]> Cypherpunks repositories - gostls13.git/log
gostls13.git
5 years agoos: change UserConfigDir on Darwin to ~/Library/Application Support
Daniel Martí [Fri, 7 Jun 2019 09:39:21 +0000 (10:39 +0100)]
os: change UserConfigDir on Darwin to ~/Library/Application Support

The old code used ~/Library/Preferences, which is documented by
Apple as:

This directory contains app-specific preference files. You
should not create files in this directory yourself. Instead, use
the NSUserDefaults class or CFPreferences API to get and set
preference values for your app.

It looks like we missed everything after the first sentence; it's
definitely not the right choice for files that Go programs and users
should be touching directly.

Instead, use ~/Library/Application Support, which is documented as:

Use this directory to store all app data files except those
associated with the user’s documents. For example, you might use
this directory to store app-created data files, configuration
files, templates, or other fixed or modifiable resources that
are managed by the app. An app might use this directory to store
a modifiable copy of resources contained initially in the app’s
bundle. A game might use this directory to store new levels
purchased by the user and downloaded from a server.

This seems in line with what UserConfigDir is for, so use it.

The documentation quotes above are obtained from the surprisingly long
link below:

https://developer.apple.com/library/archive/documentation/FileManagement/Conceptual/FileSystemProgrammingGuide/FileSystemOverview/FileSystemOverview.html

Fixes #32475.

Change-Id: Ic27a6c92d76a5d7a4d4b8eac5cd8472f67a533a4
Reviewed-on: https://go-review.googlesource.com/c/go/+/181177
Run-TryBot: Daniel Martí <mvdan@mvdan.cc>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Andrew Bonventre <andybons@golang.org>
5 years agocmd/link: fix deferreturn detector
Keith Randall [Fri, 7 Jun 2019 17:22:53 +0000 (13:22 -0400)]
cmd/link: fix deferreturn detector

The logic for detecting deferreturn calls is wrong.

We used to look for a relocation whose symbol is runtime.deferreturn
and has an offset of 0. But on some architectures, the relocation
offset is not zero. These include arm (the offset is 0xebfffffe) and
s390x (the offset is 6).

This ends up setting the deferreturn offset at 0, so we end up using
the entry point live map instead of the deferreturn live map in a
frame which defers and then segfaults.

Instead, use the IsDirectCall helper to find calls.

Fixes #32477
Update #6980

Change-Id: Iecb530a7cf6eabd7233be7d0731ffa78873f3a54
Reviewed-on: https://go-review.googlesource.com/c/go/+/181258
Run-TryBot: Keith Randall <khr@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
5 years agonet/http: support BaseContext & ConnContext for http2 Server
Brad Fitzpatrick [Fri, 7 Jun 2019 17:56:24 +0000 (17:56 +0000)]
net/http: support BaseContext & ConnContext for http2 Server

This is the net/http half of #32476. This supplies the method needed
by the other half in x/net/http2 in the already-submitted CL 181259,
which this CL also bundles in h2_bundle.go.

Thanks to Tom Thorogood (@tmthrgd) for the bug report and test.

Fixes #32476
Updates #30694

Change-Id: I79d2a280e486fbf75d116f6695fd3abb61278765
Reviewed-on: https://go-review.googlesource.com/c/go/+/181260
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
5 years agocmd/dist,cmd/compile: remove -allabis mode
Austin Clements [Thu, 6 Jun 2019 21:27:14 +0000 (17:27 -0400)]
cmd/dist,cmd/compile: remove -allabis mode

dist passes the -allabis flag to the compiler to avoid having to
recreate the cross-package ABI logic from cmd/go. However, we removed
that logic from cmd/go in CL 179863 and replaced it with a different
mechanism that doesn't depend on the build system. Hence, passing
-allabis in dist is no longer necessary.

This CL removes -allabis from dist and, since that was the only use of
it, removes support for it from the compiler as well.

Updates #31230.

Change-Id: Ib005db95755a7028f49c885785e72c3970aea4f9
Reviewed-on: https://go-review.googlesource.com/c/go/+/181079
Run-TryBot: Austin Clements <austin@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Reviewed-by: Jay Conrod <jayconrod@google.com>
5 years agosyscall: refer to Proc.Call documentation for LazyProc.Call on Windows
Matt Layher [Fri, 7 Jun 2019 17:07:32 +0000 (13:07 -0400)]
syscall: refer to Proc.Call documentation for LazyProc.Call on Windows

The documentation comment was duplicated for each of these methods, and
the LazyProc.Call documentation incorrectly mentioned that Call accepts
only 15 arguments, but it actually accepts 18 now.

