]> Cypherpunks repositories - gostls13.git/log
gostls13.git
5 years ago[dev.link] cmd: default to new object files
Cherry Zhang [Wed, 9 Oct 2019 14:22:20 +0000 (10:22 -0400)]
[dev.link] cmd: default to new object files

Switch the default to new object files.

Internal linking cgo is disabled for now, as it does not work yet
in newobj mode.

Shared libraries are also broken.

Disable some tests that are known broken for now.

Change-Id: I8ca74793423861d607a2aa7b0d89a4f4d4ca7671
Reviewed-on: https://go-review.googlesource.com/c/go/+/200161
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Jeremy Faller <jeremy@golang.org>
5 years ago[dev.link] cmd/link/internal/objfile: relocate loader to new package
Than McIntosh [Mon, 14 Oct 2019 14:06:37 +0000 (10:06 -0400)]
[dev.link] cmd/link/internal/objfile: relocate loader to new package

Third change of several to update the loader API to reflect the final
consensus version of the loader API as described in Cherry's doc.
This piece:

   - move objfile.Loader into its own separate package, and update
     clients accordingly.

This includes a few minor cleanups, including converting a couple
of loader-related functions to methods, and privatizing some of the
loader methods such as ToGlobal/ToLocal.

Change-Id: Iae20585751a45491d8b19dcffc096aadae6bbfc6
Reviewed-on: https://go-review.googlesource.com/c/go/+/200998
Run-TryBot: Than McIntosh <thanm@google.com>
Reviewed-by: Jeremy Faller <jeremy@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

5 years ago[dev.link] cmd/internal/goobj2: provide accessor methods for flags
Cherry Zhang [Mon, 14 Oct 2019 15:17:18 +0000 (11:17 -0400)]
[dev.link] cmd/internal/goobj2: provide accessor methods for flags

Per Jeremy's comment in CL 199643. This makes the code read
better.

Change-Id: If270aecd712a27fb52e3faf5a4339200327d9ffe
Reviewed-on: https://go-review.googlesource.com/c/go/+/201023
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Jeremy Faller <jeremy@golang.org>
5 years ago[dev.link] cmd/link: let cgo import overwrite contentless data symbol
Cherry Zhang [Sat, 12 Oct 2019 05:05:03 +0000 (01:05 -0400)]
[dev.link] cmd/link: let cgo import overwrite contentless data symbol

A contentless data symbol may be a declaration of a cgo-imported
variable, e.g.

//go:cgo_import_dynamic xxx
var xxx uintptr

In this case, we want to mark the symbol imported, instead of
defined with zero value.

We used to load cgo directives before loading the object file, so
we'll mark the symbol SDYNIMPORT first. But in newobj mode,
currently we load cgo directives later. Letting SDYNIMPORT
overwrite contentless data symbol makes it work in both ordering.

Change-Id: I878f52086d6cdb5a347669bf8f848a49bce87b52
Reviewed-on: https://go-review.googlesource.com/c/go/+/201020
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Jeremy Faller <jeremy@golang.org>
5 years ago[dev.link] cmd/link: add dupok symbols resolved to another package to textp
Cherry Zhang [Fri, 11 Oct 2019 21:43:32 +0000 (17:43 -0400)]
[dev.link] cmd/link: add dupok symbols resolved to another package to textp

When a dupok symbol is resolved to another package, we still need
to record its presence in the current package, as the trampoline
pass expects packages are laid out in dependency order. At the
point after deadcode where we populate symbol contents for
reachable symbols (add relocations and read symbol data), make a
note of the dupok text symbols for each package. Later in
addToTextp we will visit packages in dependency order, process
the dup text symbol list for each package and select a final lib
for each dup text symbol.

Change-Id: Ib885e0a7e2343229d853aa629e3e337111df6011
Reviewed-on: https://go-review.googlesource.com/c/go/+/200797
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Than McIntosh <thanm@google.com>
5 years ago[dev.link] cmd/internal/obj, cmd/link: add InlTree in new object files
Cherry Zhang [Tue, 8 Oct 2019 21:22:20 +0000 (17:22 -0400)]
[dev.link] cmd/internal/obj, cmd/link: add InlTree in new object files

Add InlTree to the FuncInfo aux symbol in new object files.

In the linker, change InlinedCall.Func from a Symbol to a string,
as we only use its Name. (There was a use of Func.File, but that
use is not correct anyway.) So we don't need to create a Symbol
if not necessary.

Change-Id: I38ce568ae0934cd9cb6d0b30599f1c8d75444fc9
Reviewed-on: https://go-review.googlesource.com/c/go/+/200098
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Than McIntosh <thanm@google.com>
5 years ago[dev.link] cmd/link: implement symbol overwrite logic
Cherry Zhang [Tue, 8 Oct 2019 19:35:36 +0000 (15:35 -0400)]
[dev.link] cmd/link: implement symbol overwrite logic

If two defined symbols have the same name, one contentless and
one with content, the one with content "wins". This is mainly for
go:linkname on data symbols. Support this logic in newobj mode.

Introduce an "overwrite" mechanism, letting one symbol overwrite
another. This machanism could later be used for the linker
overwriting symbol contents (e.g. -X flag).

Change-Id: I32ee7d4b82df275f11b38c3abefc99b878ff12d7
Reviewed-on: https://go-review.googlesource.com/c/go/+/200097
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Than McIntosh <thanm@google.com>
5 years ago[dev.link] cmd/internal/obj, cmd/link: use aux symbol for DWARF symbols
Cherry Zhang [Tue, 8 Oct 2019 01:10:41 +0000 (21:10 -0400)]
[dev.link] cmd/internal/obj, cmd/link: use aux symbol for DWARF symbols

Use the auxiliary symbol mechanism to connect the text symbol and
its associated DWARF symbols. This way, the linker can track the
DWARF symbols from the text symbol, without looking up special
names.

Currently, in the linker this is only used in the deadcode pass
to track which DWARF symbols are used and need to load. Later
passes still use name lookup for now.

Change-Id: I2fe49f3b1f0ecc1472ae8aa93907cff740022d8d
Reviewed-on: https://go-review.googlesource.com/c/go/+/199801
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Than McIntosh <thanm@google.com>
5 years ago[dev.link] cmd/link: create sym.Symbols after deadcode in newobj mode
Cherry Zhang [Sat, 5 Oct 2019 02:05:41 +0000 (22:05 -0400)]
[dev.link] cmd/link: create sym.Symbols after deadcode in newobj mode

With the new object files, now we can run the deadcode pass on
indices instead of Symbol structs, so we can delay creating
Symbols after the deadcode pass. Then we only need to create
reachable symbols.

Not create Symbols in LoadNew and LoadRefs, and recombine
LoadReloc into LoadFull.

Split loadcgo into two parts: the first finds root symbols, the
second create Symbols and sets attributes. The first runs before
the deadcode pass, while the second runs after.

TODO: currently there are still symbols that are not marked
reachable but still used. This includes DWARF symbols, file
symbols, and type symbols that are referenced by DWARF symbols.
We still need to create them (conservatively).

Change-Id: I695779c9312be9d49ab1683957ac3e72e1f65a1e
Reviewed-on: https://go-review.googlesource.com/c/go/+/199643
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Than McIntosh <thanm@google.com>
5 years ago[dev.link] cmd/internal/obj: convert "\" to "/" in file path
Cherry Zhang [Sat, 12 Oct 2019 18:01:57 +0000 (14:01 -0400)]
[dev.link] cmd/internal/obj: convert "\" to "/" in file path

The old code does this. Do the same.

Change-Id: Ibf32ac347d6425e19ad0bc664c6b43ab5eba9c5e
Reviewed-on: https://go-review.googlesource.com/c/go/+/201022
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Than McIntosh <thanm@google.com>
5 years ago[dev.link] cmd/internal/goobj2: separate Autolib from Pkglist in new object file
Cherry Zhang [Wed, 9 Oct 2019 14:22:02 +0000 (10:22 -0400)]
[dev.link] cmd/internal/goobj2: separate Autolib from Pkglist in new object file

In CL 196030 we decided to combine the imported package list
(Autolib) and referenced package list (PkgIdx, or Pkglist).
However, in some cases the Autolib list may contain file name,
instead of package path, e.g.
https://go.googlesource.com/go/+/refs/heads/dev.link/src/cmd/compile/internal/gc/main.go#1181
And the linker needs that to locate the file. This mostly happens
with direct invocation of the compiler and linker (i.e., not
through "go build").

Instead of letting the linker make guess of the file name based
on the package path, make Autolib a separate list.

Change-Id: If195a69462d04db515346ee67cdec925f5a69e2e
Reviewed-on: https://go-review.googlesource.com/c/go/+/200157
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Jeremy Faller <jeremy@golang.org>
5 years ago[dev.link] cmd/link: on AIX, use relocation to locate target symbol from TOC symbol
Cherry Zhang [Sat, 12 Oct 2019 01:51:01 +0000 (21:51 -0400)]
[dev.link] cmd/link: on AIX, use relocation to locate target symbol from TOC symbol

On AIX, a TOC symbol always has a relocation to its target symbol.
Instead of using name lookup to locate the target symbol, we can
just use the relocation.

Using name lookup, besides being less efficient, needs to provide
the right symbol version. In this particular case, we are looking
for a data symbol so it is almost always version 0. But in case
that it is a text symbol, we may get an ABIALIAS symbol, which
does not have its Sect set.

Change-Id: I1ecfd284b04a86bbbc450059ee89d99d40493e51
Reviewed-on: https://go-review.googlesource.com/c/go/+/201019
Reviewed-by: Than McIntosh <thanm@google.com>
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

5 years ago[dev.link] cmd/internal/obj: sort ctxt.Data on AIX
Cherry Zhang [Sat, 12 Oct 2019 05:07:31 +0000 (01:07 -0400)]
[dev.link] cmd/internal/obj: sort ctxt.Data on AIX

On AIX, TOC symbols may be created and added to ctxt.Data
concurrently. To ensure reproducible builds, sort ctxt.Data.
This implements the same logic as WriteObjFile does for old
object files.

Change-Id: I2e6e2d7755352848981544a4fb68b828a188c2ca
Reviewed-on: https://go-review.googlesource.com/c/go/+/201021
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Than McIntosh <thanm@google.com>
5 years ago[dev.link] cmd/link: support reflect.Type.Method tracking in deadcode2
Cherry Zhang [Tue, 8 Oct 2019 22:21:22 +0000 (18:21 -0400)]
[dev.link] cmd/link: support reflect.Type.Method tracking in deadcode2

When reflect.Type.Method is called, all exported methods from a
reachable type need to be conservatively live. When such a
function is called, the compiler sets an attribute to the
function, and the linker needs to check that attribute. Implement
this in the index-based deadcode pass.

Unify symbol flags and FuncInfo flags to make things simpler. In
particular, the deadcode pass can check the reflectMethod
attribute without reading in and decoding FuncInfo.

Change-Id: Ibb21e172f2996e899c6efa5551a29d0eca62df67
Reviewed-on: https://go-review.googlesource.com/c/go/+/200099
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Than McIntosh <thanm@google.com>
5 years ago[dev.link] cmd/link: implement live method tracking in deadcode2
Cherry Zhang [Fri, 4 Oct 2019 19:08:58 +0000 (15:08 -0400)]
[dev.link] cmd/link: implement live method tracking in deadcode2

This essentially replicates the logic of live method tracking and
type symbol decoding, rewritten to operate on indices instead of
Symbols.

TODO: the special handling of reflect.Type.Method has not been
implemented.

TODO: the symbol name is used too much. It ought to be a better
way to do it.

Change-Id: I860ee7a506c00833902e4870d15aea698a705dd9
Reviewed-on: https://go-review.googlesource.com/c/go/+/199078
Reviewed-by: Than McIntosh <thanm@google.com>
5 years ago[dev.link] cmd/link: move macho host files to new loader format
Jeremy Faller [Fri, 11 Oct 2019 17:58:43 +0000 (13:58 -0400)]
[dev.link] cmd/link: move macho host files to new loader format

Change-Id: I823b19c0742992dd760c6372428a1936bb7c7e70
Reviewed-on: https://go-review.googlesource.com/c/go/+/200768
Reviewed-by: Jeremy Faller <jeremy@golang.org>
Reviewed-by: Than McIntosh <thanm@google.com>
Run-TryBot: Than McIntosh <thanm@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

5 years ago[dev.link] all: merge branch 'master' into dev.link
Cherry Zhang [Fri, 11 Oct 2019 18:07:53 +0000 (14:07 -0400)]
[dev.link] all: merge branch 'master' into dev.link

Clean merge.

