]> Cypherpunks repositories - gostls13.git/log
gostls13.git
4 years agoio/fs: use testing.T.TempDir in TestWalkDir
Tobias Klauser [Sat, 13 Mar 2021 17:00:17 +0000 (18:00 +0100)]
io/fs: use testing.T.TempDir in TestWalkDir

Change-Id: I805ad51332e4efe27d47f6c6e3b0af945e0d4aa0
Reviewed-on: https://go-review.googlesource.com/c/go/+/301489
Trust: Tobias Klauser <tobias.klauser@gmail.com>
Run-TryBot: Tobias Klauser <tobias.klauser@gmail.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Emmanuel Odeke <emmanuel@orijtech.com>
4 years agocmd/compile/internal/ssa: prealloc slice
cui [Mon, 11 Jan 2021 08:44:45 +0000 (08:44 +0000)]
cmd/compile/internal/ssa: prealloc slice

Change-Id: I9943a4f931c251a69bc8244c0d7723a0a3552073
GitHub-Last-Rev: d9dd94ae4444cb0106756cdb98c1c5fa12fa5f79
GitHub-Pull-Request: golang/go#43622
Reviewed-on: https://go-review.googlesource.com/c/go/+/282992
Reviewed-by: Keith Randall <khr@golang.org>
Run-TryBot: Keith Randall <khr@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>
Trust: Emmanuel Odeke <emmanuel@orijtech.com>

4 years agoio: add error check to WriteString Example test
JulianChu [Sat, 13 Mar 2021 17:44:10 +0000 (17:44 +0000)]
io: add error check to WriteString Example test

Change-Id: I9ce1c79e5799f205aec3a4dc02645ed26bdc3581
GitHub-Last-Rev: 59b637db0154e55ddfdd55e54b9596dc3a0ad32d
GitHub-Pull-Request: golang/go#44533
Reviewed-on: https://go-review.googlesource.com/c/go/+/295389
Reviewed-by: Emmanuel Odeke <emmanuel@orijtech.com>
Reviewed-by: Robert Griesemer <gri@golang.org>
Trust: Emmanuel Odeke <emmanuel@orijtech.com>
Trust: Robert Griesemer <gri@golang.org>
Run-TryBot: Emmanuel Odeke <emmanuel@orijtech.com>
TryBot-Result: Go Bot <gobot@golang.org>

4 years agofmt: use “truncateString” not “truncate” in method doc
zfCode [Sat, 13 Mar 2021 18:00:30 +0000 (18:00 +0000)]
fmt: use “truncateString” not “truncate” in method doc

Change-Id: If1acb6a8533a782f80c7d1f0ad5155e98e1134dd
GitHub-Last-Rev: 03384a3d99dd89d802635f7ef48ce4456ec338b0
GitHub-Pull-Request: golang/go#44375
Reviewed-on: https://go-review.googlesource.com/c/go/+/293629
Reviewed-by: Emmanuel Odeke <emmanuel@orijtech.com>
Reviewed-by: Rob Pike <r@golang.org>
Trust: Rob Pike <r@golang.org>

4 years agoerrors/wrap: do not call Elem() twice
Vitaly Zdanevich [Mon, 8 Mar 2021 03:15:25 +0000 (03:15 +0000)]
errors/wrap: do not call Elem() twice

Change-Id: I2fe6037c45a0dfe25f946a92ff97b5e3fbd69bc0
GitHub-Last-Rev: 644d479a27c0eccfc0b37e1a560ca09e47b5a972
GitHub-Pull-Request: golang/go#44851
Reviewed-on: https://go-review.googlesource.com/c/go/+/299629
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Trust: Tobias Klauser <tobias.klauser@gmail.com>

4 years agoall: remove duplicate words
John Bampton [Sat, 13 Mar 2021 11:25:15 +0000 (11:25 +0000)]
all: remove duplicate words

Change-Id: Ib0469232a2b69a869e58d5d24990ad74ac96ea56
GitHub-Last-Rev: eb38e049ee1e773392ff3747e1eb2af20dd50dcd
GitHub-Pull-Request: golang/go#44805
Reviewed-on: https://go-review.googlesource.com/c/go/+/299109
Trust: Emmanuel Odeke <emmanuel@orijtech.com>
Run-TryBot: Emmanuel Odeke <emmanuel@orijtech.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Bryan C. Mills <bcmills@google.com>
4 years agoencoding/gob: ensure "duplicate type received" decoder errors surface up
Aman Karmani [Tue, 2 Mar 2021 19:52:34 +0000 (11:52 -0800)]
encoding/gob: ensure "duplicate type received" decoder errors surface up

Previously re-using a decoder with a new stream resulted in a confusing
"extra data in buffer" error message.

Change-Id: Ia4c4c3a2d4b63c59e37e53faa61a500d5ff6e5f1
Reviewed-on: https://go-review.googlesource.com/c/go/+/297949
Reviewed-by: Rob Pike <r@golang.org>
Run-TryBot: Rob Pike <r@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>
Trust: Emmanuel Odeke <emmanuel@orijtech.com>

4 years agocrypto/ecdsa: fix dead reference link
Mostyn Bramley-Moore [Tue, 16 Feb 2021 16:01:39 +0000 (16:01 +0000)]
crypto/ecdsa: fix dead reference link

The previous link broke, but it's available on the internet archive.

Fixes #39808

Change-Id: Ic2be74a1f0591600ca1acbe08e1bab8ba1e21abe
GitHub-Last-Rev: 6d6de5d2f451c6d53a1e55b62fb5a1fab0d49f10
GitHub-Pull-Request: golang/go#40165
Reviewed-on: https://go-review.googlesource.com/c/go/+/242103
Trust: Emmanuel Odeke <emmanuel@orijtech.com>
Run-TryBot: Emmanuel Odeke <emmanuel@orijtech.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Filippo Valsorda <filippo@golang.org>
4 years agonet/http: note that "HTTP/2" is invalid for ParseHTTPVersion
Josh Deprez [Tue, 10 Nov 2020 21:27:04 +0000 (21:27 +0000)]
net/http: note that "HTTP/2" is invalid for ParseHTTPVersion

Change-Id: Ieba05dea892ec9855a63b80e456bcf9188eef855
GitHub-Last-Rev: 5f7663ac4aaecb01a27a04309277240fd15759c9
GitHub-Pull-Request: golang/go#41806
Reviewed-on: https://go-review.googlesource.com/c/go/+/259758
Reviewed-by: Damien Neil <dneil@google.com>
Reviewed-by: Emmanuel Odeke <emmanuel@orijtech.com>
Trust: Damien Neil <dneil@google.com>

4 years agoruntime: fix documented alignment of 32KiB and 64KiB size classes
Matthew Dempsky [Sat, 13 Mar 2021 00:58:10 +0000 (16:58 -0800)]
runtime: fix documented alignment of 32KiB and 64KiB size classes

As Cherry pointed out on golang.org/cl/299909, the page allocator
doesn't guarantee any alignment for multi-page allocations, so object
alignments are thus implicitly capped at page alignment.

Change-Id: I6f5df27f269b095cde54056f876fe4240f69c5c7
Reviewed-on: https://go-review.googlesource.com/c/go/+/301292
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Trust: Matthew Dempsky <mdempsky@google.com>

4 years agoall: use HTML5 br tags
John Bampton [Fri, 5 Mar 2021 01:53:00 +0000 (01:53 +0000)]
all: use HTML5 br tags

In HTML5 br tags don't need a closing slash

Change-Id: Ic53c43faee08c5b1267daa9a02cc186b1c255ca1
GitHub-Last-Rev: 652208116944d01b23b8af8f1af485da5e916d32
GitHub-Pull-Request: golang/go#44283
Reviewed-on: https://go-review.googlesource.com/c/go/+/292370
Reviewed-by: Russ Cox <rsc@golang.org>
Run-TryBot: Russ Cox <rsc@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>
Trust: Emmanuel Odeke <emmanuel@orijtech.com>

4 years agomisc/cgo/testcarchive: don't use == for string equality in C code
Ian Lance Taylor [Thu, 11 Mar 2021 23:09:47 +0000 (15:09 -0800)]
misc/cgo/testcarchive: don't use == for string equality in C code

For https://gcc.gnu.org/PR99553

Change-Id: I29a7fbfd89963d4139bc19af99330d70567938ea
Reviewed-on: https://go-review.googlesource.com/c/go/+/300993
Trust: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Emmanuel Odeke <emmanuel@orijtech.com>
4 years agocmd/compile/internal/types2: use self_test.go from go/types
Robert Griesemer [Fri, 12 Mar 2021 02:10:08 +0000 (18:10 -0800)]
cmd/compile/internal/types2: use self_test.go from go/types

This CL replaces self_test.go with the (improved) version
from go/types, modified for types2.

To see the differences between go/types/self_test.go and
this version, compare against patch set 1.

Change-Id: I7ae830a17f7a0de40cc1f5063166a7247f78ec27
Reviewed-on: https://go-review.googlesource.com/c/go/+/300997
Trust: Robert Griesemer <gri@golang.org>
Reviewed-by: Robert Findley <rfindley@google.com>
4 years agocmd/compile/internal/types2: simplify error reporting API (cleanup)
Robert Griesemer [Fri, 12 Mar 2021 00:34:01 +0000 (16:34 -0800)]
cmd/compile/internal/types2: simplify error reporting API (cleanup)

- Remove specialized errorf functions for invalid AST/argument/operation.
  Instead use prefix constants with the error message.

- Replace several calls to Checker.errorf with calls to Checker.error
  if there are no arguments to format.

- Replace a handful of %s format verbs with %v to satisfy vet check.

- Add a basic test that checks that we're not using Checker.errorf when
  we should be using Checker.error.

Change-Id: I7bc7c14f3cf774689ec8cd5782ea31b6e30dbcd6
Reviewed-on: https://go-review.googlesource.com/c/go/+/300995
Trust: Robert Griesemer <gri@golang.org>
Reviewed-by: Robert Findley <rfindley@google.com>
4 years agocmd/compile: mention that -m can be increased or given multiple times
Michael Schaller [Fri, 12 Mar 2021 12:12:48 +0000 (12:12 +0000)]
cmd/compile: mention that -m can be increased or given multiple times

-m can be increased or it can be given up to 4 times to increase the verbosity of the printed optimization decisions: https://github.com/golang/go/search?q=LowerM

Change-Id: I0cc304385be052664fad455ff075846a3a63f298
GitHub-Last-Rev: 140f08529024dd17c5ca1cbad52dd1d17c6126c0
GitHub-Pull-Request: golang/go#44857
Reviewed-on: https://go-review.googlesource.com/c/go/+/299709
Reviewed-by: Keith Randall <khr@golang.org>
Reviewed-by: Robert Griesemer <gri@golang.org>
4 years agodelete favicon.ico and robots.txt
Dmitri Shuralyov [Wed, 10 Mar 2021 04:27:11 +0000 (23:27 -0500)]
delete favicon.ico and robots.txt

The favicon.ico and robots.txt files have been with us in the root
directory since 2009 and 2011 respectively. Back then, the Go repo
had content for the golang.org website, which could be run locally.
Since these files were at the root of the website, they were added
to the corresponding location in the GOROOT tree—at the root.

In 2018, work started on factoring out golang.org website content
and code into a new golang.org/x/website repository (issue 29206).

The favicon.ico and robots.txt files were copied to x/website repo,
but some more work needed to be done before they would be picked up
and served when golangorg was executed in module mode. That work is
done by now (CL 293413 and CL 293414).