To prevent further documentation drift, refer the reader to the
documentation for Proc.Call instead of duplicating it for LazyProc.Call.

In addition, note that LazyProc's Addr, Call, and Find methods each
trigger a procedure lookup.

Change-Id: I6756cf7601fba79d1414ff5a5d6eef900aa590e7
Reviewed-on: https://go-review.googlesource.com/c/go/+/181199
Run-TryBot: Matt Layher <mdlayher@gmail.com>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
5 years agocmd/go: fix error messages for go mod download in GOPATH mode
Constantin Konstantinidis [Sat, 1 Jun 2019 16:14:04 +0000 (18:14 +0200)]
cmd/go: fix error messages for go mod download in GOPATH mode

Checks if modules are enabled in GOPATH mode.
Error message returned when no version is provided. Relevant tests
updated. Test for GO111MODULE=off added.

Fixes #27783

Change-Id: I12cdaced5fa38a9c49c0ecfed4c479eb86ed061f
Reviewed-on: https://go-review.googlesource.com/c/go/+/179998
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/go: recommend -d when -m is used
Daniel Martí [Fri, 31 May 2019 09:46:00 +0000 (10:46 +0100)]
cmd/go: recommend -d when -m is used

In #32038, it was decided to remove get's -m, since one former use case
is removed, and the other can be done via -d, as pointed by Russ.

However, a user getting this short error might not realise that they can
switch to -d to skip building packages. Add a short mention to point
them in the right direction.

It's important to note "packages", because -m was a flag that acted on
modules, while -d acts on packages. Simply replacing -m with -d might
not be enough in some cases because of that distinction.

Change-Id: I0947b25c4223bdad3cd0e535848527da8db8a16d
Reviewed-on: https://go-review.googlesource.com/c/go/+/179361
Reviewed-by: Jay Conrod <jayconrod@google.com>
Reviewed-by: Bryan C. Mills <bcmills@google.com>
Run-TryBot: Daniel Martí <mvdan@mvdan.cc>
TryBot-Result: Gobot Gobot <gobot@golang.org>

5 years agocmd/go/internal/modload: propagate errors from Query in 'list -versions'
Bryan C. Mills [Thu, 6 Jun 2019 18:30:56 +0000 (14:30 -0400)]
cmd/go/internal/modload: propagate errors from Query in 'list -versions'

Fixes #32335

Change-Id: I1cf8645ecc5ba0866d9b3589a18bb500ea17f865
Reviewed-on: https://go-review.googlesource.com/c/go/+/181018
Run-TryBot: Bryan C. Mills <bcmills@google.com>
Reviewed-by: Jay Conrod <jayconrod@google.com>
5 years agotext/template: clarify the safety of Template.New
Daniel Martí [Tue, 16 Apr 2019 13:13:47 +0000 (22:13 +0900)]
text/template: clarify the safety of Template.New

In particular, the returned template isn't independent from the parent.
For example, it can't be parsed concurrently with other children
templates. Only methods which are explicitly safe for concurrent use,
like Execute, may be used concurrently.

Fixes #30281.

Change-Id: Idc84bf4199c035316cdb83b950fd4a8f2a71cd0c
Reviewed-on: https://go-review.googlesource.com/c/go/+/172297
Run-TryBot: Daniel Martí <mvdan@mvdan.cc>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Andrew Gerrand <adg@golang.org>
5 years agogo/internal/gccgoimporter: update for gofrontend export data changes
Ian Lance Taylor [Fri, 7 Jun 2019 00:19:11 +0000 (17:19 -0700)]
go/internal/gccgoimporter: update for gofrontend export data changes

This recognizes new features that the gofrontend has started emitting
in the export data to support cross-package inlinable functions.

This is a port of CL 180677 and 180758 from the gofrontend repo.

Change-Id: I48af6e71f9d8b04ba874ea0c204d39d1d461f8ad
Reviewed-on: https://go-review.googlesource.com/c/go/+/181118
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Than McIntosh <thanm@google.com>
5 years agocmd/vendor: go get -u golang.org/x/tools && go mod vendor
Daniel Martí [Sun, 2 Jun 2019 11:40:06 +0000 (12:40 +0100)]
cmd/vendor: go get -u golang.org/x/tools && go mod vendor

To pick up the structtag vet fix for 1.13.

