal/types2: translate syntax to token constants via tables
+ 2021-01-28 f7d1c5990b [dev.typeparams] cmd/compile/internal/types2: must not import a package called "init"
+ 2021-01-27 217a461f56 [dev.typeparams] cmd/compile/internal/types2: report unused packages in source order
+ 2021-01-26 08a598f8c1 [dev.typeparams] cmd/compile: fix MethodExpr handling with embedded fields
+ 2021-01-26 cecc1dfcba [dev.typeparams] test: enable excluded test fixedbugs/issue7742.go
+ 2021-01-26 e48d7d3b21 [dev.typeparams] go/constant: faster match implementation
+ 2021-01-26 d39685e5e9 [dev.typeparams] go/constant: choose internal float representations more consistently
+ 2021-01-25 34704e374f [dev.typeparams] all: merge dev.regabi (5e4a0cd) into dev.typeparams
+ 2021-01-26 c97af0036b [dev.typeparams] cmd/compile: force untyped constants from types2 to expected kind
+ 2021-01-25 3663a437a7 [dev.typeparams] go/constant: in ToFloat, convert to rational numbers, not floats
+ 2021-01-25 3432d24bab Merge "[dev.typeparams] all: merge dev.regabi (063c72f) into dev.typeparams" into dev.typeparams
+ 2021-01-25 493eb6e6ec [dev.typeparams] cmd/compile: fix -G=3 handling of blank methods
+ 2021-01-25 13f02018af [dev.typeparams] test: enable more errorcheck tests w/ -G=3
+ 2021-01-24 6d8d118762 [dev.typeparams] all: merge dev.regabi (063c72f) into dev.typeparams
+ 2021-01-24 9456804e86 [dev.typeparams] test: fix excluded files lookup so it works on Windows
+ 2021-01-23 7947df436d [dev.typeparams] test: set -G=3 and enable more errorcheck tests in run.go
+ 2021-01-23 a49e941027 [dev.typeparams] cmd/compile/internal/types2: remove MethodSet code - not used by types2
+ 2021-01-23 5347241b5e [dev.typeparams] cmd/compile/internal/types2: use same sort criteria for methods as compiler
+ 2021-01-23 2b95c28b18 [dev.typeparams] cmd/compile: refactor SelectorExpr code into helpers
+ 2021-01-23 1946a77e69 Merge "[dev.typeparams] all: merge dev.regabi (7e0a81d) into dev.typeparams" into dev.typeparams
+ 2021-01-23 6923019a71 [dev.typeparams] cmd/compile/internal/types2: factor out sorting of methods
+ 2021-01-22 6e46c8fbb5 [dev.typeparams] all: merge dev.regabi (7e0a81d) into dev.typeparams
+ 2021-01-22 e4ef30a667 [dev.typeparams] cmd/compile: refactor irgen's handling of ":="
+ 2021-01-22 626406b703 [dev.typeparams] go/types: import api_test.go changes from dev.go2go
+ 2021-01-22 12cd9cf7e0 [dev.typeparams] cmd/compile: disambiguate OXDOT in noder using types2 Selection info
+ 2021-01-22 f8654579cd [dev.typeparams] cmd/compile/internal/types2: adjust errors in branch checking code, fix a bug
+ 2021-01-21 18bd7aa625 [dev.typeparams] cmd/compile: use nil instead of syntax.ImplicitOne
+ 2021-01-21 2427f6e6c0 [dev.typeparams] cmd/compile: directly set some simple expression types
+ 2021-01-21 455c29af83 [dev.typeparams] cmd/compile/internal/types2: convert untyped arguments to delete
+ 2021-01-21 f03f934ede [dev.typeparams] cmd/compile/internal/types2: make predeclared "any" alias for interface{}
+ 2021-01-20 0f054c5be0 [dev.typeparams] cmd/dist: add -G=3 test coverage
+ 2021-01-20 89ec17be9a [dev.typeparams] cmd/compile: simplify how irgen handles qualified idents
+ 2021-01-20 fa01ade41e [dev.typeparams] go/types: add tests from dev.go2go
+ 2021-01-20 734cb8be0a [dev.typeparams] go/types: refactor untyped conversion for typeparams
+ 2021-01-20 d8796b5670 [dev.typeparams] cmd/compile/internal/types2: report type of nil based on context
+ 2021-01-19 48a3cb399d [dev.typeparams] go/types: fix some merge errors in call.go
+ 2021-01-19 3e15bf7716 [dev.typeparams] go/types: don't modify Named.underlying in validType
+ 2021-01-19 f38f862417 [dev.typeparams] go/types: strip annotations from errors
+ 2021-01-19 2e64511ac9 [dev.typeparams] go/types: unify methods in missingMethod
+ 2021-01-19 958927c824 [dev.typeparams] go/parser: error for type instances without ParseTypeParams
+ 2021-01-19 90bfc73071 [dev.typeparams] cmd/compile: cache mapped types during irgen
+ 2021-01-19 3c0a39c964 [dev.typeparams] cmd/compile/internal/types: minor fixes/cleanups around testing
+ 2021-01-15 502198c8dc [dev.typeparams] cmd/compile/internal/types2: consistently report nil type as "untyped nil"
+ 2021-01-14 82c3f0a358 [dev.typeparams] cmd/compile/internal/types2: untyped shift counts must fit into uint
+ 2021-01-14 67bf62d939 [dev.typeparams] cmd/compile/internal/types2: better error message for invalid ... use
+ 2021-01-14 ef5285fbd0 [dev.typeparams] cmd/compile: add types2-based noder
+ 2021-01-12 f065ff221b [dev.typeparams] all: merge dev.regabi (d9acf6f) into dev.typeparams
+ 2021-01-12 099599662d [dev.typeparams] cmd/compile: refactor import logic
+ 2021-01-12 106aa941df [dev.typeparams] cmd/compile: refactor DWARF scope marking
+ 2021-01-12 6a56c6c870 [dev.typeparams] go/types: import dev.go2go changes to check tests
+ 2021-01-11 9e746e4255 [dev.typeparams] cmd/compile: refactor varEmbed logic
+ 2021-01-11 3e1a87ac2a [dev.typeparams] cmd/compile: extract posMap from noder
+ 2021-01-11 2e8f29b79d [dev.typeparams] cmd/compile: add types2.Sizes implementation
+ 2021-01-11 44d1a8523a [dev.typeparams] cmd/compile/internal/types2: fixes for all.bash
+ 2021-01-11 8123bc90b8 [dev.typeparams] cmd/go: relax test expectation
+ 2021-01-11 8c5aa42c79 [dev.typeparams] cmd/compile: calculate variable sizes in walk
+ 2021-01-11 1ce0854157 [dev.typeparams] import stmt changes from dev.go2go
+ 2021-01-11 eb53a6c7cf [dev.typeparams] import operand.go changes from dev.go2go
+ 2021-01-11 81cd99858d [dev.typeparams] go/types: import expr changes from dev.go2go
+ 2021-01-08 822aeacd9e [dev.typeparams] cmd/compile/internal/syntax: remove ShortString, use String instead
+ 2021-01-08 d017a1b649 [dev.typeparams] cmd/compile/internal/syntax: add Walk node vistor from types2
+ 2021-01-08 7903214fcc [dev.typeparams] cmd/compile/internal/syntax: add ShortString tests
+ 2021-01-08 0aede1205b [dev.typeparams] cmd/compile/internal/types2: use syntax printer to print expressions
+ 2021-01-08 934f9dc0ef [dev.typeparams] cmd/compile/internal/syntax: clean up node printing API
+ 2021-01-07 5b9152de57 [dev.typeparams] all: merge dev.regabi (cb05a0a) into dev.typeparams
+ 2021-01-07 7e689f86e3 [dev.typeparams] go/types: move use and useLHS to match dev.go2go
+ 2021-01-06 d76cefed1f [dev.typeparams] go/types: remove disabled code related to type lists
+ 2021-01-06 0e286579c5 [dev.typeparams] go/types: import typexpr.go from dev.go2go
+ 2021-01-06 9546596d77 [dev.typeparams] cmd/compile/internal/types2: remove disabled code related to type lists
+ 2021-01-06 196102d046 [dev.typeparams] cmd/compile/internal/types2: review of typexpr.go
+ 2021-01-05 a8fe098a12 [dev.typeparams] all: merge dev.regabi (77365c5) into dev.typeparams
+ 2020-12-28 a800acaae1 [dev.typeparams] all: merge dev.regabi (07569da) into dev.typeparams
+ 2020-12-27 dd40bbc57b [dev.typeparams] cmd/compile: re-enable internal/types2 test
+ 2020-12-23 d99dd17827 [dev.typeparams] all: merge dev.regabi (37f138d) into dev.typeparams
+ 2020-12-22 8b1fbd8215 [dev.typeparams] all: merge dev.regabi (6d03cde) into dev.typeparams
+ 2020-12-22 23922cf83b [dev.typeparams] all: merge dev.regabi (d1d1099) into dev.typeparams
+ 2020-12-23 788dad53c5 [dev.typeparams] cmd/compile/internal/types2: disable external test temporarily
+ 2020-12-22 91cc51e005 [dev.typeparams] all: merge dev.regabi (ec741b0) into dev.typeparams
+ 2020-12-22 e02a007ffd [dev.typeparams] codereview.cfg: add config for dev.typeparams
+ 2020-12-22 41e7901ca4 [dev.typeparams] cmd/compile/internal/types2: report error for invalid main function signature
+ 2020-12-22 53c4c17b09 [dev.typeparams] all: merge dev.regabi into dev.typeparams
+ 2020-12-17 060cdbc7b5 [dev.typeparams] go/types: import object resolution from dev.go2go
+ 2020-12-17 c4f0da5750 [dev.typeparams] cmd/compile/internal/types2: remove code for implicit type arguments
+ 2020-12-16 068dd0470b [dev.typeparams] cmd/compile/internal/syntax: don't panic when providing -verify