The scope of the godoc tool has also been reduced to just serving
Go package documentation and not the website (issue 32011), so it
can provide its own favicon.ico as needed (CL 300394).

This means these two files have no more use and can be deleted.
So long and goodbye!

Change-Id: Id71bdab6317c1dc481c9d85beaaac4b4eb92d379
Reviewed-on: https://go-review.googlesource.com/c/go/+/300549
Trust: Dmitri Shuralyov <dmitshur@golang.org>
Run-TryBot: Dmitri Shuralyov <dmitshur@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Russ Cox <rsc@golang.org>
4 years agocmd/go: fix godoc formatting for text from 'go help install'
Jay Conrod [Fri, 12 Mar 2021 18:48:32 +0000 (13:48 -0500)]
cmd/go: fix godoc formatting for text from 'go help install'

Fixes #44846

Change-Id: I5a12c6437a91ce59307483ffcc70e084edc32197
Reviewed-on: https://go-review.googlesource.com/c/go/+/301329
Trust: Jay Conrod <jayconrod@google.com>
Run-TryBot: Jay Conrod <jayconrod@google.com>
Reviewed-by: Bryan C. Mills <bcmills@google.com>
Reviewed-by: Dmitri Shuralyov <dmitshur@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>

4 years agocmd/compile: minor cleanup -- remove dead code conditional on test
David Chase [Fri, 12 Mar 2021 15:56:08 +0000 (10:56 -0500)]
cmd/compile: minor cleanup -- remove dead code conditional on test

It would fail now if it were turned on.

Updsates #44816.

Change-Id: I19d94f0cb2dd84271f5304c796d7c81e1e64af25
Reviewed-on: https://go-review.googlesource.com/c/go/+/301270
Trust: David Chase <drchase@google.com>
Run-TryBot: David Chase <drchase@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
4 years agocmd/compile: test register ABI for method, interface, closure calls
David Chase [Fri, 5 Mar 2021 19:24:41 +0000 (14:24 -0500)]
cmd/compile: test register ABI for method, interface, closure calls

This is enabled with a ridiculous magic name for method,
or for last input type passed, that needs to be changed
to something inutterable before actual release.

Ridiculous method name: MagicMethodNameForTestingRegisterABI
Ridiculous last (input) type name: MagicLastTypeNameForTestingRegisterABI

RLTN is tested with strings.Contains, so you can have
MagicLastTypeNameForTestingRegisterABI1
and
MagicLastTypeNameForTestingRegisterABI2
if that is helpful

Includes test test/abi/fibish2.go

Updates #44816.

Change-Id: I592a6edc71ca9bebdd1d00e24edee1ceebb3e43f
Reviewed-on: https://go-review.googlesource.com/c/go/+/299410
Trust: David Chase <drchase@google.com>
Run-TryBot: David Chase <drchase@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
4 years agocmd/go/internal/load: always set IsImportCycle when in a cycle
Roland Shoemaker [Fri, 12 Mar 2021 17:51:50 +0000 (09:51 -0800)]
cmd/go/internal/load: always set IsImportCycle when in a cycle

When hitting an import cycle in reusePackage, and there is already
an error set, make sure IsImportCycle is set so that we don't
end up stuck in a loop.

Fixes #25830

Change-Id: Iba966aea4a637dfc34ee22782a477209ac48c9bd
Reviewed-on: https://go-review.googlesource.com/c/go/+/301289
Trust: Roland Shoemaker <roland@golang.org>
Run-TryBot: Roland Shoemaker <roland@golang.org>
Reviewed-by: Bryan C. Mills <bcmills@google.com>
Reviewed-by: Jay Conrod <jayconrod@google.com>
TryBot-Result: Go Bot <gobot@golang.org>

4 years agoruntime: simplify divmagic for span calculations
Matthew Dempsky [Thu, 11 Mar 2021 23:45:52 +0000 (15:45 -0800)]
runtime: simplify divmagic for span calculations

It's both simpler and faster to just unconditionally do two 32-bit
multiplies rather than a bunch of branching to try to avoid them.
This is safe thanks to the tight bounds derived in [1] and verified
during mksizeclasses.go.

Benchstat results below for compilebench benchmarks on my P920. See
also [2] for micro benchmarks comparing the new functions against the
originals (as well as several basic attempts at optimizing them).

name                      old time/op       new time/op       delta
Template                        295ms ± 3%        290ms ± 1%  -1.95%  (p=0.000 n=20+20)
Unicode                         113ms ± 3%        110ms ± 2%  -2.32%  (p=0.000 n=21+17)
GoTypes                         1.78s ± 1%        1.76s ± 1%  -1.23%  (p=0.000 n=21+20)
Compiler                        119ms ± 2%        117ms ± 4%  -1.53%  (p=0.007 n=20+20)
SSA                             14.3s ± 1%        13.8s ± 1%  -3.12%  (p=0.000 n=17+20)
Flate                           173ms ± 2%        170ms ± 1%  -1.64%  (p=0.000 n=20+19)
GoParser                        278ms ± 2%        273ms ± 2%  -1.92%  (p=0.000 n=20+19)
Reflect                         686ms ± 3%        671ms ± 3%  -2.18%  (p=0.000 n=19+20)
Tar                             255ms ± 2%        248ms ± 2%  -2.90%  (p=0.000 n=20+20)
XML                             335ms ± 3%        327ms ± 2%  -2.34%  (p=0.000 n=20+20)
LinkCompiler                    799ms ± 1%        799ms ± 1%    ~     (p=0.925 n=20+20)
ExternalLinkCompiler            1.90s ± 1%        1.90s ± 0%    ~     (p=0.327 n=20+20)
LinkWithoutDebugCompiler        385ms ± 1%        386ms ± 1%    ~     (p=0.251 n=18+20)
[Geo mean]                      512ms             504ms       -1.61%

name                      old user-time/op  new user-time/op  delta
Template                        286ms ± 4%        282ms ± 4%  -1.42%  (p=0.025 n=21+20)
Unicode                         104ms ± 9%        102ms ±14%    ~     (p=0.294 n=21+20)
GoTypes                         1.75s ± 3%        1.72s ± 2%  -1.36%  (p=0.000 n=21+20)
Compiler                        109ms ±11%        108ms ± 8%    ~     (p=0.187 n=21+19)
SSA                             14.0s ± 1%        13.5s ± 2%  -3.25%  (p=0.000 n=16+20)
Flate                           166ms ± 4%        164ms ± 4%  -1.34%  (p=0.032 n=19+19)
GoParser                        268ms ± 4%        263ms ± 4%  -1.71%  (p=0.011 n=18+20)
Reflect                         666ms ± 3%        654ms ± 4%  -1.77%  (p=0.002 n=18+20)
Tar                             245ms ± 5%        236ms ± 6%  -3.34%  (p=0.000 n=20+20)
XML                             320ms ± 4%        314ms ± 3%  -2.01%  (p=0.001 n=19+18)
LinkCompiler                    744ms ± 4%        747ms ± 3%    ~     (p=0.627 n=20+19)
ExternalLinkCompiler            1.71s ± 3%        1.72s ± 2%    ~     (p=0.149 n=20+20)
LinkWithoutDebugCompiler        345ms ± 6%        342ms ± 8%    ~     (p=0.355 n=20+20)
[Geo mean]                      484ms             477ms       -1.50%

[1] Daniel Lemire, Owen Kaser, Nathan Kurz. 2019. "Faster Remainder by
Direct Computation: Applications to Compilers and Software Libraries."
https://arxiv.org/abs/1902.01961

[2] https://github.com/mdempsky/benchdivmagic

Change-Id: Ie4d214e7a908b0d979c878f2d404bd56bdf374f6
Reviewed-on: https://go-review.googlesource.com/c/go/+/300994
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
Trust: Matthew Dempsky <mdempsky@google.com>
Trust: Michael Knyszek <mknyszek@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Michael Knyszek <mknyszek@google.com>
4 years agoruntime: add alignment info to sizeclasses.go comments
Matthew Dempsky [Mon, 8 Mar 2021 23:36:28 +0000 (15:36 -0800)]
runtime: add alignment info to sizeclasses.go comments

I was curious about the minimum possible alignment for each size class
and the minimum size to guarantee any particular alignment (e.g., to
know at what class size you can start assuming heap bits are byte- or
word-aligned).

Change-Id: I205b750286e8914986533c4f60712c420c3e63e9
Reviewed-on: https://go-review.googlesource.com/c/go/+/299909
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Michael Knyszek <mknyszek@google.com>
Trust: Matthew Dempsky <mdempsky@google.com>

4 years agocmd/go: include default GOEXPERIMENT in build config
Cherry Zhang [Fri, 12 Mar 2021 16:21:38 +0000 (11:21 -0500)]
cmd/go: include default GOEXPERIMENT in build config

Currently, the build config includes GOEXPERIMENT environment
variable if it is not empty, but that doesn't take the default
value (set at make.bash/bat/rc time) into consideration. This
may cause standard library packages appearing stale, as the
build config appears changed.

This CL changes it to use cfg.GOEXPERIMENT variable, which
includes the default value (if it is not overwritten).

May fix regabi and staticlockranking builders.

Change-Id: I242f887167f8e99192010be5c1a046eb88ab0c2a
Reviewed-on: https://go-review.googlesource.com/c/go/+/301269
Trust: Cherry Zhang <cherryyz@google.com>
Run-TryBot: Cherry Zhang <cherryyz@google.com>
Reviewed-by: Bryan C. Mills <bcmills@google.com>
TryBot-Result: Go Bot <gobot@golang.org>

4 years agoRevert "testing/fstest: test that ReadDirFile on a non-dir fails"
Josh Bleecher Snyder [Fri, 12 Mar 2021 15:45:18 +0000 (15:45 +0000)]
Revert "testing/fstest: test that ReadDirFile on a non-dir fails"

This reverts commit 1853411d8376570295711f9084d494d458822578.

Reason for revert: broke plan 9 builder. fixes #44967

Change-Id: Ib89448d37f7ab8bb05dbd89ce744431d807eb4da
Reviewed-on: https://go-review.googlesource.com/c/go/+/301190
Reviewed-by: Bryan C. Mills <bcmills@google.com>
Trust: Josh Bleecher Snyder <josharian@gmail.com>
Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com>

4 years agoA+C: add new e-mail addresses for Andy Pan
Andy Pan [Thu, 11 Mar 2021 06:22:16 +0000 (14:22 +0800)]
A+C: add new e-mail addresses for Andy Pan

Change-Id: I2be9cba124b407f27823b3a3778331b6118112cd
Reviewed-on: https://go-review.googlesource.com/c/go/+/300470
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Reviewed-by: Emmanuel Odeke <emmanuel@orijtech.com>
4 years agocmd/compile: fix noder.Addr() to not call typechecker
Dan Scales [Thu, 11 Mar 2021 01:27:30 +0000 (17:27 -0800)]
cmd/compile:  fix noder.Addr() to not call typechecker

Simple change to avoid calling the old typechecker in noder.Addr(). This
fixes cases where generic code calls a pointer method with a non-pointer
receiver.

Added test typeparam/lockable.go that now works with this change.

For lockable.go to work, also fix incorrect check to decide whether to
translate an OXDOT now or later. We should delay translating an OXDOT
until instantiation (because we don't know how embedding, etc. will
work) if the receiver has any typeparam, not just if the receiver type
is a simple typeparam. We also have to handle OXDOT for now in
IsAddressable(), until we can remove calls to the old typechecker in
(*irgen).funcBody().