Fixes #30846.

Change-Id: I5e011a7db1ffb9435793d533097d768f209c18e0
Reviewed-on: https://go-review.googlesource.com/c/go/+/179999
Run-TryBot: Daniel Martí <mvdan@mvdan.cc>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Bryan C. Mills <bcmills@google.com>
5 years agoruntime/internal/atomic: remove erroneous ABI wrappers
Austin Clements [Thu, 6 Jun 2019 21:23:19 +0000 (17:23 -0400)]
runtime/internal/atomic: remove erroneous ABI wrappers

CL 179862 introduced go:linkname directives to create ABI wrappers for
Store and Store64 on s390x, but a concurrent change (CL 180439)
replaced the Go definitions of these functions with assembly
definitions. This resulted in conflicting definitions for the ABI0
symbols, which led to a bootstrap linking failure.

Fix this by removing the now-incorrect go:linkname directives for
Store and Store64. This should fix the linux-s390x builders.

Updates #31230.

Change-Id: I8de8c03c23412fc217d428c0018cc56eb2f9996f
Reviewed-on: https://go-review.googlesource.com/c/go/+/181078
Reviewed-by: Bryan C. Mills <bcmills@google.com>
5 years agoruntime/internal/atomic: export more ABI0 wrappers
Austin Clements [Thu, 6 Jun 2019 21:08:07 +0000 (17:08 -0400)]
runtime/internal/atomic: export more ABI0 wrappers

Somehow I missed these two functions in CL 179863. This should fix the
linux-arm builders.

Updates #31230.

Change-Id: I3f8bef3fac331b505a55c0850b0fbc799b7c06c5
Reviewed-on: https://go-review.googlesource.com/c/go/+/181077
Run-TryBot: Austin Clements <austin@google.com>
Reviewed-by: Bryan C. Mills <bcmills@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

5 years agonet: remove non-cgo macOS resolver code
Russ Cox [Wed, 5 Jun 2019 23:59:58 +0000 (19:59 -0400)]
net: remove non-cgo macOS resolver code

The built-in Go resolver works significantly better.
In particular, the use of res_search does not support
CNAME or PTR queries and may not even be thread-safe.

This CL is essentially a revert of CL 166297 plus fixes,
including CL 180842.

See CL 180842 for additional notes about problems
with this approach.

Fixes #31705.

Change-Id: I0a30a0de2fbd04f6c461520fd34378c84aadf66c
Reviewed-on: https://go-review.googlesource.com/c/go/+/180843
Run-TryBot: Russ Cox <rsc@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
5 years agonet: fix non-cgo macOS resolver code
Russ Cox [Wed, 5 Jun 2019 16:35:29 +0000 (12:35 -0400)]
net: fix non-cgo macOS resolver code

This code was added in April in CL 166297, for #12524.
This CL fixes the following problems in the code:

 - The test for failure in the assembly stubs checked for
   64-bit -1 instead of 32-bit -1 to decide to fetch errno.

 - These C routines (res_init and res_search) don't set errno anyway,
   so the Go code using errno to decide success is incorrect.
   (The routines set h_errno, which is a racy global variable
   that can't safely be consulted, storing values in a different
   error space.)

 - The Go call passed res_search a non-NUL-terminated name.

 - The C res_search rejects calls asking for TypeALL as opposed to
   more specific answers like TypeA/TypeAAAA/TypeCNAME,
   breaking cgoLookupHost in all cases and cgoLookupIP
   except with IP-version-specific networks.

 - The DNS response packet was parsed twice, once with msg.Unpack
   (discarded), and once with the lower-level dnsmessage.Parser.
   The Parser loop was missing a call to p.SkipAllQuestions, with the
   result that no DNS response packet would ever parse successfully.

 - The parsing of the DNS response answers, if reached, behaved as if
   that the AResource and AAAAResource record contained textual
   IP addresses, while in fact they contain binary ones. The calls to
   parseIPv4 and parseIPv6 therefore would always returns nil,
   so that no useful result would be returned from the resolver.

With these fixes, cgoLookupIP can correctly resolve google.com
and return both the A and AAAA addresses.

Even after fixing all these things, TestGoLookupIP still fails,
because it is testing that in non-cgo builds the cgo stubs
correctly report "I can't handle the lookup", and as written the
code intentionally violates that expectation.