Change-Id: If9bfb0f27f41563fd5d386de9c1081542c3ce498

5 years agoruntime: move sighandler into signal_unix.go
Ian Lance Taylor [Fri, 11 Oct 2019 16:52:11 +0000 (09:52 -0700)]
runtime: move sighandler into signal_unix.go

We couldn't do this before because sighandler was compiled for nacl.

Updates #30439

Change-Id: Ieec9938b6a1796c48d251cd8b1db1a42c25f3943
Reviewed-on: https://go-review.googlesource.com/c/go/+/200739
Run-TryBot: Ian Lance Taylor <iant@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

5 years agomime/multipart: add Part.NextRawPart to avoid QP decoding
Alex Buchanan [Thu, 6 Dec 2018 06:31:53 +0000 (22:31 -0800)]
mime/multipart: add Part.NextRawPart to avoid QP decoding

NextPart has automatic handling of quoted-printable encoding,
which is sometimes undesirable. NextRawPart adds a method
for reading a part while bypassing such automatic handling.

Fixes #29090

Change-Id: I6a042a4077c64091efa3f5dbecce0d9a34ac7065
Reviewed-on: https://go-review.googlesource.com/c/go/+/152877
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
5 years agonet/mail: added support to trailing CFWS in date
Constantin Konstantinidis [Wed, 5 Sep 2018 04:35:33 +0000 (06:35 +0200)]
net/mail: added support to trailing CFWS in date

RFC 5322 date format allows CFWS after the timezone.
If CFWS is valid, it is discarded and parsing is done as before
using time.Parse().
Existing test is extended with limit cases and invalid strings.

Fixes #22661

Change-Id: I54b96d7bc384b751962a76690e7e4786217a7941
Reviewed-on: https://go-review.googlesource.com/c/go/+/117596
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
5 years agoencoding/json: add Decoder.InputOffset for offset access
Francesco Renzi [Fri, 11 Oct 2019 12:59:15 +0000 (13:59 +0100)]
encoding/json: add Decoder.InputOffset for offset access

This makes Decoder.offset public while renaming it to
Decoder.InputOffset to match encoding/xml Decoder API

Code changes made by Adam Stankiewicz [sheerun@sher.pl]

Fixes #29688

Change-Id: I86dbfd2b2da80160846e92bfa580c53d8d45e2db
Reviewed-on: https://go-review.googlesource.com/c/go/+/200677
Run-TryBot: Johan Brandhorst <johan.brandhorst@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
5 years agocmd/compile/internal/ssa: fix block AuxIntType lookup
Michael Munday [Fri, 11 Oct 2019 15:50:12 +0000 (08:50 -0700)]
cmd/compile/internal/ssa: fix block AuxIntType lookup

Avoid an out-of-range error when calling LongString on a generic
block.

Change-Id: I33ca88940d899bc71e3155bc63d2aa925cf83230
Reviewed-on: https://go-review.googlesource.com/c/go/+/200737
Run-TryBot: Michael Munday <mike.munday@ibm.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Daniel Martí <mvdan@mvdan.cc>
5 years ago[dev.link] cmd/link/internal/objfile: update deadcode2 to use new reloc hooks
Than McIntosh [Fri, 11 Oct 2019 13:21:36 +0000 (09:21 -0400)]
[dev.link] cmd/link/internal/objfile: update deadcode2 to use new reloc hooks

Update the new deadcode pass to use the revised loader interface
for querying relocations. Remove some of the previous loader relocation
methods, since they are no longer used.

Change-Id: I08cec4c05793a17698b2674068f64837a5bf4477
Reviewed-on: https://go-review.googlesource.com/c/go/+/200718
Reviewed-by: Cherry Zhang <cherryyz@google.com>
5 years ago[dev.link] cmd/link/internal/objfile: new loader method for reloc queries
Than McIntosh [Fri, 11 Oct 2019 12:56:19 +0000 (08:56 -0400)]
[dev.link] cmd/link/internal/objfile: new loader method for reloc queries

Second change of several to update the loader API to reflect the final
consensus version of the loader API as described in Cherry's doc.
This piece:

 - define new loader.Relocs() method that returns a struct
   encapsulating a set of relocations on a global symbol

Old way of examining relocations:

  nreloc := loader.NReloc(someGlobalSymbolIndex)
  for i := 0; i < nreloc; i++ {
    tgtIdx := loader.RelocSym(someGlobalSymbolIndex, i))
    ... <do something with tgtIdx>
  }

New way of examining relocations:

  relocs := d.loader.Relocs(someGlobalSymbolIndex)
  for i := 0; i < relocs.Count; i++ {
    r := relocs.At(i).Sym
    ... <do something with r.Sym>
  }

Change-Id: I5bead1d729655ea13b3396647e53aafcd3e60f97
Reviewed-on: https://go-review.googlesource.com/c/go/+/200717
Run-TryBot: Than McIntosh <thanm@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
5 years agotest/fixedbugs: bump issue21576.go's timeout to 1min
Emmanuel T Odeke [Fri, 11 Oct 2019 06:40:39 +0000 (02:40 -0400)]
test/fixedbugs: bump issue21576.go's timeout to 1min

Increases the exec timeout from 5sec to 1min, but
also print out the error value on any test failure.

Fixes #34836

Change-Id: Ida2b8bd460243491ef0f90dfe0f978dfe02a0703
Reviewed-on: https://go-review.googlesource.com/c/go/+/200519
Run-TryBot: Emmanuel Odeke <emm.odeke@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Bryan C. Mills <bcmills@google.com>
5 years agocrypto/tls: consistently use %v for error formatting
fenos [Fri, 11 Oct 2019 13:30:39 +0000 (14:30 +0100)]
crypto/tls: consistently use %v for error formatting

We should keep a consistent way of formatting errors
in this file.

Fixes #34848

Change-Id: Ibb75908504f381fccab0281a42e788ef8c716b6f
Reviewed-on: https://go-review.googlesource.com/c/go/+/200679
Run-TryBot: Johan Brandhorst <johan.brandhorst@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Daniel Martí <mvdan@mvdan.cc>
5 years agoerrors: fix typo in TODO comment
davidsbond [Fri, 11 Oct 2019 13:12:55 +0000 (14:12 +0100)]
errors: fix typo in TODO comment

Fixes #34846
Change-Id: I24b3e65fc96ec85b2821480e9396c9d1663611c4
Reviewed-on: https://go-review.googlesource.com/c/go/+/200678
Run-TryBot: Johan Brandhorst <johan.brandhorst@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Johan Brandhorst <johan.brandhorst@gmail.com>
5 years agointernal/reflectlite: skip TestMirrorWithReflect reflect package source is unavailable
Bryan C. Mills [Fri, 11 Oct 2019 12:58:52 +0000 (08:58 -0400)]
internal/reflectlite: skip TestMirrorWithReflect reflect package source is unavailable

Fixes #34847

Change-Id: Ibb58c8820cdab8b9f3755fcfba754ad0c9ca982f
Reviewed-on: https://go-review.googlesource.com/c/go/+/200657
Run-TryBot: Bryan C. Mills <bcmills@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Elias Naur <mail@eliasnaur.com>
5 years agoruntime,internal/bytealg: optimize wasmZero, wasmMove, Compare
Agniva De Sarker [Thu, 10 Oct 2019 07:01:43 +0000 (12:31 +0530)]
runtime,internal/bytealg: optimize wasmZero, wasmMove, Compare

Coalesce set/get pairs into a tee.

Change-Id: I88ccdcb148465615437bebf24145e941a037e0a5
Reviewed-on: https://go-review.googlesource.com/c/go/+/200357
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Richard Musiol <neelance@gmail.com>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
5 years agoall: remove nacl (part 3, more amd64p32)
Brad Fitzpatrick [Thu, 10 Oct 2019 16:16:54 +0000 (16:16 +0000)]
all: remove nacl (part 3, more amd64p32)

Part 1: CL 199499 (GOOS nacl)
Part 2: CL 200077 (amd64p32 files, toolchain)
Part 3: stuff that arguably should've been part of Part 2, but I forgot
        one of my grep patterns when splitting the original CL up into
        two parts.

This one might also have interesting stuff to resurrect for any future
x32 ABI support.

Updates #30439

Change-Id: I2b4143374a253a003666f3c69e776b7e456bdb9c
Reviewed-on: https://go-review.googlesource.com/c/go/+/200318
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
5 years ago[dev.link] cmd/link: escape package path in objByPkg map
Cherry Zhang [Wed, 9 Oct 2019 19:37:46 +0000 (15:37 -0400)]
[dev.link] cmd/link: escape package path in objByPkg map

The package references recorded in the object file, which are
obtained from the compiler, are escaped. We should also use the
escaped package paths in the linker for resolving package
references.

Change-Id: I42eb12df6ff24330e6dc7bed1dc8224bb3b8a106
Reviewed-on: https://go-review.googlesource.com/c/go/+/200158
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Than McIntosh <thanm@google.com>
5 years ago[dev.link] cmd/internal/goobj: replace `"".` with package prefix in newobj mode
Cherry Zhang [Wed, 9 Oct 2019 19:53:03 +0000 (15:53 -0400)]
[dev.link] cmd/internal/goobj: replace `"".` with package prefix in newobj mode

This is the behavior of the old code. Do the same.

Change-Id: I3d393d754dcbdb7e76a577252a94214d2e316651
Reviewed-on: https://go-review.googlesource.com/c/go/+/200159
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Than McIntosh <thanm@google.com>
5 years ago[dev.link] cmd/internal/obj: support -S flag in newobj mode
Cherry Zhang [Wed, 9 Oct 2019 01:43:20 +0000 (21:43 -0400)]
[dev.link] cmd/internal/obj: support -S flag in newobj mode

When the compiler's -S flag is specified, it dumps the
disassembly. Add this when writing the new style object file.

Change-Id: I4cf85e57d22d0ceea1fda6d3b59fe363573659e7
Reviewed-on: https://go-review.googlesource.com/c/go/+/200100
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Than McIntosh <thanm@google.com>
5 years ago[dev.link] cmd/link: use index for deadcode
Cherry Zhang [Sun, 29 Sep 2019 02:42:35 +0000 (22:42 -0400)]
[dev.link] cmd/link: use index for deadcode

Switch the deadcode pass to use indices instead of Symbol
structures when using new object file format. Delay loading
symbol relocations and contents fully after the deadcode pass.
The next step is not to create Symbol structures until deadcode
is done.

Method tracking logic hasn't been implemented. Currently, all
methods of a reachable type are live.

Change-Id: Iffcd06ff84e6e52bd9eb24d1220d94234d18ab6b
Reviewed-on: https://go-review.googlesource.com/c/go/+/198199
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Than McIntosh <thanm@google.com>
5 years agotest: adjust a test to work with js/wasm's background goroutine
Brad Fitzpatrick [Thu, 10 Oct 2019 19:23:03 +0000 (19:23 +0000)]
test: adjust a test to work with js/wasm's background goroutine

Fixes #34768

Change-Id: Ic73591f620cdee5bc7203483902e6ba98d2c442b
Reviewed-on: https://go-review.googlesource.com/c/go/+/200438
Reviewed-by: Bryan C. Mills <bcmills@google.com>
5 years agoruntime: remove no-op pointer writes in treap rotations
alan [Thu, 10 Oct 2019 18:56:27 +0000 (18:56 +0000)]
runtime: remove no-op pointer writes in treap rotations

Change-Id: If5a272f331fe9da09467efedd0231a4ce34db0f8
GitHub-Last-Rev: 4b81a79a92db4b51001ce6660b24c760fd3b630b
GitHub-Pull-Request: golang/go#28420
Reviewed-on: https://go-review.googlesource.com/c/go/+/144999
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Rick Hudson <rlh@golang.org>
5 years ago[dev.link] cmd/link/internal/objfile: new 'Sym' type for global symbol index
Than McIntosh [Wed, 9 Oct 2019 13:04:16 +0000 (09:04 -0400)]
[dev.link] cmd/link/internal/objfile: new 'Sym' type for global symbol index

First change of several to update the loader API to reflect the final
consensus version of the loader API as described in Cherry's doc.