+ 2020-12-16 7909d6ec28 [dev.typeparams] cmd/compile/internal/syntax: type parameters must always be named
+ 2020-12-16 3b5918c757 [dev.typeparams] cmd/compile/internal/types2: review of predicates.go
+ 2020-12-16 09abd23d9e [dev.typeparams] go/types: import predicates.go from dev.go2go
+ 2020-12-16 f38da2cbb6 [dev.typeparams] cmd/compile/internal/types2: review of unify.go
+ 2020-12-16 ceb77db24f [dev.typeparams] go/types: import some support functions from dev.go2go
+ 2020-12-15 a4d4c10340 [dev.typeparams] go/types: import lookup logic from dev.go2go
+ 2020-12-15 1306435103 [dev.typeparams] go/types: import changes to types.Info from dev.go2go
+ 2020-12-15 6b18081d01 [dev.typeparams] cmd/compile/internal/types2: don't crash if import path is missing
+ 2020-12-15 14e4267c34 [dev.typeparams] cmd/compile/internal/types2: report error for invalid (but empty) expr switch
+ 2020-12-15 96999296e6 [dev.typeparams] go/types: import unify.go and infer.go from dev.go2go
+ 2020-12-15 5aff757efc [dev.typeparams] cmd/compile/internal/gc: provide types2 selection info to noder
+ 2020-12-14 c8e73489c3 go/types: import instance.expand and subst.go from dev.go2go
+ 2020-12-14 3a912f279f [dev.typeparams] cmd/compile/internal/syntax: export NewName and use it
+ 2020-12-14 8ec9e89000 [dev.typeparams] cmd/compile/internal/types2: fix stray ')' in error
+ 2020-12-14 8fe8e29c9f [dev.typeparams] cmd/compile/internal/types2: report error for invalid type expression
+ 2020-12-14 f8930a2413 [dev.typeparams] cmd/compile/internal/types2: report invalid ... in conversions
+ 2020-12-14 5aca6e7857 [dev.typeparams] test: finish triaging all outstanding failing tests
+ 2020-12-14 df58f3368e [dev.typeparams] cmd/compile/internal/types2: don't report two errors for bad strings
+ 2020-12-14 91803a2df3 [dev.typeparams] merge: merge branch 'dev.regabi' into 'dev.typeparams'
+ 2020-12-11 a20021227e [dev.typeparams] cmd/compile/internal/types2: bring over subst.go changes from go/types
+ 2020-12-10 ddf44904f1 [dev.typeparams] test: exclude 32bit-specific test that fails on 32bit platforms (fix build)
+ 2020-12-10 dbce27d29c [dev.typeparams] cmd/compile/internal/types2: report correct line number for missing key
+ 2020-12-09 edf80c4209 [dev.typeparams] cmd/compile/internal/types2: adjusted more error messages for compiler
+ 2020-12-09 43c7b214db [dev.typeparams] cmd/compile/internal/types2: adjusted qualified identifier error message for compiler
+ 2020-12-09 6812eae2e2 [dev.typeparams] cmd/compile/internal/types2: adjust init cycle error message for compiler
+ 2020-12-09 810957b155 [dev.typeparams] cmd/compile/internal/types2: adjusted array error message for compiler
+ 2020-12-09 c32566c336 [dev.typeparams] cmd/compile/internal/types2: avoid endless recursion in Comparable predicate
+ 2020-12-08 6015c4e543 [dev.typeparams] go/*: add TODOs from CLs importing dev.go2go changes
+ 2020-12-08 b6e678573d [dev.typeparams] go/types: import universe changes from dev.go2go
+ 2020-12-08 14dc2d2345 [dev.typeparams] go/types: import the Type API from dev.go2go
+ 2020-12-07 cd15a48036 [dev.typeparams] cmd/compile/internal/types2: correct error position for inherited const init expression
+ 2020-12-04 02820d61a9 [dev.typeparams] test: enable some more errorcheck tests
+ 2020-12-03 9ff27e9fad [dev.typeparams] test: run all errorcheck tests that pass compiler with -G flag
+ 2020-12-03 7a1aa7dfaf [dev.typeparams] test: adjust more test cases to match compiler -G output
+ 2020-12-03 07d32c8183 [dev.typeparams] cmd/compile/internal/types: adjust some error messages to match the compiler
+ 2020-12-03 6b4da14dd3 [dev.typeparams] cmd/compile: provide scaffolding to get types2 types during noding
+ 2020-12-02 72cc2353f0 [dev.typeparams] go/printer: adapt changes from dev.go2go
+ 2020-12-02 ab18125567 [dev.typeparams] cmd/compile/internal/types2: no "declared but not used" errors for invalid var decls
+ 2020-12-02 036245862a [dev.typeparams] cmd/compile/internal/types2: set compiler error message for undeclared variable
+ 2020-12-01 bdc4ffe9a8 [dev.typeparams] cmd/compile/internal/types2: add Config.IgnoreBranches flag
+ 2020-12-01 72ad2f44ea [dev.typeparams] test: add scaffolding to run.go to check compiler with -G flag
+ 2020-12-01 87bc85a846 [dev.typeparams] cmd/compile/internal/types2: adjustments toward matching compiler error messages
+ 2020-12-01 fbc4c6a3ae [dev.typeparams] cmd/compile/internal/types2: remove support for type parameter pointer designation
+ 2020-11-25 ba9c35aa12 [dev.typeparams] merge dev.regabi 40f5bc4d556a into dev.typeparams
+ 2020-11-25 5c2e14872c [dev.typeparams] merge dev.regabi 41f3af9d04 into dev.typeparams
+ 2020-11-25 2c25cd5ba7 [dev.typeparams] cmd/compile/internal/types2: a type parameter is a valid type case in a type switch
+ 2020-11-24 0fb733b7f7 [dev.typeparams] go/parser: support the ParseTypeParams mode
+ 2020-11-24 b56762129e [dev.typeparams] import go2go changes to parse type parameters
+ 2020-11-24 a324aebb7d [dev.typeparams] go/types, cmd/compile/internal/types2: fix incorrect string(int) conversion (regression)
+ 2020-11-20 8fbdacf64c [dev.typeparams] cmd/compile/internal/types2: report constant overflow in binary ops
+ 2020-11-19 b1ae0a0646 [dev.typeparams] Merge branch 'master' into dev.typeparams
+ 2020-11-19 0123c9b321 [dev.typeparams] cmd/compile/internal/types2: report an error for invalid constant values
+ 2020-11-18 2140049172 [dev.typeparams] cmd/compile/internal/types2: port of https://golang.org/cl/270957
+ 2020-11-11 6877ee1e07 [dev.typeparams] cmd/compile: use existing findpkg algorithm when importing through types2
+ 2020-11-03 165ceb09f9 [dev.typeparams] merge master into dev.typeparams
+ 2020-10-28 41ff51ae00 [dev.typeparams] cmd/compile/internal/types2: review of scopes.go
+ 2020-10-27 9392b82919 [dev.typeparams] cmd/compile/internal/types: review of objset.go
+ 2020-10-27 c32ac6c15f [dev.typeparams] cmd/compile/internal/types: review of selection.go
+ 2020-10-27 ff6ab114c9 [dev.typeparams] cmd/compile/internal/types: review of gccgosizes.go
+ 2020-10-27 abb31c2558 [dev.typeparams] cmd/compile/internal/types2: review of hilbert_test.go
+ 2020-10-27 6e98406ac3 [dev.typeparams] cmd/compile/internal/types2: review of initorder.go
+ 2020-10-27 38af45b4cb [dev.typeparams] cmd/compile/internal/types2: review of package.go
+ 2020-10-27 87eab74628 [dev.typeparams] cmd/compile: enable type-checking of generic code
+ 2020-10-23 5bfd2964a6 [dev.typeparams] merge master into dev.typeparams
+ 2020-10-22 a10fe9f6e7 go/ast: import AST changes supporting typeparams from dev.go2go
+ 2020-10-21 7a8a720c80 [dev.typeparams] cmd/compile/internal/types2: add *.go2 (generic) tests
+ 2020-10-21 755d6de1d8 [dev.typeparams] cmd/compile/internal/types2: add some more tests
+ 2020-10-21 1d07306a44 [dev.typeparams] cmd/compile/internal/types2: adjust tests, enable Testdata tests
+ 2020-10-21 befc62a2c4 [dev.typeparams] cmd/compile/internal/types2: add testdata directory
+ 2020-10-21 ca36ba83ab [dev.typeparams] cmd/compile/internal/importer, types2: initial check-in of types2 and importer
+ 2020-10-20 6ff16fe3ee [dev.typeparams] cmd/compile/internal/syntax: add utility functions for testing
+ 2020-10-19 5e46c6a10f [dev.typeparams] cmd/compile/internal/syntax: add Pos method
+ 2020-10-15 e9e58a4d49 [dev.typeparams] cmd/compile/internal/syntax: fix printing of channel types
+ 2020-10-14 73f529845c [dev.typeparams] cmd/compile/internal/syntax: always use IndexExpr node for type instantiation
+ 2020-10-13 48755e06aa [dev.typeparams] cmd/compile: enable parsing of generic code with new -G flag
+ 2020-10-13 7668f02dec [dev.typeparams] cmd/compile/internal/syntax: add type parameter tests
+ 2020-10-13 b627988b0c [dev.typeparams] cmd/compile/internal/syntax: implement parsing of type parameters
+ 2020-10-13 617b633917 [dev.typeparams] cmd/compile/internal/syntax: prepare syntax nodes for type parameters
+ 2020-10-12 986cad14e2 [dev.typeparams] merge master into dev.typeparams