Change-Id: I77ee5efcef9a8f6c7133564106a32437e36ba4bb
Reviewed-on: https://go-review.googlesource.com/c/go/+/300990
Run-TryBot: Dan Scales <danscales@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Trust: Dan Scales <danscales@google.com>
Trust: Robert Griesemer <gri@golang.org>
Reviewed-by: Robert Griesemer <gri@golang.org>
4 years agointernal/poll: fix some grammar errors
Andy Pan [Thu, 11 Mar 2021 14:34:45 +0000 (22:34 +0800)]
internal/poll: fix some grammar errors

Change-Id: I25a6424bce9d372fa46e8bdd856095845d3397bf
Reviewed-on: https://go-review.googlesource.com/c/go/+/300889
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Reviewed-by: Emmanuel Odeke <emmanuel@orijtech.com>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>

4 years agocmd/internal/obj/arm64: add support for op(extended register) with RSP arguments
fanzha02 [Wed, 20 Jan 2021 09:58:21 +0000 (17:58 +0800)]
cmd/internal/obj/arm64: add support for op(extended register) with RSP arguments

Refer to ARM reference manual, like add(extended register) instructions,
the extension is encoded in the "option" field. If "Rd" or "Rn" is
RSP and "option" is "010" then LSL is preferred. Therefore, the instrution
"add Rm<<imm, RSP, RSP" or "add Rm<<imm RSP" is valid and can be encoded
as add(extended register) instruction.

But the current assembler can not handle like "op R1<<1, RSP, RSP"
instructions, this patch adds the support.

Because MVN(extended register) does not exist, remove it.

Add test cases.

Change-Id: I968749d75c6b93a4f297b39c73cc292e6b1035ad
Reviewed-on: https://go-review.googlesource.com/c/go/+/284900
Trust: fannie zhang <Fannie.Zhang@arm.com>
Run-TryBot: fannie zhang <Fannie.Zhang@arm.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
4 years agocmd/compile: call types.CheckSize() in g.typ()
Dan Scales [Thu, 11 Mar 2021 03:28:28 +0000 (19:28 -0800)]
cmd/compile:  call types.CheckSize() in g.typ()

Restore code to call types.CheckSize() in g.typ(). There are certain
cases (involving maps) where we need to do CheckSize here. In general,
the old typechecker calls CheckSize() a lot, and we want to eliminate
calling it eventually, so should get do types.CheckSize() when we create
a new concrete type.

However, the test typeparams/cons.go does not work with just calling
types.CheckSize() in g.typ() (which is why I disabled the calls
originally). The reason is that g.typ() is called recursively within
types.go, so it can be called on a partially-created recursive type,
which leads to an error in CheckSize(). So, we need to call CheckSize()
only on fully-created top-level types. So, I divided typ() into typ()
and typ1(), where typ() is now the external entry point, and typ1() is
called within types.go. Now, typ() can call CheckSize() safely.

I also added in an extra condition - we do not currently need to call
CheckSize() on non-fully-instantiated types, since they will not make it
to the backend.  That could change a bit with dictionaries.

Fixes #44895

Change-Id: I783aa7d2999dd882ddbd99a7c19a6ff6ee420102
Reviewed-on: https://go-review.googlesource.com/c/go/+/300989
Run-TryBot: Dan Scales <danscales@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Trust: Dan Scales <danscales@google.com>
Trust: Robert Griesemer <gri@golang.org>
Reviewed-by: Robert Griesemer <gri@golang.org>
4 years agocmd: move GOEXPERIMENT knob from make.bash to cmd/go
Matthew Dempsky [Thu, 24 Dec 2020 03:49:39 +0000 (19:49 -0800)]
cmd: move GOEXPERIMENT knob from make.bash to cmd/go

This CL changes GOEXPERIMENT to act like other GO[CONFIG] environment
variables. Namely, that it can be set at make.bash time to provide a
default value used by the toolchain, but then can be manually set when
running either cmd/go or the individual tools (compiler, assembler,
linker).

For example, it's now possible to test rsc.io/tmp/fieldtrack by simply
running:

GOEXPERIMENT=fieldtrack go test -gcflags=-l rsc.io/tmp/fieldtrack \
  -ldflags=-k=rsc.io/tmp/fieldtrack.tracked

without needing to re-run make.bash. (-gcflags=-l is needed because
the compiler's inlining abilities have improved, so calling a function
with a for loop is no longer sufficient to suppress inlining.)

Fixes #42681.

Change-Id: I2cf8995d5d0d05f6785a2ee1d3b54b2cfb3331ca
Reviewed-on: https://go-review.googlesource.com/c/go/+/300991
Trust: Matthew Dempsky <mdempsky@google.com>
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Bryan C. Mills <bcmills@google.com>
4 years agonet/http: revert change to generated file from CL 296152
Bryan C. Mills [Thu, 11 Mar 2021 20:04:16 +0000 (15:04 -0500)]
net/http: revert change to generated file from CL 296152

This file is generated, so the fix needs to happen upstream.
The file can then be regenerated using 'go generate net/http'.

Updates #44143

Change-Id: I13a1e7677470ba84a06976e5bbe24f4ce1e7cfb2
Reviewed-on: https://go-review.googlesource.com/c/go/+/301069
Trust: Bryan C. Mills <bcmills@google.com>
Run-TryBot: Bryan C. Mills <bcmills@google.com>
Reviewed-by: David Chase <drchase@google.com>
TryBot-Result: Go Bot <gobot@golang.org>

4 years agocmd/go: remove some fsyncs when writing files
Michael Matloob [Fri, 11 Dec 2020 22:03:17 +0000 (17:03 -0500)]
cmd/go: remove some fsyncs when writing files

cache.Trim, dowloadZip, rewriteVersionList, writeDiskCache all use
renameio.WriteFile to write their respective files to disk. For the
uses in cache.Trim and downloadZip, instead do of renameio.WriteFile,
do a truncate to the length of the file, then write the relevant bytes
so that a corrupt file (which would contain null bytes because of
the truncate) could be detected. For rewriteVersionList, use
lockedfile.Transform to do the write (which does a truncate as part of
the write too. writeDiskCache stays the same in this CL.

Also desete renameio methods that aren't used and remove the
renameio.WriteFile wrapper and just use renameio.WriteToFile which it
wraps.

There is a possibility of corrupt files in the cache (which was true
even before this CL) so later CLs will add facilities to clear corrupt
files in the cache.

Change-Id: I0d0bda40095e4cb898314315bf313e71650d8d25
Reviewed-on: https://go-review.googlesource.com/c/go/+/277412
Trust: Michael Matloob <matloob@golang.org>
Run-TryBot: Michael Matloob <matloob@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Bryan C. Mills <bcmills@google.com>
Reviewed-by: Jay Conrod <jayconrod@google.com>
4 years agocmd/compile: optimize multi-register shifts on amd64
Josh Bleecher Snyder [Fri, 8 Jan 2021 03:25:05 +0000 (19:25 -0800)]
cmd/compile: optimize multi-register shifts on amd64

amd64 can shift in bits from another register instead of filling with 0/1.
This pattern is helpful when implementing 128 bit shifts or arbitrary length shifts.
In the standard library, it shows up in pure Go math/big.

Benchmarks results on amd64 with -tags=math_big_pure_go.

name                          old time/op  new time/op  delta
NonZeroShifts/1/shrVU-8       4.45ns ± 3%  4.39ns ± 1%   -1.28%  (p=0.000 n=30+27)
NonZeroShifts/1/shlVU-8       4.13ns ± 4%  4.10ns ± 2%     ~     (p=0.254 n=29+28)
NonZeroShifts/2/shrVU-8       5.55ns ± 1%  5.63ns ± 2%   +1.42%  (p=0.000 n=28+29)
NonZeroShifts/2/shlVU-8       5.70ns ± 2%  5.14ns ± 1%   -9.82%  (p=0.000 n=29+28)
NonZeroShifts/3/shrVU-8       6.79ns ± 2%  6.35ns ± 2%   -6.46%  (p=0.000 n=28+29)
NonZeroShifts/3/shlVU-8       6.69ns ± 1%  6.25ns ± 1%   -6.60%  (p=0.000 n=28+27)
NonZeroShifts/4/shrVU-8       7.79ns ± 2%  7.06ns ± 2%   -9.48%  (p=0.000 n=30+30)
NonZeroShifts/4/shlVU-8       7.82ns ± 1%  7.24ns ± 1%   -7.37%  (p=0.000 n=28+29)
NonZeroShifts/5/shrVU-8       8.90ns ± 3%  7.93ns ± 1%  -10.84%  (p=0.000 n=29+26)
NonZeroShifts/5/shlVU-8       8.68ns ± 1%  7.92ns ± 1%   -8.76%  (p=0.000 n=29+29)
NonZeroShifts/10/shrVU-8      14.4ns ± 1%  12.3ns ± 2%  -14.79%  (p=0.000 n=28+29)
NonZeroShifts/10/shlVU-8      14.1ns ± 1%  11.9ns ± 2%  -15.55%  (p=0.000 n=28+27)
NonZeroShifts/100/shrVU-8      118ns ± 1%    96ns ± 3%  -18.82%  (p=0.000 n=30+29)
NonZeroShifts/100/shlVU-8      120ns ± 2%    98ns ± 2%  -18.46%  (p=0.000 n=29+28)
NonZeroShifts/1000/shrVU-8    1.10µs ± 1%  0.88µs ± 2%  -19.63%  (p=0.000 n=29+30)
NonZeroShifts/1000/shlVU-8    1.10µs ± 2%  0.88µs ± 2%  -20.28%  (p=0.000 n=29+28)
NonZeroShifts/10000/shrVU-8   10.9µs ± 1%   8.7µs ± 1%  -19.78%  (p=0.000 n=28+27)
NonZeroShifts/10000/shlVU-8   10.9µs ± 2%   8.7µs ± 1%  -19.64%  (p=0.000 n=29+27)
NonZeroShifts/100000/shrVU-8   111µs ± 2%    90µs ± 2%  -19.39%  (p=0.000 n=28+29)
NonZeroShifts/100000/shlVU-8   113µs ± 2%    90µs ± 2%  -20.43%  (p=0.000 n=30+27)