This piece:

 - define new loader.Sym type to encapsulate a global symbol
   index (as opposed to just using 'int')

Change-Id: I6f6483e269f80abfc7d373b2856b2c0d61b9ac24
Reviewed-on: https://go-review.googlesource.com/c/go/+/200417
Run-TryBot: Than McIntosh <thanm@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Austin Clements <austin@google.com>
Reviewed-by: Jeremy Faller <jeremy@golang.org>
5 years agocmd/go: forbid module pattern 'all' when outside a module
Jay Conrod [Thu, 10 Oct 2019 14:26:21 +0000 (10:26 -0400)]
cmd/go: forbid module pattern 'all' when outside a module

Also, in cmd/doc, avoid calling 'go list -m all' when in module mode
outside a module since it's now an error.

Fixes #32027

Change-Id: I7224c7fdf7e950bce6c058ab2a5837c27ba3b899
Reviewed-on: https://go-review.googlesource.com/c/go/+/200297
Run-TryBot: Jay Conrod <jayconrod@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Bryan C. Mills <bcmills@google.com>
5 years agocmd/compile: walk progs to generate debug_lines data
Jeremy Faller [Wed, 9 Oct 2019 14:44:33 +0000 (10:44 -0400)]
cmd/compile: walk progs to generate debug_lines data

Walking the progs is simpler than using the is_stmt symbol shenanigans.
This is a reinstatement of CL 196661, which was rolled back due to tests
failing. Unlike that original CL, this change should output the same
debug_lines data the original approach wrote.

The stats for JUST this CLC, note teh small speedup in compilation, and
the lack of difference in binary size.

name                      old time/op       new time/op       delta
Template                        229ms ± 4%        218ms ± 1%  -4.95%  (p=0.000 n=10+8)
Unicode                        92.6ms ± 9%       88.6ms ±13%    ~     (p=0.089 n=10+10)
GoTypes                         850ms ± 2%        831ms ± 4%  -2.23%  (p=0.009 n=10+10)
Compiler                        3.99s ± 1%        3.93s ± 1%  -1.29%  (p=0.000 n=10+9)
SSA                             13.7s ± 1%        13.7s ± 1%    ~     (p=0.912 n=10+10)
Flate                           140ms ± 3%        138ms ± 3%  -1.90%  (p=0.009 n=10+10)
GoParser                        172ms ± 2%        169ms ± 4%    ~     (p=0.095 n=9+10)
Reflect                         530ms ± 3%        516ms ± 5%    ~     (p=0.052 n=10+10)
Tar                             202ms ± 1%        196ms ± 3%  -2.83%  (p=0.002 n=9+10)
XML                             280ms ± 3%        270ms ± 4%  -3.48%  (p=0.009 n=10+10)
LinkCompiler                    927ms ± 2%        907ms ± 4%    ~     (p=0.052 n=10+10)
ExternalLinkCompiler            1.97s ± 2%        1.97s ± 3%    ~     (p=0.853 n=10+10)
LinkWithoutDebugCompiler        549ms ± 3%        543ms ± 5%    ~     (p=0.481 n=10+10)
StdCmd                          12.0s ± 1%        12.0s ± 1%    ~     (p=0.905 n=9+10)

name                      old user-time/op  new user-time/op  delta
Template                        372ms ±18%        344ms ±11%    ~     (p=0.190 n=10+10)
Unicode                         264ms ±23%        241ms ±43%    ~     (p=0.315 n=8+10)
GoTypes                         1.56s ±22%        1.68s ± 5%    ~     (p=0.237 n=10+8)
Compiler                        7.41s ± 2%        7.31s ± 3%    ~     (p=0.123 n=10+10)
SSA                             24.5s ± 2%        24.7s ± 1%    ~     (p=0.133 n=10+9)
Flate                           199ms ± 6%        188ms ±28%    ~     (p=0.353 n=10+10)
GoParser                        243ms ±11%        240ms ± 6%    ~     (p=0.968 n=10+9)
Reflect                         929ms ±21%        862ms ±35%    ~     (p=0.190 n=10+10)
Tar                             284ms ± 9%        296ms ±17%    ~     (p=0.497 n=9+10)
XML                             386ms ±21%        398ms ±28%    ~     (p=1.000 n=9+10)
LinkCompiler                    1.13s ± 9%        1.12s ± 8%    ~     (p=0.546 n=9+9)
ExternalLinkCompiler            2.37s ±15%        2.30s ± 9%    ~     (p=0.549 n=10+9)
LinkWithoutDebugCompiler        646ms ±10%        642ms ±13%    ~     (p=0.853 n=10+10)

name                      old alloc/op      new alloc/op      delta
Template                       36.5MB ± 0%       36.5MB ± 0%  -0.11%  (p=0.000 n=10+9)
Unicode                        28.5MB ± 0%       28.5MB ± 0%    ~     (p=0.190 n=10+10)
GoTypes                         121MB ± 0%        121MB ± 0%  -0.10%  (p=0.000 n=9+10)
Compiler                        549MB ± 0%        549MB ± 0%  -0.10%  (p=0.000 n=9+10)
SSA                            1.92GB ± 0%       1.92GB ± 0%  -0.13%  (p=0.000 n=10+10)
Flate                          23.0MB ± 0%       23.0MB ± 0%  -0.07%  (p=0.000 n=10+10)
GoParser                       27.9MB ± 0%       27.9MB ± 0%  -0.09%  (p=0.000 n=10+10)
Reflect                        77.9MB ± 0%       77.8MB ± 0%  -0.13%  (p=0.000 n=9+10)
Tar                            34.5MB ± 0%       34.4MB ± 0%  -0.09%  (p=0.000 n=10+10)
XML                            44.3MB ± 0%       44.3MB ± 0%  -0.08%  (p=0.000 n=10+10)
LinkCompiler                    229MB ± 0%        225MB ± 0%  -1.74%  (p=0.000 n=10+10)
ExternalLinkCompiler            233MB ± 0%        242MB ± 0%  +3.81%  (p=0.000 n=10+10)
LinkWithoutDebugCompiler        156MB ± 0%        152MB ± 0%  -2.29%  (p=0.000 n=10+9)

name                      old allocs/op     new allocs/op     delta
Template                         373k ± 0%         373k ± 0%  -0.21%  (p=0.000 n=10+10)
Unicode                          340k ± 0%         340k ± 0%  -0.04%  (p=0.000 n=10+10)
GoTypes                         1.33M ± 0%        1.33M ± 0%  -0.20%  (p=0.000 n=10+9)
Compiler                        5.39M ± 0%        5.38M ± 0%  -0.16%  (p=0.000 n=10+10)
SSA                             18.3M ± 0%        18.2M ± 0%  -0.15%  (p=0.000 n=10+10)
Flate                            235k ± 0%         234k ± 0%  -0.23%  (p=0.000 n=10+7)
GoParser                         309k ± 0%         308k ± 0%  -0.20%  (p=0.000 n=10+10)
Reflect                          970k ± 0%         968k ± 0%  -0.30%  (p=0.000 n=10+10)
Tar                              347k ± 0%         347k ± 0%  -0.22%  (p=0.000 n=10+10)
XML                              425k ± 0%         424k ± 0%  -0.16%  (p=0.000 n=10+10)
LinkCompiler                     602k ± 0%         601k ± 0%  -0.03%  (p=0.000 n=9+10)
ExternalLinkCompiler            1.65M ± 0%        1.65M ± 0%  -0.02%  (p=0.000 n=10+10)
LinkWithoutDebugCompiler         220k ± 0%         220k ± 0%  -0.03%  (p=0.016 n=10+9)

name                      old object-bytes  new object-bytes  delta
Template                        553kB ± 0%        553kB ± 0%  -0.01%  (p=0.000 n=10+10)
Unicode                         215kB ± 0%        215kB ± 0%    ~     (all equal)
GoTypes                        2.02MB ± 0%       2.02MB ± 0%  -0.00%  (p=0.000 n=10+10)
Compiler                       7.98MB ± 0%       7.98MB ± 0%  -0.01%  (p=0.000 n=10+10)
SSA                            27.1MB ± 0%       27.1MB ± 0%  -0.00%  (p=0.000 n=10+10)
Flate                           340kB ± 0%        340kB ± 0%  -0.01%  (p=0.000 n=10+10)
GoParser                        434kB ± 0%        434kB ± 0%  -0.00%  (p=0.000 n=10+10)
Reflect                        1.34MB ± 0%       1.34MB ± 0%  -0.01%  (p=0.000 n=10+10)
Tar                             479kB ± 0%        479kB ± 0%  -0.00%  (p=0.000 n=10+10)
XML                             618kB ± 0%        618kB ± 0%  -0.01%  (p=0.000 n=10+10)

name                      old export-bytes  new export-bytes  delta
Template                       20.4kB ± 0%       20.4kB ± 0%    ~     (all equal)
Unicode                        8.21kB ± 0%       8.21kB ± 0%    ~     (all equal)
GoTypes                        36.6kB ± 0%       36.6kB ± 0%    ~     (all equal)
Compiler                        116kB ± 0%        116kB ± 0%  +0.00%  (p=0.000 n=10+10)
SSA                             141kB ± 0%        141kB ± 0%  +0.00%  (p=0.000 n=10+10)
Flate                          5.10kB ± 0%       5.10kB ± 0%    ~     (all equal)
GoParser                       8.92kB ± 0%       8.92kB ± 0%    ~     (all equal)
Reflect                        11.8kB ± 0%       11.8kB ± 0%    ~     (all equal)
Tar                            10.9kB ± 0%       10.9kB ± 0%    ~     (all equal)
XML                            17.4kB ± 0%       17.4kB ± 0%    ~     (all equal)

name                      old text-bytes    new text-bytes    delta
HelloSize                       742kB ± 0%        742kB ± 0%    ~     (all equal)
CmdGoSize                      10.6MB ± 0%       10.6MB ± 0%    ~     (all equal)

name                      old data-bytes    new data-bytes    delta
HelloSize                      10.7kB ± 0%       10.7kB ± 0%    ~     (all equal)
CmdGoSize                       312kB ± 0%        312kB ± 0%    ~     (all equal)

name                      old bss-bytes     new bss-bytes     delta
HelloSize                       122kB ± 0%        122kB ± 0%    ~     (all equal)
CmdGoSize                       146kB ± 0%        146kB ± 0%    ~     (all equal)

name                      old exe-bytes     new exe-bytes     delta
HelloSize                      1.10MB ± 0%       1.10MB ± 0%    ~     (all equal)
CmdGoSize                      14.9MB ± 0%       14.9MB ± 0%  -0.03%  (p=0.000 n=10+10)

Change-Id: Ie078a42b29353b96654fa1f0f47d600b5a53762d
Reviewed-on: https://go-review.googlesource.com/c/go/+/200017
Reviewed-by: Jeremy Faller <jeremy@golang.org>
Reviewed-by: David Chase <drchase@google.com>
Run-TryBot: Jeremy Faller <jeremy@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

5 years agoencoding/json: support TextUnmarshaler for map keys with string underlying types
Cuong Manh Le [Thu, 10 Oct 2019 03:19:12 +0000 (10:19 +0700)]
encoding/json: support TextUnmarshaler for map keys with string underlying types

When unmarshaling to a map, the map's key type must either be a string,
an integer, or implement encoding.TextUnmarshaler. But for a user
defined type, reflect.Kind will not distinguish between the static type
and the underlying type. In:

var x MyString = "x"
t := reflect.TypeOf(x)
println(t.Kind() == reflect.String)

the Kind of x is still reflect.String, even though the static type of x
is MyString.

Moreover, checking for the map's key type is a string occurs first, so
even if the map key type MyString implements encoding.TextUnmarshaler,
it will be ignored.

To fix the bug, check for encoding.TextUnmarshaler first.

Fixes #34437

Change-Id: I780e0b084575e1dddfbb433fe03857adf71d05fb
Reviewed-on: https://go-review.googlesource.com/c/go/+/200237
Run-TryBot: Cuong Manh Le <cuong.manhle.vn@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Daniel Martí <mvdan@mvdan.cc>
5 years agoencoding/json: stop escaping U+2028 and U+2029 in Compact
Tom Thorogood [Thu, 10 Oct 2019 10:48:00 +0000 (10:48 +0000)]
encoding/json: stop escaping U+2028 and U+2029 in Compact

Compact has been inconsistently escaping only some problematic characters
(U+2028 and U+2029), but not others (<, > and &). This change addresses
this inconsistency by removing the escaping of U+2028 and U+2029.