Change-Id: Ie0796a53af1fb5db49f4c748144461d3af598575

4 years ago[dev.typeparams] all: merge master (06b86e9) into dev.typeparams
Robert Griesemer [Fri, 19 Feb 2021 19:56:01 +0000 (11:56 -0800)]
[dev.typeparams] all: merge master (06b86e9) into dev.typeparams

Merge List:

+ 2021-02-19 06b86e9803 cmd/compile:  fix check to avoid creating new closure function when typechecking inline body
+ 2021-02-19 9322eec8a2 codereview.cfg: add codereview.cfg for master branch
+ 2021-02-19 02e5a8fdfc runtime: ignore SPWRITE in syscall functions
+ 2021-02-19 fa18f224c3 runtime/pprof: disable TestMorestack on macOS under race detector
+ 2021-02-19 01eb70e3dd os: fix hex exit code print on 32-bit windows
+ 2021-02-19 49add6ad90 runtime: fix spurious stack overflow detection
+ 2021-02-19 fce2a94d84 cmd/compile: fix buglet in inlined info abstract function dwarf-gen
+ 2021-02-19 7764ee5614 runtime: fix invalid nil g check for for mips64x
+ 2021-02-19 87f425da14 cmd/go/internal/mvs: split Reqs into narrower per-function interfaces
+ 2021-02-19 4da0188c6c cmd/go/internal/modget: split resolveCandidates into two methods
+ 2021-02-19 5f2e24efb3 cmd/internal/diff: skip over Cygwin warning in diff output
+ 2021-02-19 ee7038f6a5 net: disable Windows netsh tests when netsh won't run
+ 2021-02-19 40765ffa95 os/exec: disable failing LookPathTest on windows/arm64
+ 2021-02-19 b445d6ea34 runtime/pprof: expect tests to pass on macOS
+ 2021-02-19 b110a43628 runtime: delete gosave (dead code)
+ 2021-02-19 474d5f4f4d math: remove most 387 implementations
+ 2021-02-19 c7c6c113be runtime: convert windows/arm64 assembly
+ 2021-02-19 3527caa7d6 runtime: initial windows/arm64 implementation files
+ 2021-02-19 427bd7599d runtime: generate windows/arm64 callback asm
+ 2021-02-19 f6c4b4bf96 syscall: add windows/arm64 support
+ 2021-02-19 ac024a0c7b cmd/vendor: get golang.org/x/sys@beda7e5e158
+ 2021-02-19 a3b97e7628 test: disable nilptr on windows/arm64
+ 2021-02-19 985d087782 cmd/link: add windows/arm64 support
+ 2021-02-19 95a44d2409 cmd/internal/objfile: recognize Windows ARM64 executables
+ 2021-02-19 0ca0551f02 debug/pe: recognize arm64 executables
+ 2021-02-19 0c633125f2 cmd/dist: add windows/arm64 support
+ 2021-02-19 bb6efb9609 build: set GOPATH consistently in run.bash, run.bat, run.rc
+ 2021-02-19 a1222b7535 cmd/link: add debug print in deadcode
+ 2021-02-19 1c659f2525 cmd/link: clean up windows PE generation
+ 2021-02-19 b6379f190b syscall: clean up windows a bit
+ 2021-02-19 09e059afb1 runtime: enable framepointer on all arm64
+ 2021-02-19 b19e7b518e runtime: clean up windows a bit
+ 2021-02-19 5421c37a1d runtime: fix windows/arm externalthreadhandler
+ 2021-02-19 91cc484ea9 runtime: fix time on windows/arm under WINE
+ 2021-02-19 38672d3dcf runtime: crash earlier on windows for runtime.abort
+ 2021-02-19 a1e9148e3d runtime: print hex numbers with hex prefixes in traceback debug
+ 2021-02-19 75e273fc2c runtime: fix windows/arm CONTEXT_CONTROL
+ 2021-02-19 76ab626bfc runtime: factor common code out of defs_windows_*.go
+ 2021-02-19 ece954d8b8 runtime: find g in Windows profiler using SP
+ 2021-02-19 a54f7fc0fd runtime: do not treat asmcgocall as a topofstack on g0
+ 2021-02-19 776ee4079a runtime: do not treat morestack as a topofstack
+ 2021-02-19 5ecd9e34df runtime: do not treat mcall as a topofstack
+ 2021-02-19 54da3ab385 runtime: use TOPFRAME to identify top-of-frame functions
+ 2021-02-19 fbe74dbf42 runtime: use FuncInfo SPWRITE flag to identify untraceable profile samples
+ 2021-02-19 4dd77bdc91 cmd/asm, cmd/link, runtime: introduce FuncInfo flag bits
+ 2021-02-19 aa0388f2ed runtime: remove unnecessary writes to gp.sched.g
+ 2021-02-19 6fe8981620 cmd/internal/obj/riscv: fix JMP name<>(SB)
+ 2021-02-19 01f05d8ff1 runtime: unify asmcgocall and systemstack traceback setup
+ 2021-02-19 229695a283 runtime: clean up funcID assignment
+ 2021-02-19 c80da0a33a runtime: handle nil gp in cpuprof
+ 2021-02-19 a78879ac67 runtime: move sys.DefaultGoroot to runtime.defaultGOROOT
+ 2021-02-19 8ac23a1f15 runtime: document, clean up internal/sys
+ 2021-02-19 678568a5cf runtime: delete windows setlasterror (unused)
+ 2021-02-19 0d94f989d1 runtime: clean up system calls during cgo callback init
+ 2021-02-19 e7ee3c1fa8 os: report Windows exit status in hex

Change-Id: Ib27e232dabac0604b588f27749bd8ad45a281fe3

4 years ago[dev.typeparams] cmd/compile/internal/types2: resolve decl cycle the same way as...
Robert Griesemer [Fri, 19 Feb 2021 17:39:01 +0000 (09:39 -0800)]
[dev.typeparams] cmd/compile/internal/types2: resolve decl cycle the same way as in go/types

Minor adjustment to match go/types more closely.

Change-Id: Id79c51f0ecd8cda0f5b68f6e961500f7f22f7115
Reviewed-on: https://go-review.googlesource.com/c/go/+/294270
Reviewed-by: Robert Findley <rfindley@google.com>
Trust: Robert Griesemer <gri@golang.org>

4 years agocmd/compile: fix check to avoid creating new closure function when typechecking...
Dan Scales [Fri, 19 Feb 2021 02:07:09 +0000 (18:07 -0800)]
cmd/compile:  fix check to avoid creating new closure function when typechecking inline body

By default, when typechecking a closure, tcClosure() creates a new
closure function. This should really be done separate from typechecking.
For now, we explicitly avoid creating a new closure function when
typechecking an inline body (in ImportedBody). However, the heuristic
for determining when we are typechecking an inline body was not correct
for double nested closures in an inline body, since CurFunc will then be
the inner closure, which has a body.

So, use a simple global variable to indicate when we typechecking an
inline body. The global variable is fine (just like ir.CurFunc), since
the front-end runs serially.

Fixes #44325

Change-Id: If2829fe1ebb195a7b1a240192b57fe6f04d1a36b
Reviewed-on: https://go-review.googlesource.com/c/go/+/294211
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
Trust: Dan Scales <danscales@google.com>
Run-TryBot: Dan Scales <danscales@google.com>

4 years agocodereview.cfg: add codereview.cfg for master branch
Russ Cox [Fri, 19 Feb 2021 18:39:43 +0000 (13:39 -0500)]
codereview.cfg: add codereview.cfg for master branch

The codereview sync-branch command wants all involved
branches to have codereview.cfg, and this will help us when
we transition from master to main later this year.