The assembly version is still faster, unfortunately, but the gap is narrowing.
Speedup from pure Go to assembly:

name                          old time/op  new time/op  delta
NonZeroShifts/1/shrVU-8       4.39ns ± 1%  3.45ns ± 2%  -21.36%  (p=0.000 n=27+29)
NonZeroShifts/1/shlVU-8       4.10ns ± 2%  3.47ns ± 3%  -15.42%  (p=0.000 n=28+30)
NonZeroShifts/2/shrVU-8       5.63ns ± 2%  3.97ns ± 0%  -29.40%  (p=0.000 n=29+25)
NonZeroShifts/2/shlVU-8       5.14ns ± 1%  3.77ns ± 2%  -26.65%  (p=0.000 n=28+26)
NonZeroShifts/3/shrVU-8       6.35ns ± 2%  4.79ns ± 2%  -24.52%  (p=0.000 n=29+29)
NonZeroShifts/3/shlVU-8       6.25ns ± 1%  4.42ns ± 1%  -29.29%  (p=0.000 n=27+26)
NonZeroShifts/4/shrVU-8       7.06ns ± 2%  5.64ns ± 1%  -20.05%  (p=0.000 n=30+29)
NonZeroShifts/4/shlVU-8       7.24ns ± 1%  5.34ns ± 2%  -26.23%  (p=0.000 n=29+29)
NonZeroShifts/5/shrVU-8       7.93ns ± 1%  6.56ns ± 2%  -17.26%  (p=0.000 n=26+30)
NonZeroShifts/5/shlVU-8       7.92ns ± 1%  6.27ns ± 1%  -20.79%  (p=0.000 n=29+25)
NonZeroShifts/10/shrVU-8      12.3ns ± 2%  10.2ns ± 2%  -17.21%  (p=0.000 n=29+29)
NonZeroShifts/10/shlVU-8      11.9ns ± 2%  10.5ns ± 2%  -12.45%  (p=0.000 n=27+29)
NonZeroShifts/100/shrVU-8     95.9ns ± 3%  77.7ns ± 1%  -19.00%  (p=0.000 n=29+30)
NonZeroShifts/100/shlVU-8     97.5ns ± 2%  66.8ns ± 2%  -31.47%  (p=0.000 n=28+30)
NonZeroShifts/1000/shrVU-8     884ns ± 2%   705ns ± 1%  -20.17%  (p=0.000 n=30+28)
NonZeroShifts/1000/shlVU-8     880ns ± 2%   590ns ± 1%  -32.96%  (p=0.000 n=28+25)
NonZeroShifts/10000/shrVU-8   8.74µs ± 1%  7.34µs ± 3%  -15.94%  (p=0.000 n=27+30)
NonZeroShifts/10000/shlVU-8   8.73µs ± 1%  6.00µs ± 1%  -31.25%  (p=0.000 n=27+28)
NonZeroShifts/100000/shrVU-8  89.6µs ± 2%  75.5µs ± 2%  -15.80%  (p=0.000 n=29+29)
NonZeroShifts/100000/shlVU-8  89.6µs ± 2%  68.0µs ± 3%  -24.09%  (p=0.000 n=27+30)

Change-Id: I18f58d8f5513d737d9cdf09b8f9d14011ffe3958
Reviewed-on: https://go-review.googlesource.com/c/go/+/297050
Trust: Josh Bleecher Snyder <josharian@gmail.com>
Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
4 years agohash/maphash: increase the buffer size
Josh Bleecher Snyder [Thu, 17 Dec 2020 04:28:43 +0000 (20:28 -0800)]
hash/maphash: increase the buffer size

This helps a lot for larger writes.

name            old time/op    new time/op    delta
Hash8Bytes-8      16.5ns ± 4%    16.8ns ± 2%   +1.76%  (p=0.000 n=29+30)
Hash320Bytes-8    55.0ns ± 2%    41.4ns ± 2%  -24.64%  (p=0.000 n=28+28)
Hash1K-8           190ns ± 2%     130ns ± 3%  -31.65%  (p=0.000 n=30+30)
Hash8K-8          1.57µs ± 2%    1.05µs ± 3%  -33.01%  (p=0.000 n=30+29)

name            old speed      new speed      delta
Hash8Bytes-8     485MB/s ± 4%   476MB/s ± 2%   -1.73%  (p=0.000 n=29+30)
Hash320Bytes-8  5.82GB/s ± 2%  7.72GB/s ± 3%  +32.55%  (p=0.000 n=28+29)
Hash1K-8        5.39GB/s ± 2%  7.88GB/s ± 3%  +46.32%  (p=0.000 n=30+30)
Hash8K-8        5.21GB/s ± 2%  7.77GB/s ± 3%  +49.28%  (p=0.000 n=30+29)

Updates #42710

Change-Id: Idaf4b2a8a41fc62fc16b54c9358cf2cc7009cf29
Reviewed-on: https://go-review.googlesource.com/c/go/+/278760
Trust: Josh Bleecher Snyder <josharian@gmail.com>
Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
4 years agohash/maphash: optimize Write and WriteString
Josh Bleecher Snyder [Thu, 17 Dec 2020 03:38:39 +0000 (19:38 -0800)]
hash/maphash: optimize Write and WriteString

The existing code makes copies of every byte it hashes.
When passed a large chunk of memory, Write and WriteString
can skip the copying and initSeed for most of it.

To ensure that Write, WriteByte, and WriteString continue to
generate output that depends only on the sequence of bytes,
expand the grouping test to include WriteString and interleaved calls.
Also, make the test process a lot more data, to ensure that
Write* handled full buffers correctly.

name            old time/op    new time/op    delta
Hash8Bytes-8      17.1ns ± 3%    16.5ns ± 2%   -3.26%  (p=0.000 n=29+27)
Hash320Bytes-8    74.9ns ± 2%    58.5ns ± 2%  -21.86%  (p=0.000 n=30+29)
Hash1K-8           246ns ± 3%     195ns ± 1%  -20.82%  (p=0.000 n=29+30)
Hash8K-8          1.87µs ± 2%    1.59µs ± 2%  -15.04%  (p=0.000 n=26+30)

name            old speed      new speed      delta
Hash8Bytes-8     468MB/s ± 3%   484MB/s ± 2%   +3.36%  (p=0.000 n=29+27)
Hash320Bytes-8  4.28GB/s ± 2%  5.47GB/s ± 2%  +27.97%  (p=0.000 n=30+29)
Hash1K-8        4.17GB/s ± 3%  5.26GB/s ± 1%  +26.28%  (p=0.000 n=29+30)
Hash8K-8        4.38GB/s ± 2%  5.16GB/s ± 2%  +17.70%  (p=0.000 n=26+30)

Updates #42710

Change-Id: If3cdec1580ffb3e36fab9865e5a9d089c0a34bec
Reviewed-on: https://go-review.googlesource.com/c/go/+/278758
Trust: Josh Bleecher Snyder <josharian@gmail.com>
Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
4 years agohash/maphash: manually inline setSeed
Josh Bleecher Snyder [Thu, 17 Dec 2020 04:22:58 +0000 (20:22 -0800)]
hash/maphash: manually inline setSeed

Provides a minor performance win.

name            old time/op    new time/op    delta
Hash8Bytes-8      16.5ns ± 2%    16.5ns ± 4%    ~     (p=0.407 n=27+29)
Hash320Bytes-8    58.5ns ± 2%    55.0ns ± 2%  -6.01%  (p=0.000 n=29+28)
Hash1K-8           195ns ± 1%     190ns ± 2%  -2.23%  (p=0.000 n=30+30)
Hash8K-8          1.59µs ± 2%    1.57µs ± 2%  -0.88%  (p=0.002 n=30+30)

name            old speed      new speed      delta
Hash8Bytes-8     484MB/s ± 2%   485MB/s ± 4%    ~     (p=0.417 n=27+29)
Hash320Bytes-8  5.47GB/s ± 2%  5.82GB/s ± 2%  +6.39%  (p=0.000 n=29+28)
Hash1K-8        5.26GB/s ± 1%  5.39GB/s ± 2%  +2.29%  (p=0.000 n=30+30)
Hash8K-8        5.16GB/s ± 2%  5.21GB/s ± 2%  +0.89%  (p=0.002 n=30+30)

Updates #42710

Change-Id: Ia0d7264b648f96099202de21c6b69a9c1776f6c8
Reviewed-on: https://go-review.googlesource.com/c/go/+/278759
Trust: Josh Bleecher Snyder <josharian@gmail.com>
Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
4 years agotesting/fstest: clarify TestFS docs
Josh Bleecher Snyder [Thu, 25 Feb 2021 19:10:57 +0000 (11:10 -0800)]
testing/fstest: clarify TestFS docs

The sentence starts "fsys must only contain",
which leads the reader to believe that fsys must not contain others.
The rapid reversal leads to confusion.
I had to read it several times to be sure I'd parsed it correctly.

Remove "only"; rely on the rest of the sentence to clarify.

Change-Id: I9fb7935aed4f9839344d3a00b761d20981fba864
Reviewed-on: https://go-review.googlesource.com/c/go/+/296529
Trust: Josh Bleecher Snyder <josharian@gmail.com>
Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
4 years agotesting/fstest: test that ReadDirFile on a non-dir fails
Josh Bleecher Snyder [Thu, 25 Feb 2021 17:44:35 +0000 (09:44 -0800)]
testing/fstest: test that ReadDirFile on a non-dir fails

ReadDirFile implementations should return an error for non-directories.

Change-Id: I99888562cb6cf829017904ae8c1e8887a416c4cd
Reviewed-on: https://go-review.googlesource.com/c/go/+/296391
Trust: Josh Bleecher Snyder <josharian@gmail.com>
Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
4 years agoio/fs: clarify additional File interface docs
Josh Bleecher Snyder [Thu, 25 Feb 2021 17:32:38 +0000 (09:32 -0800)]
io/fs: clarify additional File interface docs

Emphasize ReadDirFile. It isn't really optional,
and all filesystems have at least one directory (".").

The remaining two additional interfaces are optimizations.
Call them that.

Fully qualify package package io identifiers.

Change-Id: Ibc425a5dfd27e08c2c10c353f780e4a6304cfd87
Reviewed-on: https://go-review.googlesource.com/c/go/+/296390
Trust: Josh Bleecher Snyder <josharian@gmail.com>
Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
4 years agocmd/link: fix glink resolver generation on ppc64le
Paul E. Murphy [Wed, 10 Mar 2021 20:32:27 +0000 (14:32 -0600)]
cmd/link: fix glink resolver generation on ppc64le

AddSymRef grows the section to fit the relocation. This was not
being accounted for. Instead, add a relocation and explicitly
populate it so we patch the desired instructions.

Change-Id: I583147e2545aea34c854f9d35ca920c57be60b90
Reviewed-on: https://go-review.googlesource.com/c/go/+/300949
Run-TryBot: Paul Murphy <murp@ibm.com>
Run-TryBot: Lynn Boger <laboger@linux.vnet.ibm.com>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Trust: Lynn Boger <laboger@linux.vnet.ibm.com>

4 years agoruntime: support register ABI for finalizers
Michael Anthony Knyszek [Thu, 22 Oct 2020 16:55:55 +0000 (16:55 +0000)]
runtime: support register ABI for finalizers

This change modifies runfinq to properly pass arguments to finalizers in
registers via reflectcall.

For #40724.

Change-Id: I414c0eff466ef315a0eb10507994e598dd29ccb2
Reviewed-on: https://go-review.googlesource.com/c/go/+/300112
Trust: Michael Knyszek <mknyszek@google.com>
Run-TryBot: Michael Knyszek <mknyszek@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
4 years agotest: add test that caused a gofrontend internal error
Ian Lance Taylor [Thu, 11 Mar 2021 03:35:22 +0000 (19:35 -0800)]
test: add test that caused a gofrontend internal error

For #44383

Change-Id: I3610105dad3574e210e226d3ba80a4ba5a7eeaa6
Reviewed-on: https://go-review.googlesource.com/c/go/+/300789
Trust: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Reviewed-by: Than McIntosh <thanm@google.com>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
4 years agodocs: clarify when APIs use context.Background.
Matt T. Proud [Thu, 25 Feb 2021 00:03:35 +0000 (01:03 +0100)]
docs: clarify when APIs use context.Background.

The Go standard library retrofitted context support onto existing APIs
using context.Background and later offered variants that directly
supported user-defined context value specification. This commit makes
that behavior clear in documentation and suggests context-aware
alternatives if the user is looking for one.

An example motivation is supporting code for use in systems that expect
APIs to be cancelable for lifecycle correctness or load
shedding/management reasons, as alluded to in
https://blog.golang.org/context-and-structs.