Callers who need to escape the output of Compact should use HTMLEscape
which escapes <, >, &, U+2028 and U+2029.

Fixes #34070
Fixes #30357
Updates #5836

Change-Id: Icfce7691d2b8b1d9b05ba7b64d2d1e4f3b67871b
GitHub-Last-Rev: 38859fe3e2fd586bbd45175c2742f7b123836bf3
GitHub-Pull-Request: golang/go#34804
Reviewed-on: https://go-review.googlesource.com/c/go/+/200217
Reviewed-by: Daniel Martí <mvdan@mvdan.cc>
Run-TryBot: Daniel Martí <mvdan@mvdan.cc>
TryBot-Result: Gobot Gobot <gobot@golang.org>

5 years agocmd/go: forbid resolving import to modules when outside of a module
Jay Conrod [Wed, 2 Oct 2019 21:51:54 +0000 (17:51 -0400)]
cmd/go: forbid resolving import to modules when outside of a module

When in module mode outside of any module, 'go build' and most other
commands will now report an error instead of resolving a package path
to a module.

Previously, most commands would attempt to find the latest version of
a module providing the package. This could be very slow if many
packages needed to be resolved this way. Since there is no go.mod file
where module requirements can be saved, it's a repeatedly slow and
confusing experience.

After this change, 'go build' and other commands may still be used
outside of a module on packages in std and source files (.go
arguments) that only import packages in std. Listing any other package
on the command line or importing a package outside std will cause an
error.

'go get' is exempted from the new behavior, since it's expected that
'go get' resolves paths to modules at new versions.

Updates #32027

Change-Id: Ia9d3a3b4ad738ca5423472e17818d62b96a2c959
Reviewed-on: https://go-review.googlesource.com/c/go/+/198778
Run-TryBot: Jay Conrod <jayconrod@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Bryan C. Mills <bcmills@google.com>
5 years agocmd/doc: show original import error when package cannot be found
Jay Conrod [Tue, 8 Oct 2019 19:16:41 +0000 (15:16 -0400)]
cmd/doc: show original import error when package cannot be found

Updates #34669

Change-Id: I8d0ee68885e804e131f42a512080486f9b25e9dd
Reviewed-on: https://go-review.googlesource.com/c/go/+/199819
Run-TryBot: Jay Conrod <jayconrod@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Bryan C. Mills <bcmills@google.com>
5 years agogo/build: import packages in module mode when GO111MODULE is "on"
Jay Conrod [Tue, 8 Oct 2019 19:05:41 +0000 (15:05 -0400)]
go/build: import packages in module mode when GO111MODULE is "on"

go/build.Import locates package dirctories using 'go list' when in
module mode (finding, downloading, and extracting modules is
complicated, so go/build does not handle it).

Previously, Import used 'go list' if GO111MODULE was not explicitly
off and a go.mod file was present (plus some other conditions). With
this change, if GO111MODULE is "on", a go.mod file does not need to be
present.

Fixes #34669

Change-Id: I9e56871054d4b07c3fc04b6f14a5c8c8e9f3c333
Reviewed-on: https://go-review.googlesource.com/c/go/+/199818
Run-TryBot: Jay Conrod <jayconrod@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Bryan C. Mills <bcmills@google.com>
5 years agoall: remove the nacl port (part 2, amd64p32 + toolchain)
Brad Fitzpatrick [Wed, 9 Oct 2019 16:22:47 +0000 (16:22 +0000)]
all: remove the nacl port (part 2, amd64p32 + toolchain)

This is part two if the nacl removal. Part 1 was CL 199499.

This CL removes amd64p32 support, which might be useful in the future
if we implement the x32 ABI. It also removes the nacl bits in the
toolchain, and some remaining nacl bits.

Updates #30439

Change-Id: I2475d5bb066d1b474e00e40d95b520e7c2e286e1
Reviewed-on: https://go-review.googlesource.com/c/go/+/200077
Reviewed-by: Ian Lance Taylor <iant@golang.org>
5 years agodog/go1.14: properly close code tags
Ainar Garipov [Wed, 9 Oct 2019 20:48:31 +0000 (23:48 +0300)]
dog/go1.14: properly close code tags

Some code tags in the HTML were not properly closed. Close them so that
the text is rendered correctly.

Change-Id: I5c2170ffced313417f65004d53518128c34f7979
Reviewed-on: https://go-review.googlesource.com/c/go/+/200117
Reviewed-by: Ian Lance Taylor <iant@golang.org>
5 years ago[dev.link] cmd/link: change some decodetype functions to operate on bytes
Cherry Zhang [Thu, 3 Oct 2019 22:25:21 +0000 (18:25 -0400)]
[dev.link] cmd/link: change some decodetype functions to operate on bytes

Change some decodetype functions to operate on bytes nstead of
Symbol. This is in preparation of implementing live method
tracking in index-based deadcode pass, and reducing/eliminating
sym.Symbol in general.

Change-Id: Ia9809ad7b182884225e1bda577e8dbec0cd216c5
Reviewed-on: https://go-review.googlesource.com/c/go/+/199077
Reviewed-by: Austin Clements <austin@google.com>
5 years ago[dev.link] cmd/link, cmd/internal/goobj2: adopt new DWARF compilation unit logic...
Cherry Zhang [Mon, 30 Sep 2019 15:43:41 +0000 (11:43 -0400)]
[dev.link] cmd/link, cmd/internal/goobj2: adopt new DWARF compilation unit logic with new object file

The dev.link branch was not sync'd with the new DWARF compilation
unit logic change on the master branch, and the new object file
format didn't support this.

This CL adds the new DWARF CU and file table support to the new
object file format. In the old object file, the DWARF file table
is a separate section. For now, we do the same with the new
object file, keeping it as a separate block.

While here, also refactor the loader code so it is easier for the
loader to carry per-object informations.

Change-Id: I4c317941fc0a5831acbc11ce8c2a8b7421471372
Reviewed-on: https://go-review.googlesource.com/c/go/+/198198
Reviewed-by: Austin Clements <austin@google.com>
5 years ago[dev.link] cmd/link, cmd/internal/goobj2: mmap object file in -newobj mode
Cherry Zhang [Fri, 27 Sep 2019 18:49:44 +0000 (14:49 -0400)]
[dev.link] cmd/link, cmd/internal/goobj2: mmap object file in -newobj mode

With the old object file format, we use mmap (if supported) to
read object files and back symbol data with mapped read-only
memory.

Do the same with the new object file format. This also
significantly reduces number of syscalls made to read object
files.

Currently we still do mmap in object file level, not archive
level. This is probably ok, as there shouldn't be many archives
that contain more than one object. If this is a problem we can
change that later.

Change-Id: Icae3ef14d8ed6adbee1b5b48d420e2af22fd9604
Reviewed-on: https://go-review.googlesource.com/c/go/+/197797
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Than McIntosh <thanm@google.com>
5 years agocmd/go: eliminate redundancy in import error messages
Jay Conrod [Tue, 8 Oct 2019 17:16:33 +0000 (13:16 -0400)]
cmd/go: eliminate redundancy in import error messages

This change introduces a new interface, load.ImportPathError. An error
may satisfy this by providing an ImportPath method and including the
import path in its error text. modload.ImportMissingError satisfies
this interface. load.ImportErrorf also provides a convenient way to
create an error satisfying this interface with an arbitrary message.

When load.PackageError formats its error text, it may omit the last
path on the import stack if the wrapped error satisfies
ImportPathError and has a matching path.

To make this work, PackageError.Err is now an error instead of a
string. PackageError.MarshalJSON will write Err as a string for
'go list -json' output.

When go/build.Import invokes 'go list' in module mode, it now runs
with '-e' and includes '.Error' in the output format instead of
expecting the error to be in the raw stderr text. If a package error
is printed and a directory was not found, the error will be returned
without extra decoration.

Fixes #34752

Change-Id: I2d81dab7dec19e0ae9f51f6412bc9f30433a8596
Reviewed-on: https://go-review.googlesource.com/c/go/+/199840
Run-TryBot: Jay Conrod <jayconrod@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Bryan C. Mills <bcmills@google.com>
5 years agoRevert "cmd/go: fail if a test binary exits with no output"
Bryan C. Mills [Wed, 9 Oct 2019 20:13:40 +0000 (20:13 +0000)]
Revert "cmd/go: fail if a test binary exits with no output"

This reverts CL 184457.

Reason for revert: introduced failures in the regression test for #18153.

Fixes #34791
Updates #29062

Change-Id: I4040965163f809083c023be055e69b1149d6214e
Reviewed-on: https://go-review.googlesource.com/c/go/+/200106
Run-TryBot: Bryan C. Mills <bcmills@google.com>
Reviewed-by: Alexander Rakoczy <alex@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

5 years agocmd/go/internal/modfile: remove preceding empty lines when setting require
Jason A. Donenfeld [Tue, 8 Oct 2019 21:06:30 +0000 (23:06 +0200)]
cmd/go/internal/modfile: remove preceding empty lines when setting require

When rewriting a go.mod file, we currently sort all of the require
lines in a block. The way the parser works is that it considers
preceding blank lines to be empty comment lines, and preceding empty
comment lines are "owned" by their adjoining line. So when we go to sort
them, the empty lines follow around each sorted entry, which doesn't
make a whole lot of sense, since usually vertical space is inserted to
show sections, and if things get moved around by sorting, those sections
are no longer meaningful. This all results in one especially troublesome
edge case: blank lines between a block opening ("require (") and the
first block line ("golang.org/x/sys ...") are not treated the same way
and are rewritten out of existence.

Here's an example of the behavior this fixes.

Starting input file:

    require (
        golang.zx2c4.com/wireguard master

        golang.org/x/crypto latest
        golang.org/x/net latest
        golang.org/x/sys latest
        golang.org/x/text latest

        github.com/lxn/walk latest
        github.com/lxn/win latest
    )

Now we run this through `GOPROXY=direct go get -d`:

    require (

        github.com/lxn/walk v0.0.0-20190619151032-86d8802c197a
        github.com/lxn/win v0.0.0-20190716185335-d1d36f0e4f48

        golang.org/x/crypto v0.0.0-20190820162420-60c769a6c586
        golang.org/x/net v0.0.0-20190813141303-74dc4d7220e7
        golang.org/x/sys v0.0.0-20190813064441-fde4db37ae7a
        golang.org/x/text v0.3.2
        golang.zx2c4.com/wireguard v0.0.20190806-0.20190822065259-3cedc22d7b49
    )

Notice how the blank lines before lxn/walk and x/crypto were preserved.

Finally, we have this be rewritten yet again with a call to `go build`:

    require (
        github.com/lxn/walk v0.0.0-20190619151032-86d8802c197a
        github.com/lxn/win v0.0.0-20190716185335-d1d36f0e4f48

        golang.org/x/crypto v0.0.0-20190820162420-60c769a6c586
        golang.org/x/net v0.0.0-20190813141303-74dc4d7220e7
        golang.org/x/sys v0.0.0-20190813064441-fde4db37ae7a
        golang.org/x/text v0.3.2
        golang.zx2c4.com/wireguard v0.0.20190806-0.20190822065259-3cedc22d7b49
    )

In this final resting point, the first blank line has been removed.

The discrepancy between those two last stages are especially bothersome,
because it makes for lots of dirty git commits and file contents
bouncing back and forth.

This commit fixes the problem as mentioned above, getting rid of those
preceding blank lines. The output in all cases looks as it should, like
this:

    require (
        github.com/lxn/walk v0.0.0-20190619151032-86d8802c197a
        github.com/lxn/win v0.0.0-20190716185335-d1d36f0e4f48
        golang.org/x/crypto v0.0.0-20190820162420-60c769a6c586
        golang.org/x/net v0.0.0-20190813141303-74dc4d7220e7
        golang.org/x/sys v0.0.0-20190813064441-fde4db37ae7a
        golang.org/x/text v0.3.2
        golang.zx2c4.com/wireguard v0.0.20190806-0.20190822065259-3cedc22d7b49
    )

Fixes #33779

Change-Id: I11c894440bd35f343ee62db3e06a50fa871f2599
Reviewed-on: https://go-review.googlesource.com/c/go/+/199917
Run-TryBot: Jason A. Donenfeld <Jason@zx2c4.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Bryan C. Mills <bcmills@google.com>
5 years agonet/http: update bundled x/net/http2
Emmanuel T Odeke [Wed, 9 Oct 2019 18:36:09 +0000 (14:36 -0400)]
net/http: update bundled x/net/http2

Updates x/net/http2 to git rev d66e71096ffb9f08f36d9aefcae80ce319de6d68

    http2: end stream eagerly after sending the request body
    https://golang.org/cl/181157 (fixes #32254)

    all: fix typos
    https://golang.org/cl/193799

    http2: fix memory leak in random write scheduler
    https://golang.org/cl/198462 (fixes #33812)

    http2: do not sniff body if Content-Encoding is set
    https://golang.org/cl/199841 (updates #31753)

Also unskips tests from CL 199799.