Change-Id: Ia8e4c8b8c86864ed9d730e5d96be1ff386e2e1cb
Reviewed-on: https://go-review.googlesource.com/c/go/+/294291
Trust: Russ Cox <rsc@golang.org>
Run-TryBot: Russ Cox <rsc@golang.org>
Reviewed-by: Robert Findley <rfindley@google.com>
4 years agoruntime: ignore SPWRITE in syscall functions
Russ Cox [Fri, 19 Feb 2021 11:01:25 +0000 (06:01 -0500)]
runtime: ignore SPWRITE in syscall functions

netbsd/amd64's Syscall9 changes SP using ADD and SUB,
which are treated as SPWRITEs (they are not accounted for
in the sp-adjust tracking, and there are too many functions that
would report mismatched stack adjustments at RET if they were).
A traceback starting in Syscall9 as saved by entersyscall complains
about the SPWRITE-ness unnecessarily, since the PC/SP are saved
at the start of the function. Ignore SPWRITE in that case.

netbsd/arm's Syscall6 also changes SP (R13), using a direct write.
So even if we could handle the ADD/SUB in the amd64 case or
rewrote that assembly, we'd still be stuck with a more difficult
problem in this case. Ignoring the SPWRITE fixes it.

Example crashes:
https://build.golang.org/log/160fc7b051a2cf90782b75a99984fff129329e66
https://build.golang.org/log/7879e2fecdb400eee616294285e1f952e5b17301

Change-Id: I0c8e9696066e90dafed6d4a93d11697da23f0080
Reviewed-on: https://go-review.googlesource.com/c/go/+/294072
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Trust: Russ Cox <rsc@golang.org>
Run-TryBot: Russ Cox <rsc@golang.org>

4 years agoruntime/pprof: disable TestMorestack on macOS under race detector
Russ Cox [Fri, 19 Feb 2021 10:42:20 +0000 (05:42 -0500)]
runtime/pprof: disable TestMorestack on macOS under race detector

This is failing but only under the race detector.
It doesn't really seem fair to expect pprof to find
specific profile events with the race detector slowing
everything down anyway.

Change-Id: I4b353d3d63944c87884d117e07d119b2c7bf4684
Reviewed-on: https://go-review.googlesource.com/c/go/+/294071
Trust: Russ Cox <rsc@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
4 years agoos: fix hex exit code print on 32-bit windows
Russ Cox [Fri, 19 Feb 2021 10:29:51 +0000 (05:29 -0500)]
os: fix hex exit code print on 32-bit windows

We want to print hex exit codes for the large values,
but on 32-bit Windows the large values are negative.

Change-Id: I0e350b128414a9468c93eddc62d660f552c1ee05
Reviewed-on: https://go-review.googlesource.com/c/go/+/294070
Trust: Russ Cox <rsc@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
4 years agoruntime: fix spurious stack overflow detection
Russ Cox [Fri, 19 Feb 2021 10:22:35 +0000 (05:22 -0500)]
runtime: fix spurious stack overflow detection

The regabi builders are unhappy about badctxt calling throw
calling systemstack calling gosave_systemstack_switch calling
badctxt, all nosplit, repeating. This wouldn't actually happen
since after one systemstack we'd end up on the system stack
and the next one wouldn't call gosave_systemstack_switch at all.

The badctxt call itself is in a very unlikely assertion failure
inside gosave_systemstack_switch.
Keep the assertion check but call runtime.abort instead on failure,
breaking the detected (but not real) cycle.

Change-Id: Iaf5c0fc065783b8c1c6d0f62d848f023a0714b96
Reviewed-on: https://go-review.googlesource.com/c/go/+/294069
Trust: Russ Cox <rsc@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
4 years agocmd/compile: fix buglet in inlined info abstract function dwarf-gen
Than McIntosh [Wed, 17 Feb 2021 20:51:05 +0000 (15:51 -0500)]
cmd/compile: fix buglet in inlined info abstract function dwarf-gen

When generating DWARF inlined info records, it's possible to have a
local function whose only callsites are inlined away, meaning that we
emit an abstract function DIE but no regular subprogram DIE. When
emitting DWARF scope info we need to handle this case (specifically
when scoping PCs, check for the case that the func in question has
been entirely deleted).

Fixes #44344.

Change-Id: I9f5bc692f225aa4c5c23f7bd2e50bcf7fe4fc5f3
Reviewed-on: https://go-review.googlesource.com/c/go/+/293309
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Reviewed-by: Russ Cox <rsc@golang.org>
Trust: Than McIntosh <thanm@google.com>
Run-TryBot: Than McIntosh <thanm@google.com>

4 years ago[dev.typeparams] go/types, types2: revert fancy struct printing (fixes x/tools tests)
Robert Griesemer [Fri, 19 Feb 2021 02:06:01 +0000 (18:06 -0800)]
[dev.typeparams] go/types, types2: revert fancy struct printing (fixes x/tools tests)

An embedded struct field is embedded by mentioning its type.
The fact that the field name may be different and derived
from the type doesn't matter for the struct type.

Do print the embedded type rather than the derived field
name, as we have always done in the past. Remove the fancy
new code which was just plain wrong.

The struct output printing is only relevant for debugging
and test cases. Reverting to the original code (pre-generics)
fixes a couple of x/tools tests.

Unfortunately, the original code is (also) not correct for
embedded type aliases. Adjusted a gccgoimporter test
accordingly and filed issue #44410.

This is a follow-up on https://golang.org/cl/293961 which
addressed the issue only partially and left the incorrect
code in place.

Change-Id: Icb7a89c12ef7929c221fb1a5792f144f7fcd5855
Reviewed-on: https://go-review.googlesource.com/c/go/+/293962
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>
4 years agoruntime: fix invalid nil g check for for mips64x
Meng Zhuo [Fri, 19 Feb 2021 06:31:57 +0000 (14:31 +0800)]
runtime: fix invalid nil g check for for mips64x

In CL 292109 we removed unnecessary writes to gp.sched.g
but put wrong register to save g (R4 saves pointer to g) on mips64x

Change-Id: I9777846a7b0a46e1af83dcfc73b74649e0dba3c9
Reviewed-on: https://go-review.googlesource.com/c/go/+/293989
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Joel Sing <joel@sing.id.au>
Trust: Meng Zhuo <mzh@golangcn.org>
Run-TryBot: Meng Zhuo <mzh@golangcn.org>

4 years agogo/parser: improve error recovery from invalid selector exprs
Rob Findley [Fri, 19 Feb 2021 01:08:42 +0000 (20:08 -0500)]
go/parser: improve error recovery from invalid selector exprs

Before this CL, the parser consumed the next token following an invalid
selector expr no matter what it was. This leads to poor error recovery
when this next token is a closing delimiter or other reasonable element
of a stop set. As a side-effect, x/tools tests broke when parser logic
for type parameters was introduced, as they threw off the parser
synchronization to the point where the x/tools test bailed out.

This CL introduces a targeted fix that allows the x/tools tests to pass.
More general improvement for parser error recovery should be done for
go1.17.

Change-Id: I44d73d34b6063e62d16a23d24ab7cbce6500239d
Reviewed-on: https://go-review.googlesource.com/c/go/+/293792
Trust: Robert Findley <rfindley@google.com>
Run-TryBot: Robert Findley <rfindley@google.com>
Reviewed-by: Robert Griesemer <gri@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>

4 years ago[dev.typeparams] go/types: adjust printing of embedded struct fields (fixes x/tools...
Robert Griesemer [Fri, 19 Feb 2021 01:01:54 +0000 (17:01 -0800)]
[dev.typeparams] go/types: adjust printing of embedded struct fields (fixes x/tools/cmd/guru tests)

Prior to 1.16, go/types printed an embedded struct field by simply
printing its type, which may have included a package qualification.
Just printing the type is not useful with generic types and we now
must print the actual field name derived from the type - this leads
to different output for non-generic imported embedded types. Fix by
printing a package qualification in that case.

Change-Id: I2cb2484da7732428d13fdfb5fe4ec1fa1ee813a2
Reviewed-on: https://go-review.googlesource.com/c/go/+/293961
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>

4 years agocmd/go/internal/mvs: split Reqs into narrower per-function interfaces
Bryan C. Mills [Tue, 9 Feb 2021 20:38:18 +0000 (15:38 -0500)]
cmd/go/internal/mvs: split Reqs into narrower per-function interfaces

Reqs currently combines requirements with upgrades and downgrades.
However, only Upgrade needs the Upgrade method, and only Downgrade
needs the Previous method.

When we eventually add lazy loading, the lazily-loaded module graph
will not be able to compute upgrades and downgrades, so the
implementation work from here to there will be clearer if we are
explicit about which are still needed.

For #36460

Change-Id: I7bf8c2a84ce6bc4ef493a383e3d26850e9a6a6c0
Reviewed-on: https://go-review.googlesource.com/c/go/+/290771
Trust: Bryan C. Mills <bcmills@google.com>
Reviewed-by: Michael Matloob <matloob@golang.org>
Reviewed-by: Jay Conrod <jayconrod@google.com>
4 years agocmd/go/internal/modget: split resolveCandidates into two methods
Bryan C. Mills [Thu, 4 Feb 2021 14:42:48 +0000 (09:42 -0500)]
cmd/go/internal/modget: split resolveCandidates into two methods

It turns out that the existing call sites of the resolveCandidates
method pass only *either* a slice of queries or a slice of upgrades
(never both), and the behaviors triggered by the two parameters don't
overlap much at all. To clarify the two different operations, split
them into two separate methods.