This CL adds new direct tests of the pseudo-cgo routines.
The direct IP address lookups succeed, but the CNAME query
causes res_search to hang, and the PTR query fails unconditionally
(a trivial C program confirms these behaviors are due to res_search itself).

Traditionally, res_search is only intended for single-threaded use.
It is unclear whether this one is safe for use from multiple goroutines.
If you run net.test under lldb, that causes syslog messages to be
printed to standard error suggesting double-free bugs:

2019-06-05 19:52:43.505246-0400 net.test[6256:6831076] dnssd_clientstub DNSServiceRefDeallocate called with invalid DNSServiceRef 0x5c000f0 FFFFFFFF DDDDDDDD
2019-06-05 19:52:43.505274-0400 net.test[6256:6831076] dnssd_clientstub DNSServiceRefDeallocate called with invalid DNSServiceRef 0x5c000f0 FFFFFFFF DDDDDDDD
2019-06-05 19:52:43.505303-0400 net.test[6256:6831076] dnssd_clientstub DNSServiceRefDeallocate called with invalid DNSServiceRef 0x5c000f0 FFFFFFFF DDDDDDDD
2019-06-05 19:52:43.505329-0400 net.test[6256:6831076] dnssd_clientstub DNSServiceRefDeallocate called with invalid DNSServiceRef 0x5c000f0 FFFFFFFF DDDDDDDD

This res_search is from libsystem_info; a normal C program would
get res_search (#defined to res_9_search) from libresolv instead.
It is unclear what the relation between the two is.
Issue #12524 was about supporting the /etc/resolver directory tree,
but only libresolv contains code for that; libsystem_info does not.
So this code probably does not enable use of /etc/resolver.

In short:

 - Before this CL, the code clearly had never run successfully.
 - The code appears not to improve upon the usual non-cgo fallback.
 - The code carries with it no tests of improved behavior.
 - The code breaks existing tests.
 - Calling res_search does not work for PTR/CNAME queries,
   so the code breaks existing behavior, even after this CL.
 - It's unclear whether res_search is safe to call from multiple threads.
 - It's unclear whether res_search is used by any other macOS programs.

Given this, it probably makes sense to delete this code rather
than rejigger the test. This CL fixes the code first, so that there
is a working copy to bring back later if we find out that it really
is necessary.

For #31705.

Change-Id: Id2e11e8ade43098b0f90dd4d16a62ca86a7a244a
Reviewed-on: https://go-review.googlesource.com/c/go/+/180842
Run-TryBot: Russ Cox <rsc@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
5 years agoruntime: document, fix libc error checks on macOS
Russ Cox [Wed, 5 Jun 2019 18:51:57 +0000 (14:51 -0400)]
runtime: document, fix libc error checks on macOS

It matters whether we are calling a function that would
return a 32-bit or 64-bit -1 on error. A few sites were wrong
and this key detail was omitted from syscall/syscallX docs.

Change-Id: I48a421b6cc4d2d2b5e58f790cc947e3cb2f98940
Reviewed-on: https://go-review.googlesource.com/c/go/+/180841
Run-TryBot: Russ Cox <rsc@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
5 years agocmd/go: remove cross-package assembly reference discovery
Austin Clements [Fri, 31 May 2019 20:40:14 +0000 (16:40 -0400)]
cmd/go: remove cross-package assembly reference discovery

This removes the special case for finding assembly references to Go
symbols in runtime and runtime/internal/atomic. These are no longer
necessary because we've now marked all symbols in these packages that
must be accessible from assembly in other packages.

Fixes #31230.

Change-Id: I70c90b70e13b922a6669f3d46c53347f98d6fc3f
Reviewed-on: https://go-review.googlesource.com/c/go/+/179863
Run-TryBot: Austin Clements <austin@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Jay Conrod <jayconrod@google.com>
5 years agoruntime: mark all Go symbols called from assembly in other packages
Austin Clements [Fri, 31 May 2019 20:38:56 +0000 (16:38 -0400)]
runtime: mark all Go symbols called from assembly in other packages

This marks all Go symbols called from assembly in other packages with
"go:linkname" directives to ensure they get ABI wrappers.

Now that we have this go:linkname convention, this also removes the
abi0Syms definition in the runtime, which was used to give morestackc
an ABI0 wrapper. Instead, we now just mark morestackc with a
go:linkname directive.

This was tested with buildall.bash in the default configuration, with
-race, and with -gcflags=all=-d=ssa/intrinsics/off. Since I couldn't
test cgo on non-Linux configurations, I manually grepped for runtime
symbols in runtime/cgo.