Change-Id: I241c0b1cd18cad5041485be92809137a973e33bd
Reviewed-on: https://go-review.googlesource.com/c/go/+/200102
Run-TryBot: Emmanuel Odeke <emm.odeke@gmail.com>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

5 years agonet/http: do not sniff response if Content-Encoding header is set
Emmanuel T Odeke [Tue, 8 Oct 2019 07:23:08 +0000 (00:23 -0700)]
net/http: do not sniff response if Content-Encoding header is set

Fixes #31753

Change-Id: I32ec5906ef6714e19b094f67cb0f10a211a9c500
Reviewed-on: https://go-review.googlesource.com/c/go/+/199799
Run-TryBot: Emmanuel Odeke <emm.odeke@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
5 years agocmd/go/internal/list: disallow 'list -m' with '-mod=vendor'
Bryan C. Mills [Tue, 8 Oct 2019 18:51:30 +0000 (14:51 -0400)]
cmd/go/internal/list: disallow 'list -m' with '-mod=vendor'

Updates #33848

Change-Id: I81663386297282397ce1b546a8b15597bfbcea78
Reviewed-on: https://go-review.googlesource.com/c/go/+/199821
Run-TryBot: Bryan C. Mills <bcmills@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Jay Conrod <jayconrod@google.com>
5 years agocmd/go: automatically check and use vendored packages
Bryan C. Mills [Tue, 1 Oct 2019 19:48:17 +0000 (15:48 -0400)]
cmd/go: automatically check and use vendored packages

This implements the proposal described in
https://golang.org/issue/33848#issuecomment-537222782.

Fixes #33848

Change-Id: Ia34d6500ca396b6aa644b920233716c6b83ef729
Reviewed-on: https://go-review.googlesource.com/c/go/+/198319
Run-TryBot: Bryan C. Mills <bcmills@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Jay Conrod <jayconrod@google.com>
5 years agocmd/go: document multiple conditions in TestScript
Jay Conrod [Wed, 9 Oct 2019 15:10:48 +0000 (11:10 -0400)]
cmd/go: document multiple conditions in TestScript

This functionality already exists but was undocumented. Related to
comments in CL 198797.

Change-Id: Icce40bd7c362423e6ed9c20673ce3de1311e5fd5
Reviewed-on: https://go-review.googlesource.com/c/go/+/200040
Run-TryBot: Jay Conrod <jayconrod@google.com>
Reviewed-by: Bryan C. Mills <bcmills@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

5 years agotest: new testcase for gccgo compiler problem
Than McIntosh [Mon, 30 Sep 2019 18:37:05 +0000 (14:37 -0400)]
test: new testcase for gccgo compiler problem

Test case with code that caused a gccgo error while emitting export
data for an inlinable function.

Updates #34577.

Change-Id: I28b598c4c893c77f4a76bb4f2d27e5b42f702992
Reviewed-on: https://go-review.googlesource.com/c/go/+/198057
Run-TryBot: Than McIntosh <thanm@google.com>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

5 years ago[dev.link] cmd/link: load full symbol contents after deadcode pass
Cherry Zhang [Tue, 24 Sep 2019 21:31:12 +0000 (17:31 -0400)]
[dev.link] cmd/link: load full symbol contents after deadcode pass

If the new object file format is used, now we load full symbol
contents after the deadcode pass, for reachable symbols only.
We still load some informations early, like relocations and the
contents of type symbols, which are used in the deadcode pass.
If we rewrite deadcode to use index directly, we could delay more
of the loading (to sym.Symbol), and perhaps delay the creation of
sym.Symbol.

TODO: internal linking with host objects doesn't work yet.

Change-Id: I7d4880e8f150e8709ffac277e62191623440e4cf
Reviewed-on: https://go-review.googlesource.com/c/go/+/197258
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Than McIntosh <thanm@google.com>
5 years agocmd/go: fail if a test binary exits with no output
Daniel Martí [Mon, 1 Jul 2019 16:30:01 +0000 (18:30 +0200)]
cmd/go: fail if a test binary exits with no output

For example, if a test calls os.Exit(0), that could trick a 'go test'
run into not running some of the other tests, and thinking that they all
succeeded. This can easily go unnoticed and cause developers headaches.

Add a simple sanity check as part of 'go test': if the test binary
succeeds and doesn't print anything, we should error, as something
clearly went very wrong.

This is done by inspecting each of the stdout writes from the spawned
process, since we don't want to read the entirety of the output into a
buffer. We need to introduce a "buffered" bool var, as there's now an
io.Writer layer between cmd.Stdout and &buf.

A few TestMain funcs in the standard library needed fixing, as they
returned without printing anything as a means to skip testing the entire
package. For that purpose add testenv.MainMust, which prints a warning
and prints SKIP, similar to when -run matches no tests.

Finally, add tests for both os.Exit(0) and os.Exit(1), both as part of
TestMain and as part of a single test, and test that the various stdout
modes still do the right thing.

Fixes #29062.

Change-Id: Ic6f8ef3387dfc64e4cd3e8f903d7ca5f5f38d397
Reviewed-on: https://go-review.googlesource.com/c/go/+/184457
Run-TryBot: Daniel Martí <mvdan@mvdan.cc>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Bryan C. Mills <bcmills@google.com>
5 years agocmd/asm: add s390x branch-on-count instructions
Michael Munday [Thu, 15 Aug 2019 20:53:37 +0000 (21:53 +0100)]
cmd/asm: add s390x branch-on-count instructions

The branch-on-count instructions on s390x decrement the input
register and then compare its value to 0. If not equal the branch
is taken.

These instructions are useful for implementing loops with a set
number of iterations (which might be in a register).

For example, this for loop:

for i := 0; i < n; i++ {
... // i is not used or modified in the loop
}

Could be implemented using this assembly:

MOVD  Rn, Ri
loop:
...
BRCTG Ri, loop

Note that i will count down from n in the assembly whereas in the
original for loop it counted up to n which is why we can't use i
in the loop.

These instructions will only be used in hand-written codegen and
assembly for now since SSA blocks cannot currently modify values.
We could look into this in the future though.

Change-Id: Iaab93b8aa2699513b825439b8ea20d8fe2ea1ee6
Reviewed-on: https://go-review.googlesource.com/c/go/+/199977
Run-TryBot: Michael Munday <mike.munday@ibm.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
5 years agocmd/asm: fix element size encoding for VSUMQ instruction on s390x
Michael Munday [Mon, 7 Oct 2019 21:37:44 +0000 (14:37 -0700)]
cmd/asm: fix element size encoding for VSUMQ instruction on s390x

The element size for VSUMQF and VSUMQG was off by one. Fix this
and add tests for VSUM* instruction encodings.

Change-Id: I6de2dabb383e5bc6f85eef1e0f106ba949c9030b
Reviewed-on: https://go-review.googlesource.com/c/go/+/199978
Run-TryBot: Michael Munday <mike.munday@ibm.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
5 years agocmd/go: suppress more errors in package-to-module loading
Bryan C. Mills [Tue, 8 Oct 2019 18:23:43 +0000 (14:23 -0400)]
cmd/go: suppress more errors in package-to-module loading

In CL 197059, I suppressed errors if the target package was already found.
However, that does not cover the case of passing a '/v2' module path to
'go get' when the module does not contain a package at its root.

This CL is a minimal fix for that case, intended to be backportable to 1.13.

(Longer term, I intend to rework the version-validation check to treat
all mismatched paths as ErrNotExist.)

Fixes #34746
Updates #34383

Change-Id: Ia963c2ea00fae424812b8f46a4d6c2c668252147
Reviewed-on: https://go-review.googlesource.com/c/go/+/199839
Run-TryBot: Bryan C. Mills <bcmills@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Jay Conrod <jayconrod@google.com>
5 years agogo/build: allow ~ and ^ in #cgo directives for sr.ht import paths
Elias Naur [Wed, 9 Oct 2019 09:13:42 +0000 (11:13 +0200)]
go/build: allow ~ and ^ in #cgo directives for sr.ht import paths

Fixes #32260

Change-Id: Ib44ee33b8143d523875cf5a2bc5e36bf082801a4
Reviewed-on: https://go-review.googlesource.com/c/go/+/199918
Run-TryBot: Elias Naur <mail@eliasnaur.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Daniel Martí <mvdan@mvdan.cc>
Reviewed-by: Jay Conrod <jayconrod@google.com>
5 years agoall: remove the nacl port (part 1)
Brad Fitzpatrick [Tue, 8 Oct 2019 19:19:13 +0000 (19:19 +0000)]
all: remove the nacl port (part 1)

You were a useful port and you've served your purpose.
Thanks for all the play.

A subsequent CL will remove amd64p32 (including assembly files and
toolchain bits) and remaining bits. The amd64p32 removal will be
separated into its own CL in case we want to support the Linux x32 ABI
in the future and want our old amd64p32 support as a starting point.

Updates #30439

Change-Id: Ia3a0c7d49804adc87bf52a4dea7e3d3007f2b1cd
Reviewed-on: https://go-review.googlesource.com/c/go/+/199499
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

5 years agocmd/compile: split n.Noescape() into separate uses
Matthew Dempsky [Tue, 8 Oct 2019 21:21:17 +0000 (14:21 -0700)]
cmd/compile: split n.Noescape() into separate uses

n.Noescape() was overloaded for two uses: (1) to indicate a function
was annotated with //go:noescape, and (2) to indicate that certain
temporary allocations don't outlive the current statement.

The first use case is redundant with n.Func.Pragma&Noescape!=0, which
is the convention we use for checking other function-level pragmas.

The second use case is better served by renaming "Noescape" to
"Transient".

Passes toolstash-check.

Change-Id: I0f09d2d5767513894b7bf49da9cdabd04aa4a05e
Reviewed-on: https://go-review.googlesource.com/c/go/+/199822
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
5 years agogo/types: don't skip defined types when reporting cycles
Robert Griesemer [Tue, 8 Oct 2019 22:45:14 +0000 (15:45 -0700)]
go/types: don't skip defined types when reporting cycles

The newly introduced "late-stage" cycle detection for types
(https://golang.org/cl/196338/) "skips" named types on the
RHS of a type declaration when reporting a cycle. For instance,
for:

type (
   A B
   B [10]C
   C A
)

the reported cycle is:

illegal cycle in declaration of C
       C refers to
       C

because the underlying type of C resolves to [10]C (note that
cmd/compile does the same but simply says invalid recursive
type C).

This CL introduces the Named.orig field which always refers
to the RHS type in a type definition (and is never changed).
By using Named.orig rather than Named.underlying for the type
validity check, the cycle as written in the source code is
reported:

  illegal cycle in declaration of A
         A refers to
         B refers to
         C refers to
         A

Fixes #34771.

Change-Id: I41e260ceb3f9a15da87ffae6a3921bd8280e2ac4
Reviewed-on: https://go-review.googlesource.com/c/go/+/199937
Run-TryBot: Robert Griesemer <gri@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
5 years agonet/textproto: do not allow multi-line header field names
Katie Hockman [Tue, 8 Oct 2019 18:19:34 +0000 (14:19 -0400)]
net/textproto: do not allow multi-line header field names

Fixes #34702

Change-Id: I98320d54726e646a310e583283ddab676c3503e7
Reviewed-on: https://go-review.googlesource.com/c/go/+/199838
Run-TryBot: Katie Hockman <katie@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
5 years agocmd/go: respect -mod flag in fmt
Dan Peterson [Tue, 1 Oct 2019 16:53:42 +0000 (13:53 -0300)]
cmd/go: respect -mod flag in fmt

Fixes #27841

Change-Id: Ifcfd938aff8680cf7b44dfc09fde01d6105345a0
Reviewed-on: https://go-review.googlesource.com/c/go/+/198257
Run-TryBot: Dan Peterson <dpiddy@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Bryan C. Mills <bcmills@google.com>
5 years ago[dev.link] cmd/link: add symbols to Textp after deadcode pass
Cherry Zhang [Tue, 24 Sep 2019 22:12:55 +0000 (18:12 -0400)]
[dev.link] cmd/link: add symbols to Textp after deadcode pass

Currently we add all text symbols to ctxt.Textp at load time,
then the deadcode pass filters out unreachable ones. This CL
delays adding symbols to ctxt.Textp to the end of the deadcode
pass, where we only add reachable ones.