Updates #44143

Change-Id: I2d7f954ddf9b48264d5ebc8d0007058ff9bddf14
Reviewed-on: https://go-review.googlesource.com/c/go/+/296152
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Reviewed-by: Jean de Klerk <deklerk@google.com>
Trust: Jean de Klerk <deklerk@google.com>
Run-TryBot: Jean de Klerk <deklerk@google.com>
TryBot-Result: Go Bot <gobot@golang.org>

4 years agosyscall: use runtime.KeepAlive for ProcThreadAttributeList arguments
Jason A. Donenfeld [Wed, 10 Mar 2021 14:21:56 +0000 (07:21 -0700)]
syscall: use runtime.KeepAlive for ProcThreadAttributeList arguments

It turns out that if you write Go pointers to Go memory, the Go compiler
must be involved so that it generates various calls to the GC in the
process. Letting Windows write Go pointers to Go memory violated this.
So, we replace that with just a boring call to runtime.KeepAlive. That's
not a great API, but this is all internal code anyway. We fix it up
more elegantly for external consumption in x/sys/windows with CL 300369.

Fixes #44900.

Change-Id: Id6599a793af9c4815f6c9387b00796923f32cb97
Reviewed-on: https://go-review.googlesource.com/c/go/+/300349
Trust: Jason A. Donenfeld <Jason@zx2c4.com>
Run-TryBot: Jason A. Donenfeld <Jason@zx2c4.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
4 years agocrypto/rand, internal/syscall/unix: add support for getrandom syscall on solaris
Tobias Klauser [Wed, 10 Mar 2021 09:26:20 +0000 (10:26 +0100)]
crypto/rand, internal/syscall/unix: add support for getrandom syscall on solaris

The getrandom syscall is available on Solaris and Illumos, see
https://docs.oracle.com/cd/E88353_01/html/E37841/getrandom-2.html and
https://illumos.org/man/2/getrandom

Change-Id: Id1c65d6a5b2fbc80d20b43d8b32dab137ca950ca
Reviewed-on: https://go-review.googlesource.com/c/go/+/299134
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 agointernal/syscall/unix: unify GetRandom implementation
Tobias Klauser [Wed, 10 Mar 2021 09:06:11 +0000 (10:06 +0100)]
internal/syscall/unix: unify GetRandom implementation

The implementation of GetRandom for Linux, FreeBSD and DragonflyBSD can
be shared.

Also remove GRND_INSECURE on DragonflyBSD as pointed out by Ian in the
review of CL 269999.

Change-Id: I5bf4c1bd51ddb2ad600652a57e0bc1bafa1cf40d
Reviewed-on: https://go-review.googlesource.com/c/go/+/299133
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/dist: refactor test constraints for misc/cgo/testsantizers
fanzha02 [Mon, 1 Mar 2021 02:34:08 +0000 (10:34 +0800)]
cmd/dist: refactor test constraints for misc/cgo/testsantizers

Currently, the cmd/dist runs test cases in misc/cgo/testsantizers only
when memeory sanitizer is supported, but the tsan tests in
misc/cgo/testsanitizers do not require support for -msan option, which
makes tsan tests can not be run on some unsupported -msan option platforms.

Therefore, this patch moves the test constraints from cmd/dist to
msan_test.go, so that the tsan tests in misc/cgo/testsanitizers
can be run on any system where the C compiler supports -fsanitize=thread
option.

Change-Id: I779c92eedd0270050f1a0b1a69ecce50c3712bc9
Reviewed-on: https://go-review.googlesource.com/c/go/+/297774
Trust: fannie zhang <Fannie.Zhang@arm.com>
Run-TryBot: fannie zhang <Fannie.Zhang@arm.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
4 years agocmd/compile: remove 8-byte alignment requirement of stack slot on ppc64
eric fang [Tue, 2 Mar 2021 06:05:00 +0000 (06:05 +0000)]
cmd/compile: remove 8-byte alignment requirement of stack slot on ppc64

This CL applies CL 267999 to ppc64.

Fixes #42385

Change-Id: I6462d647d1abdf7cec99607c40ef4d1fed1941e5
Reviewed-on: https://go-review.googlesource.com/c/go/+/297770
Reviewed-by: eric fang <eric.fang@arm.com>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Trust: eric fang <eric.fang@arm.com>

4 years agocmd/compile: fix handling of partially inferred type arguments
Dan Scales [Wed, 10 Mar 2021 02:24:51 +0000 (18:24 -0800)]
cmd/compile: fix handling of partially inferred type arguments

In the case of partially inferred type arguments, we need to use the
IndexExpr as the key in g.info.Inferred[] rather than the CallExpr.

Added an extra fromStrings1 call in the settable.go test that tests
partially inferred type arguments. This new call uses a new concrete
type SettableString as well.

I also added another implementation fromStrings3 (derived from a go2go
tests) that typechecks but intentionally causes a panic.

Change-Id: I74d35c5a741f72f37160a96fbec939451157f392
Reviewed-on: https://go-review.googlesource.com/c/go/+/300309
Run-TryBot: Dan Scales <danscales@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Trust: Dan Scales <danscales@google.com>
Trust: Robert Griesemer <gri@golang.org>
Reviewed-by: Robert Griesemer <gri@golang.org>
4 years agocmd/internal/obj: remove param element from ppc64 optab
Paul E. Murphy [Tue, 9 Mar 2021 22:54:56 +0000 (16:54 -0600)]
cmd/internal/obj: remove param element from ppc64 optab

This is rarely used, and is implied based on the
memory type of the operand.  This is a step towards
simplifying the MOV* pseudo opcodes on ppc64.

Similarly, remove the bogus param value from AVMULESB.

Change-Id: Ibad4d045ec6d8c5163a468b2db1dfb762ef674ee
Reviewed-on: https://go-review.googlesource.com/c/go/+/300177
Run-TryBot: Paul Murphy <murp@ibm.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Lynn Boger <laboger@linux.vnet.ibm.com>
Trust: Ian Lance Taylor <iant@golang.org>

4 years agoruntime, time: disable preemption in addtimer
Michael Pratt [Wed, 10 Mar 2021 21:06:47 +0000 (16:06 -0500)]
runtime, time: disable preemption in addtimer

The timerpMask optimization updates a mask of Ps (potentially)
containing timers in pidleget / pidleput. For correctness, it depends on
the assumption that new timers can only be added to a P's own heap.

addtimer violates this assumption if it is preempted after computing pp.
That G may then run on a different P, but adding a timer to the original
P's heap.

Avoid this by disabling preemption while pp is in use.

Other uses of doaddtimer should be OK:

* moveTimers: always moves to the current P's heap
* modtimer, cleantimers, addAdjustedTimers, runtimer: does not add net
  new timers to the heap while locked

Fixes #44868

Change-Id: I4a5d080865e854931d0a3a09a51ca36879101d72
Reviewed-on: https://go-review.googlesource.com/c/go/+/300610
Trust: Michael Pratt <mpratt@google.com>
Run-TryBot: Michael Pratt <mpratt@google.com>
Reviewed-by: Michael Knyszek <mknyszek@google.com>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>

4 years agocmd/go/internal/mvs: factor out an incremental implementation
Bryan C. Mills [Thu, 29 Oct 2020 01:47:32 +0000 (21:47 -0400)]
cmd/go/internal/mvs: factor out an incremental implementation

The new Graph type implements an incremental version of the MVS
algorithm, with requirements pushed in by the caller instead of pulled
by an internal MVS traversal.

To avoid redundancy going forward (and to ensure adequate test
coverage of the incremental implementation), the existing buildList
function is reimplemented in terms of Graph.

For #36460

Change-Id: Idd0b6ab8f17cc41d83a2a4c25a95f82e9ce1eab0
Reviewed-on: https://go-review.googlesource.com/c/go/+/244760
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>
4 years agocmd/go/internal/modload: make EditBuildList report whether the build list was changed
Bryan C. Mills [Mon, 22 Feb 2021 22:05:32 +0000 (17:05 -0500)]
cmd/go/internal/modload: make EditBuildList report whether the build list was changed

For #36460

Change-Id: I8dd6e6f998a217a4287212815ce61209df6f007f
Reviewed-on: https://go-review.googlesource.com/c/go/+/296609
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>
4 years agocmd/go/internal/modload: fuse upgrading with downgrading in EditBuildList
Bryan C. Mills [Sat, 6 Feb 2021 05:50:55 +0000 (00:50 -0500)]
cmd/go/internal/modload: fuse upgrading with downgrading in EditBuildList

Previosly, EditBuildList performed an mvs.Upgrade followed by an
mvs.Downgrade, with the Downgrade building on the result of the
Upgrade. Unfortunately, that approach potentially folds in irrelevant
dependencies from the first Upgrade, which are then preserved
unnecessarily by the Downgrade (see mod_get_downup_artifact.txt).

Now, we use the initial Upgrade only to compute the maximum allowed
versions of transitive dependencies, and apply the module upgrades and
downgrades together in a single operation.

For #36460

Change-Id: I7590c137111fed4a3b06531c88d90efd49e6943a
Reviewed-on: https://go-review.googlesource.com/c/go/+/290770
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>
4 years agocmd/go: test remote lookup of packages with leading dots in path elements
Jay Conrod [Mon, 1 Mar 2021 20:18:12 +0000 (15:18 -0500)]
cmd/go: test remote lookup of packages with leading dots in path elements

Follow-up to CL 297530.

For #43985
For #34992

Change-Id: I2cfa6c41c013e627c3464c383ca42f5c9ebe521a
Reviewed-on: https://go-review.googlesource.com/c/go/+/297634
Trust: Jay Conrod <jayconrod@google.com>
Run-TryBot: Jay Conrod <jayconrod@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Bryan C. Mills <bcmills@google.com>
4 years agocmd/compile/internal: improve handling of DS form offsets on ppc64x
Lynn Boger [Mon, 8 Mar 2021 16:07:17 +0000 (10:07 -0600)]
cmd/compile/internal: improve handling of DS form offsets on ppc64x

In the ppc64 ISA DS form loads and stores are restricted to offset
fields that are a multiple of 4. This is currently handled with checks
in the rules that generate MOVDload, MOVWload, MOVDstore and
MOVDstorezero to prevent invalid instructions from getting to the
assembler.

An unhandled case was discovered which led to the search for a better
solution to this problem. Now, instead of checking the offset in the
rules, this will be detected when processing these Ops in
ssaGenValues in ppc64/ssa.go. If the offset is not valid, the address
of the symbol to be loaded or stored will be computed using the base
register + offset, and that value used in the new base register.
With the full address in the base register, the offset field can be
zero in the instruction.

Updates #44739

Change-Id: I4f3c0c469ae70a63e3add295c9b55ea0e30ef9b3
Reviewed-on: https://go-review.googlesource.com/c/go/+/299789
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: Cherry Zhang <cherryyz@google.com>
4 years agoruntime: check partial lock ranking order
Michael Pratt [Tue, 9 Mar 2021 21:13:23 +0000 (16:13 -0500)]
runtime: check partial lock ranking order

To ease readability we typically keep the partial order lists sorted by
rank. This isn't required for correctness, it just makes it (slightly)
easier to read the lists.

Currently we must notice out-of-order entries during code review, which
is an error-prone process.

Add a test to enforce ordering, and fix the errors that have crept in.
Most of the existing errors were misordered lockRankHchan or
lockRankPollDesc.

While we're here, I've moved the correctness check that the partial
ordering satisfies the total ordering from init to a test case. This
will allow us to catch these errors without even running
staticlockranking.