For #36460

Change-Id: I64651637734fd44fea68740a3cdfbacfb73c19b6
Reviewed-on: https://go-review.googlesource.com/c/go/+/289697
Trust: Bryan C. Mills <bcmills@google.com>
Run-TryBot: Bryan C. Mills <bcmills@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Michael Matloob <matloob@golang.org>
Reviewed-by: Jay Conrod <jayconrod@google.com>
4 years agocmd/internal/diff: skip over Cygwin warning in diff output
Russ Cox [Thu, 18 Feb 2021 18:01:10 +0000 (13:01 -0500)]
cmd/internal/diff: skip over Cygwin warning in diff output

This happens on Windows. Don't let it stop us.

Change-Id: Ie2115d5825e1c2217f237ed373adb35594a5aaff
Reviewed-on: https://go-review.googlesource.com/c/go/+/293850
Trust: Russ Cox <rsc@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
4 years agonet: disable Windows netsh tests when netsh won't run
Russ Cox [Thu, 18 Feb 2021 16:47:47 +0000 (11:47 -0500)]
net: disable Windows netsh tests when netsh won't run

On my Surface Pro X running the insider preview,
running "netsh help" from Powershell started from the task bar works.
But running "powershell" at a cmd.exe prompt and then running
"netsh help" produces missing DLL errors.
These aren't our fault, so just skip the netsh-based tests if this happens.

Change-Id: I13a17e01143d823d3b5242d827db056bd253e3e9
Reviewed-on: https://go-review.googlesource.com/c/go/+/293849
Trust: Russ Cox <rsc@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
4 years agoos/exec: disable failing LookPathTest on windows/arm64
Russ Cox [Thu, 18 Feb 2021 16:27:23 +0000 (11:27 -0500)]
os/exec: disable failing LookPathTest on windows/arm64

For #44379.

Change-Id: I9a3cf4d511a8286117f877c2ff9dbde56fa55983
Reviewed-on: https://go-review.googlesource.com/c/go/+/293709
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Trust: Russ Cox <rsc@golang.org>

4 years agoruntime/pprof: expect tests to pass on macOS
Russ Cox [Mon, 15 Feb 2021 22:02:30 +0000 (17:02 -0500)]
runtime/pprof: expect tests to pass on macOS

macOS tests have been disabled since CL 12429045 (Aug 2013).
At the time, macOS required a kernel patch to get a working profiler
(https://research.swtch.com/macpprof), which we didn't want
to require, of course.

macOS has improved - it no longer requires the kernel patch - but
we never updated the list of exceptions.

As far as I can tell, the builders have no problem passing the pprof test now.
(It is possible that the iOS builders have trouble, but that is now a different GOOS.)

Remove the exception for macOS. The test should now pass.

Fixes #6047.

Change-Id: Iab49036cacc1025e56f515bd19d084390c2f5357
Reviewed-on: https://go-review.googlesource.com/c/go/+/292229
Trust: Russ Cox <rsc@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
4 years agoruntime: delete gosave (dead code)
Russ Cox [Mon, 15 Feb 2021 16:26:58 +0000 (11:26 -0500)]
runtime: delete gosave (dead code)

Change-Id: Ie811526534df8622d89c5b1b81dbe19ece1c962b
Reviewed-on: https://go-review.googlesource.com/c/go/+/292110
Trust: Russ Cox <rsc@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
4 years agomath: remove most 387 implementations
Russ Cox [Thu, 11 Feb 2021 04:41:23 +0000 (23:41 -0500)]
math: remove most 387 implementations

The Surface Pro X's 386 simulator is not completely faithful to a real 387.
The most egregious problem is that it computes Log2(8) as 2.9999999999999996,
but it has some other subtler problems as well. All the problems occur in
routines that we don't even bother with assembly for on amd64.
If the speed of Go code is OK on amd64 it should be OK on 386 too.
Just remove all the 386-only assembly functions.

This leaves Ceil, Floor, Trunc, Hypot, and Sqrt in 386 assembly,
all of which are also in assembly on amd64 and all of which pass
their tests on Surface Pro X.

Compared to amd64, the 386 port omits assembly for Min, Max, and Log.
It never had Min and Max, and this CL deletes Log because Log2 wasn't
even correct. (None of the other architectures have assembly Log either.)

Change-Id: I5eb6c61084467035269d4098a36001447b7a0601
Reviewed-on: https://go-review.googlesource.com/c/go/+/291229
Trust: Russ Cox <rsc@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
4 years agoruntime: convert windows/arm64 assembly
Russ Cox [Mon, 1 Feb 2021 20:12:08 +0000 (15:12 -0500)]
runtime: convert windows/arm64 assembly

The assembly is mostly a straightforward conversion of the
equivalent arm assembly.

This CL is part of a stack adding windows/arm64
support (#36439), intended to land in the Go 1.17 cycle.

Change-Id: I61b15d712ade4d3a7285c7680de8e0987aacba10
Reviewed-on: https://go-review.googlesource.com/c/go/+/288828
Trust: Russ Cox <rsc@golang.org>
Trust: Jason A. Donenfeld <Jason@zx2c4.com>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
4 years agoruntime: initial windows/arm64 implementation files
Russ Cox [Fri, 22 Jan 2021 15:30:10 +0000 (10:30 -0500)]
runtime: initial windows/arm64 implementation files

This CL adds a few small files - defs, os, and rt0 - to start
on windows/arm64 support for the runtime.

It also copies sys_windows_arm.s to sys_windows_arm64.s,
with the addition of "#ifdef NOT_PORTED" around the entire file.
This is meant to make future CLs easier to review, since the
general pattern is to translate the 32-bit ARM assembly into
64-bit ARM assembly.

This CL is part of a stack adding windows/arm64
support (#36439), intended to land in the Go 1.17 cycle.

Change-Id: I922037eb3890e77bac48281ecaa8e489595675be
Reviewed-on: https://go-review.googlesource.com/c/go/+/288827
Trust: Russ Cox <rsc@golang.org>
Trust: Jason A. Donenfeld <Jason@zx2c4.com>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Reviewed-by: Alex Brainman <alex.brainman@gmail.com>
Reviewed-by: Jason A. Donenfeld <Jason@zx2c4.com>
4 years agoruntime: generate windows/arm64 callback asm
Russ Cox [Fri, 22 Jan 2021 20:16:13 +0000 (15:16 -0500)]
runtime: generate windows/arm64 callback asm

This CL is part of a stack adding windows/arm64
support (#36439), intended to land in the Go 1.17 cycle.

Change-Id: I5e2b589797808626bcca771cdf860d5cb85586cb
Reviewed-on: https://go-review.googlesource.com/c/go/+/288826
Trust: Russ Cox <rsc@golang.org>
Trust: Jason A. Donenfeld <Jason@zx2c4.com>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Reviewed-by: Alex Brainman <alex.brainman@gmail.com>
Reviewed-by: Jason A. Donenfeld <Jason@zx2c4.com>
4 years agosyscall: add windows/arm64 support
Russ Cox [Wed, 27 Jan 2021 02:18:01 +0000 (21:18 -0500)]
syscall: add windows/arm64 support

types_windows_arm64.go is a copy of types_windows_amd64.go.
All that matters for these types seems to be that they are 64-bit vs 32-bit.

This CL is part of a stack adding windows/arm64
support (#36439), intended to land in the Go 1.17 cycle.

Change-Id: Ia7788d6e88e5db899371c75dc7dea7d912a689ad
Reviewed-on: https://go-review.googlesource.com/c/go/+/288825
Trust: Russ Cox <rsc@golang.org>
Trust: Jason A. Donenfeld <Jason@zx2c4.com>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Reviewed-by: Alex Brainman <alex.brainman@gmail.com>
Reviewed-by: Jason A. Donenfeld <Jason@zx2c4.com>
4 years agocmd/vendor: get golang.org/x/sys@beda7e5e158
Russ Cox [Wed, 27 Jan 2021 02:23:55 +0000 (21:23 -0500)]
cmd/vendor: get golang.org/x/sys@beda7e5e158

This brings in the windows/arm64 support, along with other recent changes.

This CL is part of a stack adding windows/arm64
support (#36439), intended to land in the Go 1.17 cycle.

Change-Id: Ifaca710093376c658ecf91239aa05cc33af98a31
Reviewed-on: https://go-review.googlesource.com/c/go/+/288824
Trust: Russ Cox <rsc@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Reviewed-by: Alex Brainman <alex.brainman@gmail.com>
4 years agotest: disable nilptr on windows/arm64
Russ Cox [Wed, 27 Jan 2021 06:01:18 +0000 (01:01 -0500)]
test: disable nilptr on windows/arm64

The address space starts at 4GB, so dummy is too far out.

This CL is part of a stack adding windows/arm64
support (#36439), intended to land in the Go 1.17 cycle.

Change-Id: I5f67e268ce729086d9f9fc8541722fabccfd0145
Reviewed-on: https://go-review.googlesource.com/c/go/+/288823
Trust: Russ Cox <rsc@golang.org>
Trust: Jason A. Donenfeld <Jason@zx2c4.com>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Reviewed-by: Alex Brainman <alex.brainman@gmail.com>
Reviewed-by: Jason A. Donenfeld <Jason@zx2c4.com>
4 years agocmd/link: add windows/arm64 support
Russ Cox [Wed, 27 Jan 2021 17:35:16 +0000 (12:35 -0500)]
cmd/link: add windows/arm64 support

This CL is part of a stack adding windows/arm64
support (#36439), intended to land in the Go 1.17 cycle.