Change-Id: Ie83b2958f915c5aaa004b8c5ed1f1bc275f4d1db
Reviewed-on: https://go-review.googlesource.com/c/go/+/197257
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Than McIntosh <thanm@google.com>
5 years ago[dev.link] cmd/link: remove runtime.gcbits special case
Cherry Zhang [Mon, 7 Oct 2019 16:44:49 +0000 (12:44 -0400)]
[dev.link] cmd/link: remove runtime.gcbits special case

Follow-up of Than's comment on CL 196032. For runtime.gcbits
symbols, the compiler already sets the "local" attribute
(cmd/compile/internal/gc/reflect.go:dgcptrmask). No need of the
special handling in the linker.

Change-Id: I7c53a818ed0375083f04655f3565c1c9c45f2401
Reviewed-on: https://go-review.googlesource.com/c/go/+/199642
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Than McIntosh <thanm@google.com>
5 years agonet/http: use err as error var in server.Serve
Davor Kapsa [Tue, 8 Oct 2019 15:26:19 +0000 (17:26 +0200)]
net/http: use err as error var in server.Serve

Change-Id: Icbf97d640fb26eed646f9e85c1f1c94b1469ca4f
Reviewed-on: https://go-review.googlesource.com/c/go/+/199778
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

5 years agogo/types: remove objSet type in favor of explicit map type (cleanup)
Robert Griesemer [Tue, 1 Oct 2019 19:34:06 +0000 (12:34 -0700)]
go/types: remove objSet type in favor of explicit map type (cleanup)

Avoid confusion between (now gone) objSet and objset types.
Also: rename visited -> seen in initorder.go.

No functional changes.

Change-Id: Ib0aa25e006eee55a79a739194d0d26190354a9f2
Reviewed-on: https://go-review.googlesource.com/c/go/+/198044
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
5 years agogo/types: simplify some code and remove TODOs (cleanup)
Robert Griesemer [Tue, 1 Oct 2019 16:59:53 +0000 (09:59 -0700)]
go/types: simplify some code and remove TODOs (cleanup)

- remove Checker.cycle in favor of using a "seen" map
- rename Checker.typeCycle -> Checker.cycle
- remove TODO in api.go since the API is frozen

Change-Id: I182a8215978dad54e9c6e79c21c5ec88ec802349
Reviewed-on: https://go-review.googlesource.com/c/go/+/198042
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
5 years agogo/types: fix cycle detection
Robert Griesemer [Wed, 18 Sep 2019 23:49:20 +0000 (16:49 -0700)]
go/types: fix cycle detection

For Go 1.13, we rewrote the go/types cycle detection scheme. Unfortunately,
it was a bit too clever and introduced a bug (#34333). Here's an example:

type A struct {
f1 *B
f2 B
}

type B A

When type-checking this code, the first cycle A->*B->B->A (via field f1)
is ok because there's a pointer indirection. Though in the process B is
considered "type-checked" (and painted/marked from "grey" to black").
When type-checking f2, since B is already completely set up, go/types
doesn't complain about the invalid cycle A->B->A (via field f2) anymore.
On the other hand, with the fields f1, f2 swapped:

type A struct {
f2 B
f1 *B
}

go/types reports an error because the cycle A->B->A is type-checked first.
In general, we cannot know the "right" order in which types need to be
type-checked.

This CL fixes the issue as follows:

1) The global object path cycle detection does not take (pointer, function,
   reference type) indirections into account anymore for cycle detection.
   That mechanism was incorrect to start with and the primary cause for this
   issue. As a consequence we don't need Checker.indirectType and indir anymore.

2) After processing type declarations, Checker.validType is called to
   verify that a type doesn't expand indefinitively. This corresponds
   essentially to cmd/compile's dowidth computation (without size computation).

3) Cycles involving only defined types (e.g.: type (A B; B C; C A))
   require separate attention as those must now be detected when resolving
   "forward chains" of type declarations. Checker.underlying was changed
   to detect these cycles.

All three cycle detection mechanism use an object path ([]Object) to
report cycles. The cycle error reporting mechanism is now factored out
into Checker.cycleError and used by all three mechanisms. It also makes
an attempt to report the cycle starting with the "first" (earliest in the
source) object.

Fixes #34333.

Change-Id: I2c6446445e47344cc2cd034d3c74b1c345b8c1e6
Reviewed-on: https://go-review.googlesource.com/c/go/+/196338
Run-TryBot: Robert Griesemer <gri@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
5 years agocmd/compile: use vsx loads and stores for LoweredMove, LoweredZero on ppc64x
Lynn Boger [Fri, 4 Oct 2019 20:03:36 +0000 (16:03 -0400)]
cmd/compile: use vsx loads and stores for LoweredMove, LoweredZero on ppc64x

This improves the code generated for LoweredMove and LoweredZero by
using LXVD2X and STXVD2X to move 16 bytes at a time. These instructions
are now used if the size to be moved or zeroed is >= 64. These same
instructions have already been used in the asm implementations for
memmove and memclr.

Some examples where this shows an improvement on power8:

MakeSlice/Byte                                  27.3ns ± 1%     25.2ns ± 0%    -7.69%
MakeSlice/Int16                                 40.2ns ± 0%     35.2ns ± 0%   -12.39%
MakeSlice/Int                                   94.9ns ± 1%     77.9ns ± 0%   -17.92%
MakeSlice/Ptr                                    129ns ± 1%      103ns ± 0%   -20.16%
MakeSlice/Struct/24                              176ns ± 1%      131ns ± 0%   -25.67%
MakeSlice/Struct/32                              200ns ± 1%      142ns ± 0%   -29.09%
MakeSlice/Struct/40                              220ns ± 2%      156ns ± 0%   -28.82%
GrowSlice/Byte                                  81.4ns ± 0%     73.4ns ± 0%    -9.88%
GrowSlice/Int16                                  118ns ± 1%       98ns ± 0%   -17.03%
GrowSlice/Int                                    178ns ± 1%      134ns ± 1%   -24.65%
GrowSlice/Ptr                                    249ns ± 4%      212ns ± 0%   -14.94%
GrowSlice/Struct/24                              294ns ± 5%      215ns ± 0%   -27.08%
GrowSlice/Struct/32                              315ns ± 1%      248ns ± 0%   -21.49%
GrowSlice/Struct/40                              382ns ± 4%      289ns ± 1%   -24.38%
ExtendSlice/IntSlice                             109ns ± 1%       90ns ± 1%   -17.51%
ExtendSlice/PointerSlice                         142ns ± 2%      118ns ± 0%   -16.75%
ExtendSlice/NoGrow                              6.02ns ± 0%     5.88ns ± 0%    -2.33%
Append                                          27.2ns ± 0%     27.6ns ± 0%    +1.38%
AppendGrowByte                                  4.20ms ± 3%     2.60ms ± 0%   -38.18%
AppendGrowString                                 134ms ± 3%      102ms ± 2%   -23.62%
AppendSlice/1Bytes                              5.65ns ± 0%     5.67ns ± 0%    +0.35%
AppendSlice/4Bytes                              6.40ns ± 0%     6.55ns ± 0%    +2.34%
AppendSlice/7Bytes                              8.74ns ± 0%     8.84ns ± 0%    +1.14%
AppendSlice/8Bytes                              5.68ns ± 0%     5.70ns ± 0%    +0.40%
AppendSlice/15Bytes                             9.31ns ± 0%     9.39ns ± 0%    +0.86%
AppendSlice/16Bytes                             14.0ns ± 0%      5.8ns ± 0%   -58.32%
AppendSlice/32Bytes                             5.72ns ± 0%     5.68ns ± 0%    -0.66%
AppendSliceLarge/1024Bytes                       918ns ± 8%      615ns ± 1%   -33.00%
AppendSliceLarge/4096Bytes                      3.25µs ± 1%     1.92µs ± 1%   -40.84%
AppendSliceLarge/16384Bytes                     8.70µs ± 2%     4.69µs ± 0%   -46.08%
AppendSliceLarge/65536Bytes                     18.1µs ± 3%      7.9µs ± 0%   -56.30%
AppendSliceLarge/262144Bytes                    69.8µs ± 2%     25.9µs ± 0%   -62.91%
AppendSliceLarge/1048576Bytes                    258µs ± 1%       93µs ± 0%   -63.96%

Change-Id: I21625dbe231a2029ddb9f7d73f5a6417b35c1e49
Reviewed-on: https://go-review.googlesource.com/c/go/+/199639
Run-TryBot: Lynn Boger <laboger@linux.vnet.ibm.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
5 years agocmd/internal/obj/arm64: add error checking for system registers.
diaxu01 [Wed, 21 Aug 2019 09:31:35 +0000 (09:31 +0000)]
cmd/internal/obj/arm64: add error checking for system registers.

This CL adds system register error checking test cases. There're two kinds of
error test cases:

1. illegal combination.
MRS should be used in this way: MRS <system register>, <general register>.
MSR should be used in this way: MSR <general register>, <system register>.
Error usage examples:
MRS     R8, VTCR_EL2    // ERROR "illegal combination"
MSR     VTCR_EL2, R8    // ERROR "illegal combination"

2. illegal read or write access.
Error usage examples:
MSR     R7, MIDR_EL1    // ERROR "expected writable system register or pstate"
MRS     OSLAR_EL1, R3   // ERROR "expected readable system register"

This CL reads system registers readable and writeable property to check whether
they're used with legal read or write access. This property is named AccessFlags
in sysRegEnc.go, and it is automatically generated by modifing the system register
generator.

Change-Id: Ic83d5f372de38d1ecd0df1ca56b354ee157f16b4
Reviewed-on: https://go-review.googlesource.com/c/go/+/194917
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

5 years agocmd/compile: regenerate known formats in fmtmap_test.go
Tobias Klauser [Tue, 8 Oct 2019 11:17:02 +0000 (13:17 +0200)]
cmd/compile: regenerate known formats in fmtmap_test.go

This fixes TestFormats after CL 198037

Change-Id: I3fb7d667f7c2a1fd88a320482310d33b75e068c4
Reviewed-on: https://go-review.googlesource.com/c/go/+/199777
Run-TryBot: Tobias Klauser <tobias.klauser@gmail.com>
Reviewed-by: Michael Munday <mike.munday@ibm.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

5 years agosyscall: don't use deprecated syscalls on linux/arm64
Tobias Klauser [Sat, 5 Oct 2019 11:44:36 +0000 (13:44 +0200)]
syscall: don't use deprecated syscalls on linux/arm64

Reimplement syscall wrappers for linux/arm64 in terms of supported
syscalls (or in case of Ustat make it return ENOSYS) and remove the
manually added SYS_* consts for the deprecated syscalls. Adapted from
golang.org/x/sys/unix where this is already done since CL 119655.

Change-Id: I94ab48a4645924df3822497d0575f1a1573d509f
Reviewed-on: https://go-review.googlesource.com/c/go/+/199140
Run-TryBot: Tobias Klauser <tobias.klauser@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
5 years agocmd/compile: add SSA rules for s390x compare-and-branch instructions
Michael Munday [Tue, 17 Sep 2019 14:29:31 +0000 (07:29 -0700)]
cmd/compile: add SSA rules for s390x compare-and-branch instructions

This commit adds SSA rules for the s390x combined compare-and-branch
instructions. These have a shorter encoding than separate compare
and branch instructions and they also don't clobber the condition
code (a.k.a. flag register) reducing pressure on the flag allocator.

I have deleted the 'loop_test.go' file and replaced it with a new
codegen test which performs a wider range of checks.