Change-Id: I9c11abe49ea26c556439822bb6a3183129600c3b
Reviewed-on: https://go-review.googlesource.com/c/go/+/300171
Trust: Michael Pratt <mpratt@google.com>
Trust: Dan Scales <danscales@google.com>
Run-TryBot: Michael Pratt <mpratt@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Dan Scales <danscales@google.com>
4 years agoflag: panic if flag name begins with - or contains =
KimMachineGun [Wed, 10 Mar 2021 12:55:56 +0000 (12:55 +0000)]
flag: panic if flag name begins with - or contains =

Fixes #41792

Change-Id: I9b4aae8a899e3c3ac9532d27932d275cfb1fab48
GitHub-Last-Rev: f06b1e17674bf77bdc2d3e798df4c4379748c8d2
GitHub-Pull-Request: golang/go#42737
Reviewed-on: https://go-review.googlesource.com/c/go/+/271788
Reviewed-by: Emmanuel Odeke <emmanuel@orijtech.com>
Reviewed-by: Rob Pike <r@golang.org>
Run-TryBot: Emmanuel Odeke <emmanuel@orijtech.com>
TryBot-Result: Go Bot <gobot@golang.org>
Trust: Rob Pike <r@golang.org>

4 years agoencoding/xml: prevent infinite loop while decoding
Katie Hockman [Mon, 1 Mar 2021 14:54:00 +0000 (09:54 -0500)]
encoding/xml: prevent infinite loop while decoding

This change properly handles a TokenReader which
returns an EOF in the middle of an open XML
element.

Thanks to Sam Whited for reporting this.

Fixes CVE-2021-27918
Fixes #44913

Change-Id: Id02a3f3def4a1b415fa2d9a8e3b373eb6cb0f433
Reviewed-on: https://team-review.git.corp.google.com/c/golang/go-private/+/1004594
Reviewed-by: Russ Cox <rsc@google.com>
Reviewed-by: Roland Shoemaker <bracewell@google.com>
Reviewed-by: Filippo Valsorda <valsorda@google.com>
Reviewed-on: https://go-review.googlesource.com/c/go/+/300391
Trust: Katie Hockman <katie@golang.org>
Run-TryBot: Katie Hockman <katie@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Alexander Rakoczy <alex@golang.org>
Reviewed-by: Filippo Valsorda <filippo@golang.org>
4 years agoarchive/zip: fix panic in Reader.Open
Roland Shoemaker [Tue, 2 Mar 2021 18:00:53 +0000 (10:00 -0800)]
archive/zip: fix panic in Reader.Open

When operating on a Zip file that contains a file prefixed with "../",
Open(...) would cause a panic in toValidName when attempting to strip
the prefixed path components.

Fixes CVE-2021-27919
Fixes #44916

Change-Id: Ic755d8126cb0897e2cbbdacf572439c38dde7b35
Reviewed-on: https://team-review.git.corp.google.com/c/golang/go-private/+/1004761
Reviewed-by: Filippo Valsorda <valsorda@google.com>
Reviewed-by: Russ Cox <rsc@google.com>
Reviewed-by: Katie Hockman <katiehockman@google.com>
Reviewed-on: https://go-review.googlesource.com/c/go/+/300489
Trust: Katie Hockman <katie@golang.org>
Run-TryBot: Katie Hockman <katie@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Alexander Rakoczy <alex@golang.org>
Reviewed-by: Filippo Valsorda <filippo@golang.org>
4 years agocmd/compile: deal with helper generic types that add methods to T
Dan Scales [Sat, 6 Mar 2021 06:07:56 +0000 (22:07 -0800)]
cmd/compile:  deal with helper generic types that add methods to T

Deal with cases like: 'type P[T any] T' (used to add methods to an
arbitrary type T), In this case, P[T] has kind types.TTYPEPARAM (as does
T itself), but requires more code to substitute than a simple TTYPEPARAM
T. See the comment near the beginning of subster.typ() in stencil.go.

Add new test absdiff.go. This test has a case for complex types (which
I've commented out) that will only work when we deal better with Go
builtins in generic functions (like real and imag).

Remove change in fmt.go for TTYPEPARAMS that is no longer needed (since
all TTYPEPARAMS have a sym) and was sometimes causing an extra prefix
when formatting method names.

Separate out the setting of a TTYPEPARAM bound, since it can reference
the TTYPEPARAM being defined, so must be done separately. Also, we don't
currently (and may not ever) need bounds after types2 typechecking.

Change-Id: Id173057e0c4563b309b95e665e9c1151ead4ba77
Reviewed-on: https://go-review.googlesource.com/c/go/+/300049
Run-TryBot: Dan Scales <danscales@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Trust: Dan Scales <danscales@google.com>
Trust: Robert Griesemer <gri@golang.org>
Reviewed-by: Robert Griesemer <gri@golang.org>
4 years agocmd/gofmt: fix const association to avoid inaccurate comment
Rob Findley [Wed, 10 Mar 2021 16:33:23 +0000 (11:33 -0500)]
cmd/gofmt: fix const association to avoid inaccurate comment

The const parseTypeParams was grouped with printer-related consts in
gofmt.go, implicitly suggesting that it must be kept in sync with
go/format/format.go.

Change-Id: Ia65dc15c27fef2c389f963071252adee32ec6bd6
Reviewed-on: https://go-review.googlesource.com/c/go/+/300451
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 agogo/types: add missing build tag to api_go1.18_test.go
Rob Findley [Wed, 10 Mar 2021 16:30:24 +0000 (11:30 -0500)]
go/types: add missing build tag to api_go1.18_test.go

This file has a go:build comment without a corresponding +build comment.

Change-Id: Id01604242a14d8ead16ffb9aa1b45eef7706956a
Reviewed-on: https://go-review.googlesource.com/c/go/+/300450
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 agoruntime/race: update dead link
Michael Pratt [Wed, 10 Mar 2021 16:41:04 +0000 (11:41 -0500)]
runtime/race: update dead link

LLVM changed their main branch name, so this link didn't work anymore.

Change-Id: I4c3a67b26e2bda012071281e29ea3c932c185130
Reviewed-on: https://go-review.googlesource.com/c/go/+/300469
Trust: Michael Pratt <mpratt@google.com>
Run-TryBot: Michael Pratt <mpratt@google.com>
Reviewed-by: Dmitry Vyukov <dvyukov@google.com>
TryBot-Result: Go Bot <gobot@golang.org>

4 years agocmd/compile: remove ".fp" fake arg
Cherry Zhang [Mon, 8 Mar 2021 20:19:47 +0000 (15:19 -0500)]
cmd/compile: remove ".fp" fake arg

No longer needed with previous CLs.

Change-Id: I9a1c11092a2736c190fa8e8ddfbb913b708957eb
Reviewed-on: https://go-review.googlesource.com/c/go/+/300155
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 agoruntime: update signature of reflectcall functions
Cherry Zhang [Wed, 10 Mar 2021 03:58:55 +0000 (22:58 -0500)]
runtime: update signature of reflectcall functions

reflectcall tail calls runtime.call{16,32,...} functions, so they
have the same signature as reflectcall. It is important for them
to have the correct arg map, because those functions, as well as
the function being reflectcall'd, could move the stack. When that
happens, its pointer arguments, in particular regArgs, need to be
adjusted. Otherwise it will still point to the old stack, causing
memory corruption.

This only caused failures on the regabi builder because it is the
only place where internal/abi.RegArgs is not a zero-sized type.

May fix #44821.

Change-Id: Iab400ea6b60c52360d0b43a793f6bfe50ca9989b
Reviewed-on: https://go-review.googlesource.com/c/go/+/300154
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 agocmd/go: avoid password prompts in TestScript/mod_get_private_vcs
Bryan C. Mills [Wed, 10 Mar 2021 04:31:41 +0000 (23:31 -0500)]
cmd/go: avoid password prompts in TestScript/mod_get_private_vcs

In some cases, this test would prompt for interactive SSH passwords in
order to authenticate to github.com over SSH. Setting GIT_SSH_COMMAND
to /bin/false prevents that, while still provoking the desired Git
failure mode.

Updates #44904.

Change-Id: Idc9fe9f47d2ccb6c8a4ea988b73d9c8c774e4079
Reviewed-on: https://go-review.googlesource.com/c/go/+/300156
Trust: Bryan C. Mills <bcmills@google.com>
Trust: Kevin Burke <kev@inburke.com>
Run-TryBot: Bryan C. Mills <bcmills@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Kevin Burke <kev@inburke.com>
4 years agocrypto/rand: supports for getrandom syscall in DragonFlyBSD
David Carlier [Tue, 9 Mar 2021 19:13:37 +0000 (19:13 +0000)]
crypto/rand: supports for getrandom syscall in DragonFlyBSD

Since the 5.7 release, DragonFlyBSD supports as well
the getrandom function, the actual stable is 5.8.

Change-Id: I2b8fc468771b10ac12b38ea7e8e5314342de6375
GitHub-Last-Rev: c5c496f41898d58f2c6f3ccc81f754792f49edbe
GitHub-Pull-Request: golang/go#42617
Reviewed-on: https://go-review.googlesource.com/c/go/+/269999
Run-TryBot: Ian Lance Taylor <iant@golang.org>
Trust: Ian Lance Taylor <iant@golang.org>
Trust: Tobias Klauser <tobias.klauser@gmail.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Tobias Klauser <tobias.klauser@gmail.com>
4 years agotesting: fix typo in a comment
fanzha02 [Wed, 10 Mar 2021 02:09:39 +0000 (10:09 +0800)]
testing: fix typo in a comment

Change-Id: I781808327be84113cd55c52bc214b821cd166114
Reviewed-on: https://go-review.googlesource.com/c/go/+/300269
Trust: fannie zhang <Fannie.Zhang@arm.com>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
4 years agointernal/poll: implement a pipe pool for splice() call
Andy Pan [Thu, 19 Nov 2020 11:09:14 +0000 (19:09 +0800)]
internal/poll: implement a pipe pool for splice() call

In scenarios where splice() is called, splice() is usually called not just once, but many times,
which means that a lot of pipes will be created and destroyed frequently, costing an amount of system resources
and slowing down performance, thus I suggest that we add a pipe pool for reusing pipes.

Benchmark tests:

goos: linux
goarch: amd64
pkg: internal/poll
cpu: AMD EPYC 7K62 48-Core Processor

name                  old time/op    new time/op    delta
SplicePipe-8            1.36µs ± 1%    0.02µs ± 0%   -98.57%  (p=0.001 n=7+7)
SplicePipeParallel-8     747ns ± 4%       4ns ± 0%   -99.41%  (p=0.001 n=7+7)

name                  old alloc/op   new alloc/op   delta
SplicePipe-8             24.0B ± 0%      0.0B       -100.00%  (p=0.001 n=7+7)
SplicePipeParallel-8     24.0B ± 0%      0.0B       -100.00%  (p=0.001 n=7+7)

name                  old allocs/op  new allocs/op  delta
SplicePipe-8              1.00 ± 0%      0.00       -100.00%  (p=0.001 n=7+7)
SplicePipeParallel-8      1.00 ± 0%      0.00       -100.00%  (p=0.001 n=7+7)

Fixes #42740

Change-Id: Idff654b7264342084e089b5ba796c87c380c471b
Reviewed-on: https://go-review.googlesource.com/c/go/+/271537
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>
Trust: Brad Fitzpatrick <bradfitz@golang.org>

4 years agocmd/go: allow '+' in package import paths in module mode
Bryan C. Mills [Tue, 9 Mar 2021 22:23:04 +0000 (17:23 -0500)]
cmd/go: allow '+' in package import paths in module mode

This change upgrades x/mod to pull in the fix from CL 300149.