Updates #31230.

Change-Id: I6c8aa56be2ca6802dfa2bf159e49c411b9071bf1
Reviewed-on: https://go-review.googlesource.com/c/go/+/179862
Run-TryBot: Austin Clements <austin@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Michael Knyszek <mknyszek@google.com>
5 years agocmd/compile: make the second argument to go:linkname optional
Austin Clements [Fri, 31 May 2019 19:45:06 +0000 (15:45 -0400)]
cmd/compile: make the second argument to go:linkname optional

The //go:linkname directive can be used to make a symbol accessible to
another package (when it wouldn't normally be). Sometimes you want to
do this without actually changing the symbol's object file symbol
name; for example, in gccgo this makes unexported symbols non-static,
and in gc this provides ABI0 wrappers for Go symbols so they can be
called from assembly in other packages. Currently, this results in
stutter like

   //go:linkname entersyscall runtime.entersyscall

This CL makes the second argument to go:linkname optional for the case
where the intent is simply to expose the symbol rather than to rename
it in the object file.

Updates #31230.

Change-Id: Id06d9c4b2ec3d8e27f9b8a0d65212ab8048d734f
Reviewed-on: https://go-review.googlesource.com/c/go/+/179861
Run-TryBot: Austin Clements <austin@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: David Chase <drchase@google.com>
5 years agocmd/compile: generate ABI wrappers for //go:linkname'd symbols
Austin Clements [Fri, 31 May 2019 19:32:22 +0000 (15:32 -0400)]
cmd/compile: generate ABI wrappers for //go:linkname'd symbols

Calling a Go symbol from assembly in another package currently results
in a link failure because the Go symbol is defined as ABIInternal, but
the assembly call is from ABI0. In general this is okay because you
shouldn't do this anyway, but there are special cases where this is
necessary, especially between the runtime and packages closely tied to
the runtime in std.

Currently, we address this for runtime symbols with a hack in cmd/go
that knows to scan related packages when building the symabis file for
the runtime and runtime/internal/atomic. However, in addition to being
a messy solution in the first place, this hack causes races in cmd/go
that are difficult to work around.

We considered creating dummy references from assembly in the runtime
to these symbols, just to make sure they get ABI0 wrappers. However,
there are a fairly large number of these symbols on some platforms,
and it can vary significantly depending on build flags (e.g., race
mode), so even this solution is fairly unpalatable.

This CL addresses this by providing a way to mark symbols in Go code
that should be made available to assembly in other packages. Rather
than introduce a new pragma, we lightly expand the meaning of
"//go:linkname", since that pragma already generally indicates that
you're making the symbol available in a way it wasn't before. This
also dovetails nicely with the behavior of go:linkname in gccgo, which
makes unexported symbols available to other packages.

Follow-up CLs will make use of this and then remove the hack from
cmd/go.

Updates #31230.

Change-Id: I23060c97280626581f025c5c01fb8d24bb4c5159
Reviewed-on: https://go-review.googlesource.com/c/go/+/179860
Run-TryBot: Austin Clements <austin@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: David Chase <drchase@google.com>
5 years agocmd/compile, runtime: make atomic loads/stores sequentially consistent on s390x
Michael Munday [Tue, 4 Jun 2019 18:17:41 +0000 (19:17 +0100)]
cmd/compile, runtime: make atomic loads/stores sequentially consistent on s390x

The z/Architecture does not guarantee that a load following a store
will not be reordered with that store, unless they access the same
address. Therefore if we want to ensure the sequential consistency
of atomic loads and stores we need to perform serialization
operations after atomic stores.

We do not need to serialize in the runtime when using StoreRel[ease]
and LoadAcq[uire]. The z/Architecture already provides sufficient
ordering guarantees for these operations.

name              old time/op  new time/op  delta
AtomicLoad64-16   0.51ns ± 0%  0.51ns ± 0%     ~     (all equal)
AtomicStore64-16  0.51ns ± 0%  0.60ns ± 9%  +16.47%  (p=0.000 n=17+20)
AtomicLoad-16     0.51ns ± 0%  0.51ns ± 0%     ~     (all equal)
AtomicStore-16    0.51ns ± 0%  0.60ns ± 9%  +16.50%  (p=0.000 n=18+20)

Fixes #32428.