Object sizes from compilebench:

name                      old object-bytes  new object-bytes  delta
Template                        562kB ± 0%        561kB ± 0%   -0.28%  (p=0.000 n=10+10)
Unicode                         217kB ± 0%        217kB ± 0%   -0.17%  (p=0.000 n=10+10)
GoTypes                        2.03MB ± 0%       2.02MB ± 0%   -0.59%  (p=0.000 n=10+10)
Compiler                       8.16MB ± 0%       8.11MB ± 0%   -0.62%  (p=0.000 n=10+10)
SSA                            27.4MB ± 0%       27.0MB ± 0%   -1.45%  (p=0.000 n=10+10)
Flate                           356kB ± 0%        356kB ± 0%   -0.12%  (p=0.000 n=10+10)
GoParser                        438kB ± 0%        436kB ± 0%   -0.51%  (p=0.000 n=10+10)
Reflect                        1.37MB ± 0%       1.37MB ± 0%   -0.42%  (p=0.000 n=10+10)
Tar                             485kB ± 0%        483kB ± 0%   -0.39%  (p=0.000 n=10+10)
XML                             630kB ± 0%        621kB ± 0%   -1.45%  (p=0.000 n=10+10)
[Geo mean]                     1.14MB            1.13MB        -0.60%

name                      old text-bytes    new text-bytes    delta
HelloSize                       763kB ± 0%        754kB ± 0%   -1.30%  (p=0.000 n=10+10)
CmdGoSize                      10.7MB ± 0%       10.6MB ± 0%   -0.91%  (p=0.000 n=10+10)
[Geo mean]                     2.86MB            2.82MB        -1.10%

Change-Id: Ibca55d9c0aa1254aee69433731ab5d26a43a7c18
Reviewed-on: https://go-review.googlesource.com/c/go/+/198037
Run-TryBot: Michael Munday <mike.munday@ibm.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
5 years agocmd/compile: don't use statictmps for small object in slice literal
Cuong Manh Le [Sat, 28 Sep 2019 16:30:08 +0000 (23:30 +0700)]
cmd/compile: don't use statictmps for small object in slice literal

Fixes #21561

Change-Id: I89c59752060dd9570d17d73acbbaceaefce5d8ce
Reviewed-on: https://go-review.googlesource.com/c/go/+/197560
Run-TryBot: Cuong Manh Le <cuong.manhle.vn@gmail.com>
Run-TryBot: Keith Randall <khr@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

5 years agosyscall: on wasm, do not panic if "process" global is not defined
Richard Musiol [Mon, 7 Oct 2019 22:58:26 +0000 (00:58 +0200)]
syscall: on wasm, do not panic if "process" global is not defined

When running wasm in the browser, the "process" global is not defined.
This causes functions like os.Getpid() to panic, which is unusual.
For example on Windows os.Getpid() returns -1 and does not panic.

This change adds a dummy polyfill for "process" which returns -1 or an
error. It also extends the polyfill for "fs".

Fixes #34627
Replaces CL 199357

Change-Id: Ifeb12fe7e152c517848933a9ab5f6f749896dcef
Reviewed-on: https://go-review.googlesource.com/c/go/+/199698
Run-TryBot: Richard Musiol <neelance@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
5 years agocmd/compile: suppress statement marks on interior of switch tree
David Chase [Thu, 3 Oct 2019 16:19:42 +0000 (12:19 -0400)]
cmd/compile: suppress statement marks on interior of switch tree

The lines on nodes within the IF-tree generated for switch
statements looks like control flow so the lines get marked
as statement boundaries.  Except for the first/root comparison,
explicitly disable the marks.

Change-Id: I64b966ed8e427cdc6b816ff6b6a2eb754346edc7
Reviewed-on: https://go-review.googlesource.com/c/go/+/198738
Run-TryBot: David Chase <drchase@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Jeremy Faller <jeremy@golang.org>
5 years agocmd/compile: remove useless block-indirection in type switch
Matthew Dempsky [Mon, 7 Oct 2019 20:33:39 +0000 (13:33 -0700)]
cmd/compile: remove useless block-indirection in type switch

Previously, when emitting type switches without an explicit "case nil"
clause, we would emit:

    if x == nil { goto Lnil }
    ...
    Lnil: goto Ldefault

But we can instead just emit:

    if x == nil { goto Ldefault }

Doesn't pass toolstash-check; seems like it causes some harmless
instruction scheduling changes.

Change-Id: Ie233dda26756911e93a08b3db40407ba38694c62
Reviewed-on: https://go-review.googlesource.com/c/go/+/199644
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
Reviewed-by: Keith Randall <khr@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

5 years agoruntime: speed up receive on empty closed channel
Ben Schwartz [Mon, 10 Jun 2019 16:29:23 +0000 (12:29 -0400)]
runtime: speed up receive on empty closed channel

Currently, nonblocking receive on an open channel is about
700 times faster than nonblocking receive on a closed channel.
This change makes closed channels equally fast.

Fixes #32529

relevant benchstat output:
name                       old time/op    new time/op    delta
MakeChan/Byte-40            140ns ± 4%     137ns ± 7%   -2.38%  (p=0.023 n=17+19)
MakeChan/Int-40             174ns ± 5%     173ns ± 6%     ~     (p=0.437 n=18+19)
MakeChan/Ptr-40             315ns ±15%     301ns ±15%     ~     (p=0.051 n=20+20)
MakeChan/Struct/0-40        123ns ± 8%      99ns ±11%  -19.18%  (p=0.000 n=20+17)
MakeChan/Struct/32-40       297ns ± 8%     241ns ±18%  -19.13%  (p=0.000 n=20+20)
MakeChan/Struct/40-40       344ns ± 5%     273ns ±23%  -20.49%  (p=0.000 n=20+20)
ChanNonblocking-40         0.32ns ± 2%    0.32ns ± 2%   -1.25%  (p=0.000 n=19+18)
SelectUncontended-40       5.72ns ± 1%    5.71ns ± 2%     ~     (p=0.326 n=19+19)
SelectSyncContended-40     10.9µs ±10%    10.6µs ± 3%   -2.77%  (p=0.009 n=20+16)
SelectAsyncContended-40    1.00µs ± 0%    1.10µs ± 0%  +10.75%  (p=0.000 n=18+19)
SelectNonblock-40          1.22ns ± 2%    1.21ns ± 4%     ~     (p=0.141 n=18+19)
ChanUncontended-40          240ns ± 4%     233ns ± 4%   -2.82%  (p=0.000 n=20+20)
ChanContended-40           86.7µs ± 0%    82.7µs ± 0%   -4.64%  (p=0.000 n=20+19)
ChanSync-40                 294ns ± 7%     284ns ± 9%   -3.44%  (p=0.006 n=20+20)
ChanSyncWork-40            38.4µs ±19%    34.0µs ± 4%  -11.33%  (p=0.000 n=20+18)
ChanProdCons0-40           1.50µs ± 1%    1.63µs ± 0%   +8.53%  (p=0.000 n=19+19)
ChanProdCons10-40          1.17µs ± 0%    1.18µs ± 1%   +0.44%  (p=0.000 n=19+20)
ChanProdCons100-40          985ns ± 0%     959ns ± 1%   -2.64%  (p=0.000 n=20+20)
ChanProdConsWork0-40       1.50µs ± 0%    1.60µs ± 2%   +6.54%  (p=0.000 n=18+20)
ChanProdConsWork10-40      1.26µs ± 0%    1.26µs ± 2%   +0.40%  (p=0.015 n=20+19)
ChanProdConsWork100-40     1.27µs ± 0%    1.22µs ± 0%   -4.15%  (p=0.000 n=20+19)
SelectProdCons-40          1.50µs ± 1%    1.53µs ± 1%   +1.95%  (p=0.000 n=20+20)
ChanCreation-40            82.1ns ± 5%    81.6ns ± 7%     ~     (p=0.483 n=19+19)
ChanSem-40                  877ns ± 0%     719ns ± 0%  -17.98%  (p=0.000 n=18+19)
ChanPopular-40             1.75ms ± 2%    1.78ms ± 3%   +1.76%  (p=0.002 n=20+19)
ChanClosed-40               215ns ± 1%       0ns ± 6%  -99.82%  (p=0.000 n=20+18)

Change-Id: I6d5ca4f1530cc9e1a9f3ef553bbda3504a036448
Reviewed-on: https://go-review.googlesource.com/c/go/+/181543
Run-TryBot: Keith Randall <khr@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
5 years agocmd/compile: reimplement parameter leak encoding
Matthew Dempsky [Fri, 27 Sep 2019 00:21:50 +0000 (17:21 -0700)]
cmd/compile: reimplement parameter leak encoding

Currently, escape analysis is able to record at most one dereference
when a parameter leaks to the heap; that is, at call sites, it can't
distinguish between any of these three functions:

    func x1(p ****int) { sink = *p }
    func x2(p ****int) { sink = **p }
    func x3(p ****int) { sink = ***p }

Similarly, it's limited to recording parameter leaks to only the first
4 parameters, and only up to 6 dereferences.

All of these limitations are due to the awkward encoding scheme used
at the moment.

This CL replaces the encoding scheme with a simple [8]uint8 array,
which can handle up to the first 7 parameters, and up to 254
dereferences, which ought to be enough for anyone. And if not, it's
much more easily increased.

Shrinks export data size geometric mean for Kubernetes by 0.07%.

Fixes #33981.

Change-Id: I10a94b9accac9a0c91490e0d6d458316f5ca1e13
Reviewed-on: https://go-review.googlesource.com/c/go/+/197680
Reviewed-by: Cherry Zhang <cherryyz@google.com>
5 years agocmd/compile: introduce EscLeaks abstraction
Matthew Dempsky [Thu, 26 Sep 2019 22:55:58 +0000 (15:55 -0700)]
cmd/compile: introduce EscLeaks abstraction

This CL better abstracts away the parameter leak info that was
directly encoded into the uint16 value. Followup CL will rewrite the
implementation.

Passes toolstash-check.

Updates #33981.

Change-Id: I27f81d26f5dd2d85f5b0e5250ca529819a1f11c2
Reviewed-on: https://go-review.googlesource.com/c/go/+/197679
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
5 years agoruntime: do not omit stack trace of goroutine that handles async events
Richard Musiol [Sun, 6 Oct 2019 20:39:08 +0000 (22:39 +0200)]
runtime: do not omit stack trace of goroutine that handles async events

On wasm there is a special goroutine that handles asynchronous events.
Blocking this goroutine often causes a deadlock. However, the stack
trace of this goroutine was omitted when printing the deadlock error.

This change adds an exception so the goroutine is not considered as
an internal system goroutine and the stack trace gets printed, which
helps with debugging the deadlock.

Updates #32764

Change-Id: Icc8f5ba3ca5a485d557b7bdd76bf2f1ffb92eb3e
Reviewed-on: https://go-review.googlesource.com/c/go/+/199537
Run-TryBot: Richard Musiol <neelance@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
5 years agocmd/link: produce valid binaries with large data section on wasm
Richard Musiol [Sat, 5 Oct 2019 22:49:52 +0000 (00:49 +0200)]
cmd/link: produce valid binaries with large data section on wasm

CL 170950 had a regression that makes the compiler produce
an invalid wasm binary if the data section is too large.
Loading such a binary gives the following error:
"LinkError: WebAssembly.instantiate(): data segment is out of bounds"

This change fixes the issue by ensuring that the minimum size of the
linear memory is larger than the end of the data section.

Fixes #34395.

Change-Id: I0c8629de7ffd0d85895ad31bf8c9d45fef197a57
Reviewed-on: https://go-review.googlesource.com/c/go/+/199358
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

5 years agocmd/compile: improve write barrier removal
Keith Randall [Mon, 7 Oct 2019 00:38:39 +0000 (17:38 -0700)]
cmd/compile: improve write barrier removal

We're allowed to remove a write barrier when both the old
value in memory and the new value we're writing are not heap pointers.

Improve both those checks a little bit.

A pointer is known to not be a heap pointer if it is read from
read-only memory. This sometimes happens for loads of pointers
from string constants in read-only memory.

Do a better job of tracking which parts of memory are known to be
zero.  Before we just kept track of a range of offsets in the most
recently allocated object. For code that initializes the new object's
fields in a nonstandard order, that tracking is imprecise. Instead,
keep a bit map of the first 64 words of that object, so we can track
precisely what we know to be zeroed.

The new scheme is only precise up to the first 512 bytes of the object.
After that, we'll use write barriers unnecessarily. Hopefully most
initializers of large objects will use typedmemmove, which does only one
write barrier check for the whole initialization.