Fixes #44776.

Change-Id: I273f41df2abfff76d91315b7f19fce851c8770d8
Reviewed-on: https://go-review.googlesource.com/c/go/+/300176
Trust: Bryan C. Mills <bcmills@google.com>
Run-TryBot: Bryan C. Mills <bcmills@google.com>
Reviewed-by: Jay Conrod <jayconrod@google.com>
TryBot-Result: Go Bot <gobot@golang.org>

4 years agocmd/compile/internal/types2: remove concept of finals
Robert Griesemer [Wed, 10 Mar 2021 01:14:15 +0000 (17:14 -0800)]
cmd/compile/internal/types2: remove concept of finals

This is a 1:1 port of the respective change in go/types
in https://golang.org/cl/299590.

Change-Id: I65ad723f2e21e3d95fc0b94665e0121e31871a48
Reviewed-on: https://go-review.googlesource.com/c/go/+/300250
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 agocmd/compile/internal/types2: better error reporting framework (starting point)
Robert Griesemer [Thu, 25 Feb 2021 22:54:04 +0000 (14:54 -0800)]
cmd/compile/internal/types2: better error reporting framework (starting point)

Until now, errors which came with additional details (e.g., a declaration
cycle error followed by the list of objects involved in the cycle, one per
line) were reported as an ordinary error followed by "secondary" errors,
with the secondary errors marked as such by having a tab-indented error
message.

This approach often required clients to filter these secondary errors
(as they are not new errors, they are just clarifying a previously
reported error).

This CL introduces a new internal error_ type which permits accumulating
various error information that may then be reported as a single error.

Change-Id: I25b2f094facd37e12737e517f7ef8853d465ff77
Reviewed-on: https://go-review.googlesource.com/c/go/+/296689
Trust: Robert Griesemer <gri@golang.org>
Reviewed-by: Robert Findley <rfindley@google.com>
4 years agogo/types: improve the positioning of broken import errors
Rob Findley [Tue, 9 Mar 2021 19:22:38 +0000 (14:22 -0500)]
go/types: improve the positioning of broken import errors

The heuristic gopls uses to guess error spans can get tripped-up on
certain valid characters in an import path (for example '-').

Update the error for broken imports to capture the full import path
span, so that gopls doesn't need to rely on heuristics.

Change-Id: Ieb8e0dce11933643f701b32271ff5f3477fecaaa
Reviewed-on: https://go-review.googlesource.com/c/go/+/300169
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/asm,cmd/compile: support 5 operand RLWNM/RLWMI on ppc64
Paul E. Murphy [Tue, 5 Jan 2021 22:44:43 +0000 (16:44 -0600)]
cmd/asm,cmd/compile: support 5 operand RLWNM/RLWMI on ppc64

These instructions are actually 5 argument opcodes as specified
by the ISA.  Prior to this patch, the MB and ME arguments were
merged into a single bitmask operand to workaround the limitations
of the ppc64 assembler backend.

This limitation no longer exists. Thus, we can pass operands for
these opcodes without having to merge the MB and ME arguments in
the assembler frontend or compiler backend.

Likewise, support for 4 operand variants is unchanged.

Change-Id: Ib086774f3581edeaadfd2190d652aaaa8a90daeb
Reviewed-on: https://go-review.googlesource.com/c/go/+/298750
Reviewed-by: Lynn Boger <laboger@linux.vnet.ibm.com>
Reviewed-by: Carlos Eduardo Seo <carlos.seo@linaro.org>
Trust: Carlos Eduardo Seo <carlos.seo@linaro.org>

4 years agoruntime, cmd/link/internal/ld: disable memory profiling when data unreachable
Brad Fitzpatrick [Mon, 8 Mar 2021 04:52:48 +0000 (20:52 -0800)]
runtime, cmd/link/internal/ld: disable memory profiling when data unreachable

If runtime.MemProfile is unreachable, default to not collecting any
memory profiling samples, to save memory on the hash table.

Fixes #42347

Change-Id: I9a4894a5fc77035fe59b1842e1ec77a1182e70c1
Reviewed-on: https://go-review.googlesource.com/c/go/+/299671
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Trust: Keith Randall <khr@golang.org>

4 years agonet: don't append a dot to TXT records on Plan 9
Fazlul Shahriar [Thu, 25 Feb 2021 20:16:25 +0000 (15:16 -0500)]
net: don't append a dot to TXT records on Plan 9

TXT records are not domain names, so no need to call absDomainName.
The output now matches the pure Go resolver.

Fixes #44619

Change-Id: I1ebf09152ff5c0446d2e2b4c26671358892d9dc9
Reviewed-on: https://go-review.googlesource.com/c/go/+/296589
Reviewed-by: David du Colombier <0intro@gmail.com>
Run-TryBot: David du Colombier <0intro@gmail.com>
TryBot-Result: Go Bot <gobot@golang.org>
Trust: David du Colombier <0intro@gmail.com>
Trust: Ian Lance Taylor <iant@golang.org>

4 years agocmd/compile: fix failure to communicate between ABIinfo producer&consumer
David Chase [Sun, 7 Mar 2021 19:00:10 +0000 (14:00 -0500)]
cmd/compile: fix failure to communicate between ABIinfo producer&consumer

ABI info producer and consumer had different ideas for register
order for parameters.

Includes a test, includes improvements to debugging output.

Updates #44816.

Change-Id: I4812976f7a6c08d6fc02aac1ec0544b1f141cca6
Reviewed-on: https://go-review.googlesource.com/c/go/+/299570
Trust: David Chase <drchase@google.com>
Reviewed-by: Than McIntosh <thanm@google.com>
4 years agocmd/compile: fix confusion in generating SelectN index
David Chase [Sun, 7 Mar 2021 01:59:40 +0000 (20:59 -0500)]
cmd/compile: fix confusion in generating SelectN index

Old: return the ABI register index of the result (wrong!)
New: return the index w/in sequence of result registers (right!)

Fixed bug:
genCaller0/genCaller0.go:43:9: internal compiler error: 'Caller0':
   panic during schedule while compiling Caller0:
runtime error: index out of range [10] with length 9

Updates #44816.

Change-Id: I1111e283658a2d6422986ae3d61bd95d1b9bde5e
Reviewed-on: https://go-review.googlesource.com/c/go/+/299549
Trust: David Chase <drchase@google.com>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
4 years agocmd/compile: fix broken type+offset calc for register args
David Chase [Sat, 6 Mar 2021 02:09:40 +0000 (21:09 -0500)]
cmd/compile: fix broken type+offset calc for register args

Includes more enhancements to debugging output.

Updates #44816.

Change-Id: I5b21815cf37ed21e7dec6c06f538090f32260203
Reviewed-on: https://go-review.googlesource.com/c/go/+/299409
Trust: David Chase <drchase@google.com>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
4 years agocmd/compile: fix OpArg decomposer for registers in expandCalls
David Chase [Sat, 6 Mar 2021 00:56:13 +0000 (19:56 -0500)]
cmd/compile: fix OpArg decomposer for registers in expandCalls

Includes test taken from
https://github.com/golang/go/issues/44816#issuecomment-791618179
and improved debugging output.

Updates #44816

Change-Id: I94aeb9c5255f175fe80727be29d218bad54bf7ea
Reviewed-on: https://go-review.googlesource.com/c/go/+/299389
Trust: David Chase <drchase@google.com>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
4 years agocmd/compile: remove skipping of implicit operations during export
Keith Randall [Mon, 8 Mar 2021 07:48:02 +0000 (23:48 -0800)]
cmd/compile: remove skipping of implicit operations during export

We'll need to attach types to these operations, so we need to
represent them in the import/export data.

Some of the operations use a selector indicating a different package,
so we need to provide an option to encode the package of a selector.
The default selector() function can't encode that extra information,
as selector's exact encoding is used by go/types.

Change-Id: I4c110fe347b3d915f88a722834bc4058baea7854
Reviewed-on: https://go-review.googlesource.com/c/go/+/299771
Trust: Keith Randall <khr@golang.org>
Run-TryBot: Keith Randall <khr@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Dan Scales <danscales@google.com>
4 years agocmd/compile: detect duplicate importing earlier
Keith Randall [Fri, 5 Mar 2021 17:50:28 +0000 (09:50 -0800)]
cmd/compile: detect duplicate importing earlier

Change-Id: I05ba944e189a884b727e40a9526d212612c3e923
Reviewed-on: https://go-review.googlesource.com/c/go/+/299770
Trust: Keith Randall <khr@golang.org>
Run-TryBot: Keith Randall <khr@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Dan Scales <danscales@google.com>
4 years agoruntime: add pollDesc partial edges
Michael Pratt [Tue, 9 Mar 2021 15:41:16 +0000 (10:41 -0500)]
runtime: add pollDesc partial edges

gscan is taken during stack growth, which may occur while pollDesc is
held.

mallocgc may also be called while pollDesc is held. mallocgc may take
mheap or mheapSpecial. The former exists, but is out of order; the
latter is missing.

Fixes #44881

Change-Id: Ie25935d9d433e813c11a528ee47255b317a09f41
Reviewed-on: https://go-review.googlesource.com/c/go/+/300009
Trust: Michael Pratt <mpratt@google.com>
Trust: Dan Scales <danscales@google.com>
Reviewed-by: Dan Scales <danscales@google.com>
Reviewed-by: Michael Knyszek <mknyszek@google.com>
Run-TryBot: Michael Pratt <mpratt@google.com>

4 years agocmd/compile: add debugging mode for import/export
Keith Randall [Fri, 5 Mar 2021 17:49:28 +0000 (09:49 -0800)]
cmd/compile: add debugging mode for import/export

Just add a simple magic number with each op, to detect when
the reader gets desynchronized from the writer.

Change-Id: Iac7dab7f465b0021b1d7ae31c8f8a353ac3663a2
Reviewed-on: https://go-review.googlesource.com/c/go/+/299769
Trust: Keith Randall <khr@golang.org>
Run-TryBot: Keith Randall <khr@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Dan Scales <danscales@google.com>
4 years agocmd/compile: get instantiated generic types working with interfaces
Dan Scales [Wed, 3 Mar 2021 21:33:27 +0000 (13:33 -0800)]
cmd/compile:  get instantiated generic types working with interfaces

Get instantiatiated generic types working with interfaces, including
typechecking assignments to interfaces and instantiating all the methods
properly. To get it all working, this change includes:

 - Add support for substituting in interfaces in subster.typ()

 - Fill in the info for the methods for all instantiated generic types,
   so those methods will be available for later typechecking (by the old
   typechecker) when assigning an instantiated generic type to an
   interface. We also want those methods available so we have the list
   when we want to instantiate all methods of an instantiated type. We
   have both for instantiated types encountered during the initial noder
   phase, and for instantiated types created during stenciling of a
   function/method.

 - When we first create a fully-instantiated generic type (whether
   during initial noder2 pass or while instantiating a method/function),
   add it to a list so that all of its methods will also be
   instantiated. This is needed so that an instantiated type can be
   assigned to an interface.

 - Properly substitute type names in the names of instantiated methods.

 - New accessor methods for types.Type.RParam.

 - To deal with generic types which are empty structs (or just don't use
   their type params anywhere), we want to set HasTParam if a named type
   has any type params that are not fully instantiated, even if the
   type param is not used in the type.