Change-Id: I397c1d238bb18cbe78b3fca00910660cf1d66b8d
Reviewed-on: https://go-review.googlesource.com/c/go/+/288822
Trust: Russ Cox <rsc@golang.org>
Trust: Jason A. Donenfeld <Jason@zx2c4.com>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Reviewed-by: Alex Brainman <alex.brainman@gmail.com>
Reviewed-by: Jason A. Donenfeld <Jason@zx2c4.com>
4 years agocmd/internal/objfile: recognize Windows ARM64 executables
Russ Cox [Fri, 22 Jan 2021 14:23:49 +0000 (09:23 -0500)]
cmd/internal/objfile: recognize Windows ARM64 executables

This CL is part of a stack adding windows/arm64
support (#36439), intended to land in the Go 1.17 cycle.

Change-Id: I5ec459063d394b9f434d1b8b5030960b45061038
Reviewed-on: https://go-review.googlesource.com/c/go/+/288821
Trust: Russ Cox <rsc@golang.org>
Trust: Jason A. Donenfeld <Jason@zx2c4.com>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Reviewed-by: Alex Brainman <alex.brainman@gmail.com>
Reviewed-by: Jason A. Donenfeld <Jason@zx2c4.com>
4 years agodebug/pe: recognize arm64 executables
Russ Cox [Fri, 22 Jan 2021 14:32:38 +0000 (09:32 -0500)]
debug/pe: recognize arm64 executables

We still need to add test data, but as yet we haven't identified
a good Windows arm64 compiler to produce small binaries.

This CL is part of a stack adding windows/arm64
support (#36439), intended to land in the Go 1.17 cycle.

Change-Id: Ifbecb9a6e25f7af38e20b7d7830df7f5efe2798a
Reviewed-on: https://go-review.googlesource.com/c/go/+/288820
Trust: Russ Cox <rsc@golang.org>
Trust: Jason A. Donenfeld <Jason@zx2c4.com>
Run-TryBot: Russ Cox <rsc@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Reviewed-by: Alex Brainman <alex.brainman@gmail.com>
Reviewed-by: Jason A. Donenfeld <Jason@zx2c4.com>
4 years agoMerge "[dev.typeparams] all: merge master (eb98272) into dev.typeparams" into dev...
Gerrit Code Review [Fri, 19 Feb 2021 00:36:17 +0000 (00:36 +0000)]
Merge "[dev.typeparams] all: merge master (eb98272) into dev.typeparams" into dev.typeparams

4 years agocmd/dist: add windows/arm64 support
Russ Cox [Fri, 22 Jan 2021 14:25:40 +0000 (09:25 -0500)]
cmd/dist: add windows/arm64 support

- Add Windows SystemInfo constant for arm64
- Add windows/arm64 to GOOS/GOARCH list

This CL is part of a stack adding windows/arm64
support (#36439), intended to land in the Go 1.17 cycle.

Change-Id: I6109bd87512b5cb1d227d7a85fd0ac20eb2259e5
Reviewed-on: https://go-review.googlesource.com/c/go/+/288819
Trust: Russ Cox <rsc@golang.org>
Trust: Jason A. Donenfeld <Jason@zx2c4.com>
Reviewed-by: Alex Brainman <alex.brainman@gmail.com>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
4 years agobuild: set GOPATH consistently in run.bash, run.bat, run.rc
Russ Cox [Wed, 27 Jan 2021 02:14:43 +0000 (21:14 -0500)]
build: set GOPATH consistently in run.bash, run.bat, run.rc

We used to clear GOPATH in all the build scripts.
Clearing GOPATH is misleading at best, since you just end up
with the default GOPATH (%USERPROFILE%\go on Windows).
Unless that's your GOROOT, in which case you end up with a
fatal error from the go command (#43938).

run.bash changed to setting GOPATH=/dev/null, which has no
clear analogue on Windows.

run.rc still clears GOPATH.

Change them all to set GOPATH to a non-existent directory
/nonexist-gopath or c:\nonexist-gopath.

Change-Id: I51edd66d37ff6a891b0d0541d91ecba97fbbb03d
Reviewed-on: https://go-review.googlesource.com/c/go/+/288818
Trust: Russ Cox <rsc@golang.org>
Trust: Jason A. Donenfeld <Jason@zx2c4.com>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Reviewed-by: Alex Brainman <alex.brainman@gmail.com>
Reviewed-by: Jason A. Donenfeld <Jason@zx2c4.com>
4 years agocmd/link: add debug print in deadcode
Russ Cox [Fri, 22 Jan 2021 16:13:32 +0000 (11:13 -0500)]
cmd/link: add debug print in deadcode

This matches the prints that deadcode prints later
as the algorithm progresses under -v=2.
It helps to see the initial conditions with -v=2 as well.

Change-Id: I06ae86fe9bd8314d003148f3d941832c9b10aef1
Reviewed-on: https://go-review.googlesource.com/c/go/+/288817
Trust: Russ Cox <rsc@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
4 years agocmd/link: clean up windows PE generation
Russ Cox [Wed, 27 Jan 2021 16:50:58 +0000 (11:50 -0500)]
cmd/link: clean up windows PE generation

A bunch of places are a bit too picky about the architecture.
Simplify them.

Also use a large PEBASE for 64-bit systems.
This more closely matches what is usually used on Windows x86-64
and is required for Windows arm64.
Unfortunately, we still need a special case for x86-64 because
of some cgo relocations. This may be fixable separately.

This CL is part of a stack adding windows/arm64
support (#36439), intended to land in the Go 1.17 cycle.
This CL is, however, not windows/arm64-specific.
It is cleanup meant to make the port (and future ports) easier.

Change-Id: I65212d28ad4d8c40e2b70dc29f7fce072babecb5
Reviewed-on: https://go-review.googlesource.com/c/go/+/288816
Trust: Russ Cox <rsc@golang.org>
Trust: Jason A. Donenfeld <Jason@zx2c4.com>
Reviewed-by: Alex Brainman <alex.brainman@gmail.com>
Reviewed-by: Jason A. Donenfeld <Jason@zx2c4.com>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
4 years agosyscall: clean up windows a bit
Russ Cox [Wed, 27 Jan 2021 16:39:48 +0000 (11:39 -0500)]
syscall: clean up windows a bit

The files being deleted contain no code.
They exist because back when we used Makefiles
that listed all the Go sources to be compiled, we wrote
patterns like syscall_$GOOS_$GOARCH.go,
and it was easier to create dummy empty files
than introduce conditionals to not look for that
file on Windows.

Now that we have the go command instead,
we don't need those dummy files.

This CL is part of a stack adding windows/arm64
support (#36439), intended to land in the Go 1.17 cycle.
This CL is, however, not windows/arm64-specific.
It is cleanup meant to make the port (and future ports) easier.

Change-Id: Ie0066d1dd2bf09802c74c6a496276e8c593e4bc2
Reviewed-on: https://go-review.googlesource.com/c/go/+/288815
Trust: Russ Cox <rsc@golang.org>
Trust: Jason A. Donenfeld <Jason@zx2c4.com>
Reviewed-by: Alex Brainman <alex.brainman@gmail.com>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Reviewed-by: Jason A. Donenfeld <Jason@zx2c4.com>
4 years agoruntime: enable framepointer on all arm64
Russ Cox [Wed, 27 Jan 2021 16:34:42 +0000 (11:34 -0500)]
runtime: enable framepointer on all arm64

Frame pointers were already enabled on linux, darwin, ios,
but not freebsd, android, openbsd, netbsd.

But the space was reserved on all platforms, leading to
two different arm64 framepointer conditions in different
parts of the code, one of which had no name
(framepointer_enabled || GOARCH == "arm64",
which might have been "framepointer_space_reserved").

So on the disabled systems, the stack layouts were still
set up for frame pointers and the only difference was not
actually maintaining the FP register in the generated code.

Reduce complexity by just enabling the frame pointer
completely on all the arm64 systems.

This commit passes on freebsd, android, netbsd.
I have not been able to try it on openbsd.

This CL is part of a stack adding windows/arm64
support (#36439), intended to land in the Go 1.17 cycle.
This CL is, however, not windows/arm64-specific.
It is cleanup meant to make the port (and future ports) easier.

Change-Id: I83bd23369d24b76db4c6a648fa74f6917819a093
Reviewed-on: https://go-review.googlesource.com/c/go/+/288814
Trust: Russ Cox <rsc@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
4 years agoruntime: clean up windows a bit
Russ Cox [Thu, 28 Jan 2021 14:23:35 +0000 (09:23 -0500)]
runtime: clean up windows a bit

Document the various hard-coded architecture checks
or remove them in favor of more general checks.
This should be a no-op now but will make the arm64 port
have fewer diffs.

This CL is part of a stack adding windows/arm64
support (#36439), intended to land in the Go 1.17 cycle.
This CL is, however, not windows/arm64-specific.
It is cleanup meant to make the port (and future ports) easier.