Fixes #34723
Update #21561

Change-Id: Idf6e1b7d525042fb67961302d4fc6f941393cac8
Reviewed-on: https://go-review.googlesource.com/c/go/+/199558
Run-TryBot: Keith Randall <khr@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
5 years agointernal/reflectlite: add missing copyright header
Cuong Manh Le [Mon, 7 Oct 2019 15:04:06 +0000 (22:04 +0700)]
internal/reflectlite: add missing copyright header

Change-Id: Id43870de6608ef2e8c0ebef82fd710b2c3061e66
Reviewed-on: https://go-review.googlesource.com/c/go/+/199599
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Reviewed-by: Tobias Klauser <tobias.klauser@gmail.com>
5 years agocmd/compile: reuse dead register before reusing register holding constant
Keith Randall [Mon, 7 Oct 2019 06:03:28 +0000 (23:03 -0700)]
cmd/compile: reuse dead register before reusing register holding constant

For commuting ops, check whether the second argument is dead before
checking if the first argument is rematerializeable. Reusing the register
holding a dead value is always best.

Fixes #33580

Change-Id: I7372cfc03d514e6774d2d9cc727a3e6bf6ce2657
Reviewed-on: https://go-review.googlesource.com/c/go/+/199559
Run-TryBot: Keith Randall <khr@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: David Chase <drchase@google.com>
5 years agobytes: add endian base compare test
Meng Zhuo [Mon, 7 Oct 2019 03:30:01 +0000 (11:30 +0800)]
bytes: add endian base compare test

The current bytes test suit didn't come with endian based test
which causing #34549 can passed the try-bot.
This test will failed when little endian architecture simply using
load and compare uint.

Update #34549

Change-Id: I0973c2cd505ce21c2bed1deeb7d526f1e872118d
Reviewed-on: https://go-review.googlesource.com/c/go/+/198358
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

5 years agointernal/reflectlite: add type mirror with reflect test
Cuong Manh Le [Sat, 5 Oct 2019 10:46:09 +0000 (17:46 +0700)]
internal/reflectlite: add type mirror with reflect test

Add test to check that struct type in reflectlite is mirror of reflect.
Note that the test does not check the field types, only check for number
of fields and field name are the same.

Updates #34486

Change-Id: Id5f9b26d35faec97863dd1fe7e5eab37d4913181
Reviewed-on: https://go-review.googlesource.com/c/go/+/199280
Run-TryBot: Cuong Manh Le <cuong.manhle.vn@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
5 years agocmd/compile: reduce amount of code generated for block rewrite rules
Michael Munday [Wed, 2 Oct 2019 10:02:41 +0000 (11:02 +0100)]
cmd/compile: reduce amount of code generated for block rewrite rules

Add a Reset method to blocks that allows us to reduce the amount of
code we generate for block rewrite rules.

Thanks to Cherry for suggesting a similar fix to this in CL 196557.

Compilebench result:

name                      old time/op       new time/op       delta
Template                        211ms ± 1%        211ms ± 1%   -0.30%  (p=0.028 n=19+20)
Unicode                        83.7ms ± 3%       83.0ms ± 2%   -0.79%  (p=0.029 n=18+19)
GoTypes                         757ms ± 1%        755ms ± 1%   -0.31%  (p=0.034 n=19+19)
Compiler                        3.51s ± 1%        3.50s ± 1%   -0.20%  (p=0.013 n=18+18)
SSA                             11.7s ± 1%        11.7s ± 1%   -0.38%  (p=0.000 n=19+19)
Flate                           131ms ± 1%        130ms ± 1%   -0.32%  (p=0.024 n=18+18)
GoParser                        162ms ± 1%        162ms ± 1%     ~     (p=0.059 n=20+18)
Reflect                         471ms ± 0%        470ms ± 0%   -0.24%  (p=0.045 n=20+17)
Tar                             187ms ± 1%        186ms ± 1%     ~     (p=0.157 n=20+20)
XML                             255ms ± 1%        255ms ± 1%     ~     (p=0.461 n=19+20)
LinkCompiler                    754ms ± 2%        755ms ± 2%     ~     (p=0.919 n=17+17)
ExternalLinkCompiler            2.82s ±16%        2.37s ±10%  -15.94%  (p=0.000 n=20+20)
LinkWithoutDebugCompiler        439ms ± 4%        442ms ± 6%     ~     (p=0.461 n=18+19)
StdCmd                          25.8s ± 2%        25.5s ± 1%   -0.95%  (p=0.000 n=20+20)

name                      old user-time/op  new user-time/op  delta
Template                        240ms ± 8%        238ms ± 7%     ~     (p=0.301 n=20+20)
Unicode                         107ms ±18%        104ms ±13%     ~     (p=0.149 n=20+20)
GoTypes                         883ms ± 3%        888ms ± 2%     ~     (p=0.211 n=20+20)
Compiler                        4.22s ± 1%        4.20s ± 1%     ~     (p=0.077 n=20+18)
SSA                             14.1s ± 1%        14.1s ± 2%     ~     (p=0.192 n=20+20)
Flate                           145ms ±10%        148ms ± 5%     ~     (p=0.126 n=20+18)
GoParser                        186ms ± 7%        186ms ± 7%     ~     (p=0.779 n=20+20)
Reflect                         538ms ± 3%        541ms ± 3%     ~     (p=0.192 n=20+20)
Tar                             218ms ± 4%        217ms ± 6%     ~     (p=0.835 n=19+20)
XML                             298ms ± 5%        298ms ± 5%     ~     (p=0.749 n=19+20)
LinkCompiler                    818ms ± 5%        825ms ± 8%     ~     (p=0.461 n=20+20)
ExternalLinkCompiler            1.55s ± 4%        1.53s ± 5%     ~     (p=0.063 n=20+18)
LinkWithoutDebugCompiler        460ms ±12%        460ms ± 7%     ~     (p=0.925 n=20+20)

name                      old object-bytes  new object-bytes  delta
Template                        554kB ± 0%        554kB ± 0%     ~     (all equal)
Unicode                         215kB ± 0%        215kB ± 0%     ~     (all equal)
GoTypes                        2.01MB ± 0%       2.01MB ± 0%     ~     (all equal)
Compiler                       7.97MB ± 0%       7.97MB ± 0%   +0.00%  (p=0.000 n=20+20)
SSA                            26.8MB ± 0%       26.9MB ± 0%   +0.27%  (p=0.000 n=20+20)
Flate                           340kB ± 0%        340kB ± 0%     ~     (all equal)
GoParser                        434kB ± 0%        434kB ± 0%     ~     (all equal)
Reflect                        1.34MB ± 0%       1.34MB ± 0%     ~     (all equal)
Tar                             480kB ± 0%        480kB ± 0%     ~     (all equal)
XML                             622kB ± 0%        622kB ± 0%     ~     (all equal)

name                      old export-bytes  new export-bytes  delta
Template                       20.4kB ± 0%       20.4kB ± 0%     ~     (all equal)
Unicode                        8.21kB ± 0%       8.21kB ± 0%     ~     (all equal)
GoTypes                        36.6kB ± 0%       36.6kB ± 0%     ~     (all equal)
Compiler                        115kB ± 0%        115kB ± 0%   +0.08%  (p=0.000 n=20+20)
SSA                             141kB ± 0%        141kB ± 0%   +0.07%  (p=0.000 n=20+20)
Flate                          5.11kB ± 0%       5.11kB ± 0%     ~     (all equal)
GoParser                       8.93kB ± 0%       8.93kB ± 0%     ~     (all equal)
Reflect                        11.8kB ± 0%       11.8kB ± 0%     ~     (all equal)
Tar                            10.9kB ± 0%       10.9kB ± 0%     ~     (all equal)
XML                            17.4kB ± 0%       17.4kB ± 0%     ~     (all equal)

name                      old text-bytes    new text-bytes    delta
HelloSize                       742kB ± 0%        742kB ± 0%     ~     (all equal)
CmdGoSize                      10.7MB ± 0%       10.7MB ± 0%     ~     (all equal)

name                      old data-bytes    new data-bytes    delta
HelloSize                      10.7kB ± 0%       10.7kB ± 0%     ~     (all equal)
CmdGoSize                       312kB ± 0%        312kB ± 0%     ~     (all equal)

name                      old bss-bytes     new bss-bytes     delta
HelloSize                       122kB ± 0%        122kB ± 0%     ~     (all equal)
CmdGoSize                       146kB ± 0%        146kB ± 0%     ~     (all equal)

name                      old exe-bytes     new exe-bytes     delta
HelloSize                      1.10MB ± 0%       1.10MB ± 0%     ~     (all equal)
CmdGoSize                      14.9MB ± 0%       14.9MB ± 0%     ~     (all equal)

Change-Id: Ic89a8e62423b3d9fd9391159e0663acf450803b5
Reviewed-on: https://go-review.googlesource.com/c/go/+/198419
Run-TryBot: Michael Munday <mike.munday@ibm.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Reviewed-by: Daniel Martí <mvdan@mvdan.cc>
5 years agoos: re-enable TestPipeThreads on darwin
Shenghou Ma [Sun, 6 Oct 2019 12:03:44 +0000 (08:03 -0400)]
os: re-enable TestPipeThreads on darwin

CL 197938 actually fixes those regression on Darwin as syscalls
are no longer labeled as always blocking and consume a thread.

Fixes #33953
Fixes #32326

Change-Id: I82c98516c23cd36f762bc5433d7b71ea8939a0ac
Reviewed-on: https://go-review.googlesource.com/c/go/+/199477
Run-TryBot: Minux Ma <minux@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Tobias Klauser <tobias.klauser@gmail.com>
5 years agounicode/utf8: add link to formal UTF-8 description.
Serhat Giydiren [Sat, 5 Oct 2019 13:12:13 +0000 (16:12 +0300)]
unicode/utf8: add link to formal UTF-8 description.

Fixes #31590

Change-Id: I7fd6dcc5c34496776439ff0295f18b5fb5cb538a
Reviewed-on: https://go-review.googlesource.com/c/go/+/199141
Reviewed-by: Emmanuel Odeke <emm.odeke@gmail.com>
5 years agointernal/reflectlite: updates reflectlite to match runtime rtype/mapType
Cuong Manh Le [Fri, 27 Sep 2019 09:38:14 +0000 (16:38 +0700)]
internal/reflectlite: updates reflectlite to match runtime rtype/mapType

CL 191198 updated runtime rtype and mapType without adopting the changes
to reflectlite, causing mismatch between them.

This CL updates those changes to reflectlite.

Fixes #34486

Change-Id: I2bb043673d997f97bb0b12c4ad471474803b2160
Reviewed-on: https://go-review.googlesource.com/c/go/+/197559
Run-TryBot: Cuong Manh Le <cuong.manhle.vn@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Bryan C. Mills <bcmills@google.com>
Reviewed-by: Keith Randall <khr@golang.org>
5 years agoruntime: add doc to remind adopting changes to reflectlite
Cuong Manh Le [Fri, 27 Sep 2019 09:42:46 +0000 (16:42 +0700)]
runtime: add doc to remind adopting changes to reflectlite

Updates #34486

Change-Id: Iec9a5d120013aaa287eccf2999b3f2b831be070e
Reviewed-on: https://go-review.googlesource.com/c/go/+/197558
Reviewed-by: Ian Lance Taylor <iant@golang.org>
5 years agocmd/go: avoid ssh known_hosts prompts on test
Daniel Martí [Sat, 5 Oct 2019 14:08:07 +0000 (15:08 +0100)]
cmd/go: avoid ssh known_hosts prompts on test

TestScripts/mod_get_svn would stop with the following prompt if the real
user didn't have vcs-test.golang.org in their known_hosts file:

The authenticity of host 'vcs-test.golang.org (35.184.38.56)' can't be established.
ECDSA key fingerprint is SHA256:[...]
Are you sure you want to continue connecting (yes/no/[fingerprint])?

This was bad because it relied on the user's real ssh known_hosts file.
Worse even, if the user didn't expert or notice the prompt, it could
hang a 'go test' run for quite a while.

Work around that by forcing svn to not use ssh at all. Other potentially
better approaches were tried, but none worked on svn 1.12.2 with openssh
8.0p1.

Fixes #33883.

Change-Id: I2f925fa892f2fa53c77d86b0034141162517ee69
Reviewed-on: https://go-review.googlesource.com/c/go/+/199142
Run-TryBot: Daniel Martí <mvdan@mvdan.cc>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>