 - In subst.typ() and elsewhere, always set sym.Def for a new forwarding
   type we are creating, so we always create a single unique type for
   each generic type instantiation. This handles recursion within a
   type, and also recursive relationships across many types or methods.
   We remove the seen[] hashtable, which was serving the same purpose,
   but for subst.typ() only. We now handle all kinds of recursive types.

 - We don't seem to need to force types.CheckSize() on
   created/substituted generic types anymore, so commented out for now.

 - Add an RParams accessor to types2.Signature, and also a new
   exported types2.AsSignature() function.

Change-Id: If6c5dd98427b20bfe9de3379cc16f83df9c9b632
Reviewed-on: https://go-review.googlesource.com/c/go/+/298449
Run-TryBot: Dan Scales <danscales@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Trust: Dan Scales <danscales@google.com>
Reviewed-by: Robert Griesemer <gri@golang.org>
4 years agonet: use io.Discard in TestSendfileOnWriteTimeoutExceeded
Tobias Klauser [Tue, 9 Mar 2021 09:38:30 +0000 (10:38 +0100)]
net: use io.Discard in TestSendfileOnWriteTimeoutExceeded

Replace ioutil.Discard which is deprecated as of Go 1.16.

This was already done in CL 263142 but accidentially re-introduced in
CL 285914.

Change-Id: Ife0944d416294b1ba7c8f6b602aa68a3b9213c50
Reviewed-on: https://go-review.googlesource.com/c/go/+/299989
Trust: Tobias Klauser <tobias.klauser@gmail.com>
Run-TryBot: Tobias Klauser <tobias.klauser@gmail.com>
Reviewed-by: Emmanuel Odeke <emmanuel@orijtech.com>
TryBot-Result: Go Bot <gobot@golang.org>

4 years agocmd/link/internal/ld: deflake TestWindowsIssue36495
Brad Fitzpatrick [Tue, 9 Mar 2021 02:44:27 +0000 (18:44 -0800)]
cmd/link/internal/ld: deflake TestWindowsIssue36495

Over a dozen of the ld tests were missing closes. That was less
obvious before CL 299670 started using T.TempDir instead, which fails
a test when the tempdir can't be cleaned up (as it can't on Windows
when things are still open), insteading of leaving tempdirs around on
disk after the test.

Most of the missing closes were fixed in CL 299670, but the builders
helpfully pointed out that I missed at least this one.

Change-Id: I35f695bb7cbfba31e16311c5af965c148f9d7943
Reviewed-on: https://go-review.googlesource.com/c/go/+/299929
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Dan Scales <danscales@google.com>
Reviewed-by: Alex Brainman <alex.brainman@gmail.com>
Trust: Dan Scales <danscales@google.com>
Trust: Alex Brainman <alex.brainman@gmail.com>

4 years agocmd/compile: make ValAndOff.{Val,Off} return an int32
Alberto Donizetti [Fri, 5 Mar 2021 10:22:13 +0000 (11:22 +0100)]
cmd/compile: make ValAndOff.{Val,Off} return an int32

The ValAndOff type is a 64bit integer holding a 32bit value and a
32bit offset in each half, but for historical reasons its Val and Off
methods returned an int64. This was convenient when AuxInt was always
an int64, but now that AuxInts are typed we can return int32 from Val
and Off and get rid of a several casts and now unnecessary range
checks.

This change:

- changes the Val and Off methods to return an int32 (from int64);
- adds Val64 and Off64 methods for convenience in the few remaining
  places (in the ssa.go files) where Val and Off are stored in int64
  fields;
- deletes makeValAndOff64, renames makeValAndOff32 to makeValAndOff
- deletes a few ValAndOff methods that are now unused;
- removes several validOff/validValAndOff check that will always
  return true.

Passes:

  GOARCH=amd64 gotip build -toolexec 'toolstash -cmp' -a std
  GOARCH=386 gotip build -toolexec 'toolstash -cmp' -a std
  GOARCH=s390x gotip build -toolexec 'toolstash -cmp' -a std

(the three GOARCHs with SSA rules files impacted by the change).

Change-Id: I2abbbf42188c798631b94d3a55ca44256f140be7
Reviewed-on: https://go-review.googlesource.com/c/go/+/299149
Trust: Alberto Donizetti <alb.donizetti@gmail.com>
Trust: Keith Randall <khr@golang.org>
Run-TryBot: Alberto Donizetti <alb.donizetti@gmail.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
4 years agoruntime: document netpollclose
Tobias Klauser [Tue, 2 Mar 2021 08:10:49 +0000 (09:10 +0100)]
runtime: document netpollclose

Document that network poller implementations need to define
netpollclose.

Change-Id: Idc73dea7cfd503d4de7e1d95902f0f102cf5ed8f
Reviewed-on: https://go-review.googlesource.com/c/go/+/297809
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/compile: remove 4-byte alignment requirement of stack slot on arm
eric fang [Tue, 2 Mar 2021 06:02:55 +0000 (06:02 +0000)]
cmd/compile: remove 4-byte alignment requirement of stack slot on arm

This CL applies CL 267999 to arm.

Updates #42385

Change-Id: Iad82aafcb7b0a5a77a4bea32f648320f57a17cdd
Reviewed-on: https://go-review.googlesource.com/c/go/+/297773
Reviewed-by: eric fang <eric.fang@arm.com>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Trust: eric fang <eric.fang@arm.com>
Run-TryBot: eric fang <eric.fang@arm.com>

4 years agocmd/vet: bring in sigchanyzer to report unbuffered channels to signal.Notify
Emmanuel T Odeke [Sun, 7 Mar 2021 07:57:33 +0000 (23:57 -0800)]
cmd/vet: bring in sigchanyzer to report unbuffered channels to signal.Notify

Brings in the static analyzer "sigchanyzer", that we created at
Orijtech, Inc, and already submitted in CL 274352, as

    golang.org/x/tools/go/analysis/passes/sigchanyzer

and add it to cmd/vet as one of the passes.

Fixes #9399

Change-Id: I83708b8ea5ca8ede5ee31efab55cbce7419434ab
Reviewed-on: https://go-review.googlesource.com/c/go/+/299532
Run-TryBot: Emmanuel Odeke <emmanuel@orijtech.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Cuong Manh Le <cuong.manhle.vn@gmail.com>
Reviewed-by: Bryan C. Mills <bcmills@google.com>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Trust: Cuong Manh Le <cuong.manhle.vn@gmail.com>
Trust: Bryan C. Mills <bcmills@google.com>

4 years agoos/signal: remove comments about SA_RESTART
Ian Lance Taylor [Wed, 3 Mar 2021 16:04:40 +0000 (08:04 -0800)]
os/signal: remove comments about SA_RESTART

It's no longer necessary for non-Go signal handlers to use SA_RESTART.

For #20400
Fixes #44761

Change-Id: Ie3c7fecc631a4a2822331754296ea09b308e1391
Reviewed-on: https://go-review.googlesource.com/c/go/+/298269
Trust: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
Reviewed-by: Bryan C. Mills <bcmills@google.com>
TryBot-Result: Go Bot <gobot@golang.org>

4 years agocmd/link: use testing.T.TempDir in tests
Brad Fitzpatrick [Mon, 8 Mar 2021 04:52:39 +0000 (20:52 -0800)]
cmd/link: use testing.T.TempDir in tests

Change-Id: I6fc8c9ee6d2246bfd874eb58b411e34ddbeaf723
Reviewed-on: https://go-review.googlesource.com/c/go/+/299670
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Reviewed-by: Keith Randall <khr@golang.org>
4 years agocmd/compile: fix width not calculated for imported type
Cuong Manh Le [Mon, 8 Mar 2021 18:52:10 +0000 (01:52 +0700)]
cmd/compile: fix width not calculated for imported type

The compiler currently has problem that some imported type is missing
size calculation. The problem is not triggered until CL 283313 merged,
due to the compiler can compile the functions immediately when it sees
them, so during SSA generation, size calculation is still ok.

CL 283313 makes the compiler always push functions to compile queue,
then drain from it for compiling function. During this process, the
types calculation size is disabled, so calculating size during SSA now
make the compiler crashes.

To fix this, we can just always calculate type size during typechecking,
when importing type from other packages.

Fixes #44732

Change-Id: I8d00ea0b5aadd432154908280e55d85c75f3ce92
Reviewed-on: https://go-review.googlesource.com/c/go/+/299689
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: Dan Scales <danscales@google.com>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
4 years agocmd/link/internal/ld: fix typo in a comment
Brad Fitzpatrick [Mon, 8 Mar 2021 04:32:38 +0000 (20:32 -0800)]
cmd/link/internal/ld: fix typo in a comment

Change-Id: I6d0fafd38f0ad9392f163a9d8cd94d103bf2a059
Reviewed-on: https://go-review.googlesource.com/c/go/+/299669
Reviewed-by: Keith Randall <khr@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
4 years agocmd/compile: minor doc improvements
Josh Bleecher Snyder [Sun, 7 Mar 2021 22:48:08 +0000 (14:48 -0800)]
cmd/compile: minor doc improvements

These are left over from comments I failed to leave on CL 249463;
apparently I never hit "Reply".

Change-Id: Ia3f8a900703c347f8f98581ec1ac172c0f72cd9e
Reviewed-on: https://go-review.googlesource.com/c/go/+/299589
Trust: Josh Bleecher Snyder <josharian@gmail.com>
Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
4 years agoencoding/binary: limit bytes read by Uvarint to <= 10
Emmanuel T Odeke [Sun, 7 Mar 2021 07:14:21 +0000 (23:14 -0800)]
encoding/binary: limit bytes read by Uvarint to <= 10

Limits the number of bytes that can be consumed by Uvarint
to MaxVarintLen64 (10) to avoid wasted computations.
With this change, if Uvarint reads more than MaxVarintLen64
bytes, it'll return the erroring byte count of n=-(MaxVarintLen64+1)
which is -11, as per the function signature.

Updated some tests to reflect the new change in expectations of n
when the number of bytes to be read exceeds the limits..

Fixes #41185

Change-Id: Ie346457b1ddb0214b60c72e81128e24d604d083d
Reviewed-on: https://go-review.googlesource.com/c/go/+/299531
Run-TryBot: Emmanuel Odeke <emmanuel@orijtech.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
Trust: Emmanuel Odeke <emmanuel@orijtech.com>

4 years agocmd/compile: improve IsNonNil rule on riscv64
Joel Sing [Fri, 5 Mar 2021 17:46:07 +0000 (04:46 +1100)]
cmd/compile: improve IsNonNil rule on riscv64

IsNonNil is readily implemented using SNEZ on riscv64, removing over 8,000
instructions from the go binary. Other rules will improve on this sequence,
however in this case it makes sense to use a direct simplification.

Change-Id: Ib4068599532398afcd05f51d160673ef5fb5e5a0
Reviewed-on: https://go-review.googlesource.com/c/go/+/299230
Trust: Joel Sing <joel@sing.id.au>
Reviewed-by: Michael Munday <mike.munday@lowrisc.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Go Bot <gobot@golang.org>

4 years agocmd/compile: rename internal-abi.md to abi-internal.md
Matthew Dempsky [Sun, 7 Mar 2021 07:25:25 +0000 (23:25 -0800)]
cmd/compile: rename internal-abi.md to abi-internal.md

Allows muscle-memoried tab completion of cmd/compile/internal/...
paths to work again.

Change-Id: Ib54a5f2cc9fabcb876c2e62635828ab28b565501
Reviewed-on: https://go-review.googlesource.com/c/go/+/299530
Reviewed-by: Austin Clements <austin@google.com>
Trust: Matthew Dempsky <mdempsky@google.com>