Change-Id: Ifd6b19e44e8c9ca4a0d2590f314928ce235821b3
Reviewed-on: https://go-review.googlesource.com/c/go/+/288813
Trust: Russ Cox <rsc@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Reviewed-by: Alex Brainman <alex.brainman@gmail.com>
4 years agoruntime: fix windows/arm externalthreadhandler
Russ Cox [Wed, 27 Jan 2021 16:27:25 +0000 (11:27 -0500)]
runtime: fix windows/arm externalthreadhandler

Externalthreadhandler was not handling its own stack correctly.
It incorrectly referred to the saved LR slot (uninitialized, it turned out)
as holding the return value from the called function.

Externalthreadhandler is used to call two different functions:
profileloop1 and ctrlhandler1.
Profileloop1 does not return, so no harm done.
Ctrlhandler1 returns a boolean indicating whether the handler
took care of the control event (if true, no other handlers run).

It's hard to say exactly what uninitialized values are likely to
have been returned instead of ctrlhandler1's result, but it
probably wasn't helping matters.

Change-Id: Ia02f1c033df618cb82c2193b3a8241ed048a8b18
Reviewed-on: https://go-review.googlesource.com/c/go/+/288812
Trust: Russ Cox <rsc@golang.org>
Trust: Jason A. Donenfeld <Jason@zx2c4.com>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Reviewed-by: Jason A. Donenfeld <Jason@zx2c4.com>
4 years agoruntime: fix time on windows/arm under WINE
Russ Cox [Wed, 27 Jan 2021 16:22:02 +0000 (11:22 -0500)]
runtime: fix time on windows/arm under WINE

This code has clearly never run successfully,
since one of the “tail calls" calls the wrong function,
and both of them appear in functions with stack frames
that are never going to be properly unwound.
Probably there is no windows/arm under WINE at all.
But might as well fix the code.

Change-Id: I5fa62274b3661bc6bce098657b5bcf11d59655eb
Reviewed-on: https://go-review.googlesource.com/c/go/+/288811
Trust: Russ Cox <rsc@golang.org>
Trust: Jason A. Donenfeld <Jason@zx2c4.com>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Reviewed-by: Jason A. Donenfeld <Jason@zx2c4.com>
4 years agoruntime: crash earlier on windows for runtime.abort
Russ Cox [Wed, 27 Jan 2021 16:15:33 +0000 (11:15 -0500)]
runtime: crash earlier on windows for runtime.abort

The isAbort check was wrong for non-x86 systems.
That was causing the exception chain to be passed back to Windows.
That was causing some other kind of fault - not sure what.
That was leading back to lastcontinuehandler to print a larger
stack trace, and then the throwing = 1 print added runtime.abort,
which made TestAbort pass even though it wasn't really working.

Recognize abort properly and handle it as Go, not as something
for Windows to try to handle.

Keep the throwing = 1 print, because more detail on throw is
always better.

This CL is part of a stack adding windows/arm64
support (#36439), intended to land in the Go 1.17 cycle.
This CL is, however, not windows/arm64-specific.
It is cleanup meant to make the port (and future ports) easier.

Change-Id: If614f4ab2884bd90410d29e28311bf969ceeac09
Reviewed-on: https://go-review.googlesource.com/c/go/+/288810
Trust: Russ Cox <rsc@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
4 years agoruntime: print hex numbers with hex prefixes in traceback debug
Russ Cox [Wed, 27 Jan 2021 02:26:01 +0000 (21:26 -0500)]
runtime: print hex numbers with hex prefixes in traceback debug

If traceback fails, it prints a helpful hex dump of the stack.
But the hex numbers have no 0x prefix, which might make it
a little unclear that they are hex.
We only print two per line, so there is plenty of room for the 0x.
Print it, which lets us delete a custom hex formatter.

Also, in the translated <name+off> hints, print off in hex
(with a 0x prefix). The offsets were previously decimal, which
could have been confused for hex since none of the hex had
0x prefixes. And decimal is kind of useless anyway since the
offsets shown in the main traceback are hex, so you can't
easily match them up without mental base conversions.

Just print hex everywhere, clearly marked by 0x.

This CL is part of a stack adding windows/arm64
support (#36439), intended to land in the Go 1.17 cycle.
This CL is, however, not windows/arm64-specific.
It is cleanup meant to make the port (and future ports) easier.

Change-Id: I72d26a4e41ada38b620bf8fe3576d787a2e59b47
Reviewed-on: https://go-review.googlesource.com/c/go/+/288809
Trust: Russ Cox <rsc@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
4 years agoruntime: fix windows/arm CONTEXT_CONTROL
Russ Cox [Wed, 27 Jan 2021 16:02:33 +0000 (11:02 -0500)]
runtime: fix windows/arm CONTEXT_CONTROL

The constant was wrong, and the “right” constant doesn't work either.
But with the actually-right constant (and possibly earlier fixes in this
stack as well), profiling now works.

Change-Id: If8caff1da556826db40961fb9bcfe2b1f31ea9f9
Reviewed-on: https://go-review.googlesource.com/c/go/+/288808
Trust: Russ Cox <rsc@golang.org>
Trust: Jason A. Donenfeld <Jason@zx2c4.com>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Reviewed-by: Jason A. Donenfeld <Jason@zx2c4.com>
4 years agoruntime: factor common code out of defs_windows_*.go
Russ Cox [Wed, 27 Jan 2021 16:01:00 +0000 (11:01 -0500)]
runtime: factor common code out of defs_windows_*.go

Also give up on the fiction that these files can be regenerated.
They contain many manual edits, and they're fairly small anyway.

This CL is part of a stack adding windows/arm64
support (#36439), intended to land in the Go 1.17 cycle.
This CL is, however, not windows/arm64-specific.
It is cleanup meant to make the port (and future ports) easier.

Change-Id: Ib4e4e20a43d8beb1d5390fd184160c33607641f6
Reviewed-on: https://go-review.googlesource.com/c/go/+/288807
Trust: Russ Cox <rsc@golang.org>
Trust: Jason A. Donenfeld <Jason@zx2c4.com>
Reviewed-by: Alex Brainman <alex.brainman@gmail.com>
Reviewed-by: Jason A. Donenfeld <Jason@zx2c4.com>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
4 years agoruntime: find g in Windows profiler using SP
Russ Cox [Wed, 27 Jan 2021 16:06:04 +0000 (11:06 -0500)]
runtime: find g in Windows profiler using SP

The architecture-specific interpretation of m->tls[0]
is unnecessary and fragile. Delete it.

This CL is part of a stack adding windows/arm64
support (#36439), intended to land in the Go 1.17 cycle.
This CL is, however, not windows/arm64-specific.
It is cleanup meant to make the port (and future ports) easier.

Change-Id: I927345e52fa2f1741d4914478a29d1fb8acb0dc3
Reviewed-on: https://go-review.googlesource.com/c/go/+/288806
Trust: Russ Cox <rsc@golang.org>
Trust: Jason A. Donenfeld <Jason@zx2c4.com>
Reviewed-by: Jason A. Donenfeld <Jason@zx2c4.com>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
4 years agoruntime: do not treat asmcgocall as a topofstack on g0
Russ Cox [Fri, 29 Jan 2021 14:55:03 +0000 (09:55 -0500)]
runtime: do not treat asmcgocall as a topofstack on g0

This was added in 2018 to fix a runtime crash during unwind
during a unhandled-panic-induced crash.
(See https://golang.org/cl/90895 and #23576.)
Clearly we cannot unwind past this function, and the change
did stop the unwind. But it's not a top-of-stack function, and
the real issue is that SP is changed.

The new SPWRITE bit takes care of this instead, so we can drop
it from the topofstack function.

At this point the topofstack function is only checking the
TOPFRAME bit, so we can inline that into the one call site.

This CL is part of a stack adding windows/arm64
support (#36439), intended to land in the Go 1.17 cycle.
This CL is, however, not windows/arm64-specific.
It is cleanup meant to make the port (and future ports) easier.

Change-Id: I856552298032770e48e06c95a20823a1dbd5e38c
Reviewed-on: https://go-review.googlesource.com/c/go/+/288805
Trust: Russ Cox <rsc@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
4 years agoruntime: do not treat morestack as a topofstack
Russ Cox [Fri, 29 Jan 2021 14:05:10 +0000 (09:05 -0500)]
runtime: do not treat morestack as a topofstack

I added morestack to this list in 2013 with an explanation
that they were needed if we “start a garbage collection on g0
during a stack split or unsplit”.
(https://golang.org/cl/11533043)

This explanation no longer applies for a handful of reasons,
most importantly that if we did stop a stack scan in the middle
of a call to morestack, we'd ignore pointers above the split,
which would lead to memory corruption. But we don't scan
goroutine stacks during morestack now, so that can't happen.
If we did see morestack during a GC, that would be a good time
to crash the program.

The real problem with morestack is during profiling, as noted
in the code review conversation during 2013. And in profiling
we just need to know to stop and not unwind further, which
the new SPWRITE bit will do for us.

So remove from topofstack and let the program crash if GC
sees morestack and otherwise let the SPWRITE stop morestack
unwinding during profiling.

This CL is part of a stack adding windows/arm64
support (#36439), intended to land in the Go 1.17 cycle.
This CL is, however, not windows/arm64-specific.
It is cleanup meant to make the port (and future ports) easier.

Change-Id: I06d95920b18c599c7c46f64c21028104978215d7
Reviewed-on: https://go-review.googlesource.com/c/go/+/288804
Trust: Russ Cox <rsc@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
4 years agoruntime: do not treat mcall as a topofstack
Russ Cox [Thu, 28 Jan 2021 22:17:38 +0000 (17:17 -0500)]
runtime: do not treat mcall as a topofstack

I added mcall to this list in 2013 without explaining why.
(https://codereview.appspot.com/11085043/diff/61001/src/pkg/runtime/traceback_x86.c)
I suspect I was stopping crashes during profiling where the unwind
tried to walk up past mcall and got confused.

mcall is not something you can unwind past, because it switches
stacks, but it's also not something you should expect as a
standard top-of-stack frame. So if you do see it during say
a garbage collection stack walk, it would be important to crash
instead of silently stopping the walk prematurely.

This CL removes it from the topofstack list to avoid the silent stop.
Now that mcall is detected as SPWRITE, that will stop the
unwind (with a crash if encountered during GC, which we want).

This CL is part of a stack adding windows/arm64
support (#36439), intended to land in the Go 1.17 cycle.
This CL is, however, not windows/arm64-specific.
It is cleanup meant to make the port (and future ports) easier.

Change-Id: I666487ce24efd72292f2bc3eac7fe0477e16bddd
Reviewed-on: https://go-review.googlesource.com/c/go/+/288803
Trust: Russ Cox <rsc@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
4 years agoruntime: use TOPFRAME to identify top-of-frame functions
Russ Cox [Thu, 28 Jan 2021 21:22:52 +0000 (16:22 -0500)]
runtime: use TOPFRAME to identify top-of-frame functions

No change to actual runtime, but helps reduce the laundry list
of functions.

mcall, morestack, and asmcgocall are not actually top-of-frame,
so those need more attention in follow-up CLs.

mstart moved to assembly so that it can be marked TOPFRAME.

Since TOPFRAME also tells DWARF consumers not to unwind
this way, this change should also improve debuggers a
marginal amount.

This CL is part of a stack adding windows/arm64
support (#36439), intended to land in the Go 1.17 cycle.
This CL is, however, not windows/arm64-specific.
It is cleanup meant to make the port (and future ports) easier.

Change-Id: If1e0d46ca973de5e46b62948d076f675f285b5d9
Reviewed-on: https://go-review.googlesource.com/c/go/+/288802
Trust: Russ Cox <rsc@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
4 years agoruntime: use FuncInfo SPWRITE flag to identify untraceable profile samples
Russ Cox [Thu, 28 Jan 2021 21:10:58 +0000 (16:10 -0500)]
runtime: use FuncInfo SPWRITE flag to identify untraceable profile samples

The old code was very clever about predicting whether a traceback was safe.
That cleverness has not aged well. In particular, the setsSP function is missing
a bunch of functions that write to SP and will confuse traceback.
And one such function - jmpdefer - was handled as a special case in
gentraceback instead of simply listing it in setsSP.

Throw away all the clever prediction about whether traceback will crash.
Instead, make traceback NOT crash, by checking whether the function
being walked writes to SP.

This CL is part of a stack adding windows/arm64
support (#36439), intended to land in the Go 1.17 cycle.
This CL is, however, not windows/arm64-specific.
It is cleanup meant to make the port (and future ports) easier.

Change-Id: I3d55fe257a22745e4919ac4dc9a9378c984ba0da
Reviewed-on: https://go-review.googlesource.com/c/go/+/288801
Trust: Russ Cox <rsc@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
4 years agocmd/asm, cmd/link, runtime: introduce FuncInfo flag bits
Russ Cox [Thu, 28 Jan 2021 20:21:33 +0000 (15:21 -0500)]
cmd/asm, cmd/link, runtime: introduce FuncInfo flag bits

The runtime traceback code has its own definition of which functions
mark the top frame of a stack, separate from the TOPFRAME bits that
exist in the assembly and are passed along in DWARF information.
It's error-prone and redundant to have two different sources of truth.
This CL provides the actual TOPFRAME bits to the runtime, so that
the runtime can use those bits instead of reinventing its own category.

This CL also adds a new bit, SPWRITE, which marks functions that
write directly to SP (anything but adding and subtracting constants).
Such functions must stop a traceback, because the traceback has no
way to rederive the SP on entry. Again, the runtime has its own definition
which is mostly correct, but also missing some functions. During ordinary
goroutine context switches, such functions do not appear on the stack,
so the incompleteness in the runtime usually doesn't matter.
But profiling signals can arrive at any moment, and the runtime may
crash during traceback if it attempts to unwind an SP-writing frame
and gets out-of-sync with the actual stack. The runtime contains code
to try to detect likely candidates but again it is incomplete.
Deriving the SPWRITE bit automatically from the actual assembly code
provides the complete truth, and passing it to the runtime lets the
runtime use it.

This CL is part of a stack adding windows/arm64
support (#36439), intended to land in the Go 1.17 cycle.
This CL is, however, not windows/arm64-specific.
It is cleanup meant to make the port (and future ports) easier.

Change-Id: I227f53b23ac5b3dabfcc5e8ee3f00df4e113cf58
Reviewed-on: https://go-review.googlesource.com/c/go/+/288800
Trust: Russ Cox <rsc@golang.org>
Trust: Jason A. Donenfeld <Jason@zx2c4.com>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Reviewed-by: Jason A. Donenfeld <Jason@zx2c4.com>
4 years agoruntime: remove unnecessary writes to gp.sched.g
Russ Cox [Mon, 15 Feb 2021 14:25:55 +0000 (09:25 -0500)]
runtime: remove unnecessary writes to gp.sched.g

A g's sched.g is set in newproc1:

newg.sched.g = guintptr(unsafe.Pointer(newg))

After that, it never changes. Yet lots of assembly code does
"g.sched.g = g" unnecessarily. Remove all those lines to avoid
confusion about whether it ever changes.

Also, split gogo into two functions, one that does the nil g check
and a second that does the actual switch. This way, if the nil g check
fails, we get a stack trace showing the call stack that led to the failure.
(The SP write would otherwise cause the stack trace to abort.)

Also restore the proper nil g check in a handful of assembly functions.
(There is little point in checking for nil g *after* installing it as the real g.)

Change-Id: I22866b093f901f765de1d074e36eeec10366abfb
Reviewed-on: https://go-review.googlesource.com/c/go/+/292109
Trust: Russ Cox <rsc@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
4 years agocmd/internal/obj/riscv: fix JMP name<>(SB)
Russ Cox [Mon, 15 Feb 2021 18:58:45 +0000 (13:58 -0500)]
cmd/internal/obj/riscv: fix JMP name<>(SB)

It was being rejected. Now it isn't and can be used in the runtime.

Change-Id: I4626bf9fc2e0bc26fffb87d11bede459964324b3
Reviewed-on: https://go-review.googlesource.com/c/go/+/292129
Trust: Russ Cox <rsc@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
4 years agoruntime: unify asmcgocall and systemstack traceback setup
Russ Cox [Tue, 2 Feb 2021 03:58:28 +0000 (22:58 -0500)]
runtime: unify asmcgocall and systemstack traceback setup

Both asmcgocall and systemstack need to save the calling Go code's
context for use by traceback, but they do it differently.
Systemstack's appraoch is better, because it doesn't require a
special case in traceback.
So make them both use that.

While we are here, the fake mstart caller in systemstack is
no longer needed and can be removed.
(traceback knows to stop in systemstack because of the writes to SP.)

Also remove the fake mstarts in sys_windows_*.s.

And while we are there, fix the control flow guard code in sys_windows_arm.s.
The current code is using pointers to a stack frame that technically is gone
once we hit the RET instruction. Clearly it's working OK, but better not to depend
on data below SP being preserved, even for just a few instructions.
Store the value we need in other registers instead.
(This code is only used for pushing a sigpanic call, which does not
actually return to the site of the fault and therefore doesn't need to
preserve any of the registers.)

This CL is part of a stack adding windows/arm64
support (#36439), intended to land in the Go 1.17 cycle.
This CL is, however, not windows/arm64-specific.
It is cleanup meant to make the port (and future ports) easier.

Change-Id: Id1e3ef5e54f7ad786e4b87043f2626eba7c3bbd9
Reviewed-on: https://go-review.googlesource.com/c/go/+/288799
Trust: Russ Cox <rsc@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
4 years agoruntime: clean up funcID assignment
Russ Cox [Thu, 28 Jan 2021 14:32:55 +0000 (09:32 -0500)]
runtime: clean up funcID assignment

Large enum sets should be sorted by name when the
values don't matter, as they don't here. Do that.

Also replace the large switch with a map lookup.

This CL is part of a stack adding windows/arm64
support (#36439), intended to land in the Go 1.17 cycle.
This CL is, however, not windows/arm64-specific.
It is cleanup meant to make the port (and future ports) easier.

Change-Id: Ibe727b5d8866bf4c40c96020e1f4632bde7efd59
Reviewed-on: https://go-review.googlesource.com/c/go/+/288798
Trust: Russ Cox <rsc@golang.org>
Trust: Jason A. Donenfeld <Jason@zx2c4.com>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Reviewed-by: Jason A. Donenfeld <Jason@zx2c4.com>
Reviewed-by: Ian Lance Taylor <iant@golang.org>