Change-Id: I88d19a4010c46070e4fff4b41587efe4c628d4d9
Reviewed-on: https://go-review.googlesource.com/c/go/+/180439
Run-TryBot: Michael Munday <mike.munday@ibm.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Austin Clements <austin@google.com>
5 years agocmd/compile: correct capitalization in recordFlags parameter
David Chase [Thu, 6 Jun 2019 13:56:20 +0000 (09:56 -0400)]
cmd/compile: correct capitalization in recordFlags parameter

Tool refactoring smallStacks into smallFrames helpfully
"corrected" the capitalization in a string, this undoes
the help.

This is necessary to ensure correct (re)building when the
flag is used to research stack-marking GC latency bugs.

Updates #27732.

Change-Id: Ib7c8d4a36c9e4f9612559be68bd481f9d9cc69f1
Reviewed-on: https://go-review.googlesource.com/c/go/+/180958
Run-TryBot: David Chase <drchase@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
5 years agosyscall: fix skip condition in skipUnprivilegedUserClone
Meng Zhuo [Thu, 6 Jun 2019 11:17:01 +0000 (19:17 +0800)]
syscall: fix skip condition in skipUnprivilegedUserClone

This is a follow up CL of CL 180877:
It will skip test create user namespaces under 3 conditions:

1. sysctl file is missing
2. file reads nothing
3. user don't have permission to create namespaces

Change-Id: I25f00a6b67213bf98d654972388637789978e1fe
Reviewed-on: https://go-review.googlesource.com/c/go/+/180937
Run-TryBot: Tobias Klauser <tobias.klauser@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Tobias Klauser <tobias.klauser@gmail.com>
5 years agocmd/compile: add -smallframes gc flag for GC latency diagnosis
David Chase [Wed, 5 Jun 2019 18:53:28 +0000 (14:53 -0400)]
cmd/compile: add -smallframes gc flag for GC latency diagnosis

Shrinks the size of things that can be stack allocated from
10M to 128k for declared variables and from 64k to 16k for
implicit allocations (new(T), &T{}, etc).

Usage: "go build -gcflags -smallframes hello.go"

An earlier GOEXPERIMENT version of this caused only one
problem, when a gc-should-detect-oversize-stack test no
longer had an oversized stack to detect.  The change was
converted to a flag to make it easier to access (for
diagnosing "long" GC-related single-thread pauses) and to
remove interference with the test.

Includes test to verify behavior.

Updates #27732.

Change-Id: I1255d484331e77185e07c78389a8b594041204c2
Reviewed-on: https://go-review.googlesource.com/c/go/+/180817
Run-TryBot: David Chase <drchase@google.com>
Reviewed-by: Keith Randall <khr@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

5 years agosyscall: skip test if unprivileged_userns_clone sysctl is missing
Meng Zhuo [Thu, 6 Jun 2019 10:07:13 +0000 (18:07 +0800)]
syscall: skip test if unprivileged_userns_clone sysctl is missing

The original test (CL 166460) didn't check the existence of
/proc/sys/kernel/unprivileged_userns_clone and continue the test
if the file doesn't exist.

Fixes #32459

Change-Id: Iab4938252fcaded32b61e17edf68f966c2565582
Reviewed-on: https://go-review.googlesource.com/c/go/+/180877
Run-TryBot: Tobias Klauser <tobias.klauser@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Tobias Klauser <tobias.klauser@gmail.com>
5 years agoruntime: fix non-tab indentation in lookup_darwin_*.s
Russ Cox [Wed, 5 Jun 2019 17:59:00 +0000 (13:59 -0400)]
runtime: fix non-tab indentation in lookup_darwin_*.s

Change-Id: Ie00494f098bd2bce9bfd1b18dbf9543cf46faad6
Reviewed-on: https://go-review.googlesource.com/c/go/+/180840
Run-TryBot: Russ Cox <rsc@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
5 years agoruntime: fix scattered non-tab indentation in assembly
Russ Cox [Wed, 5 Jun 2019 17:57:47 +0000 (13:57 -0400)]
runtime: fix scattered non-tab indentation in assembly

Change-Id: I6940a4c747f2da871263afa6a4e3386395d5cf54
Reviewed-on: https://go-review.googlesource.com/c/go/+/180839
Run-TryBot: Russ Cox <rsc@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
5 years agonet: fix conf.teardown call in TestGoLookupIPOrderFallbackToFile
Russ Cox [Wed, 5 Jun 2019 16:33:30 +0000 (12:33 -0400)]
net: fix conf.teardown call in TestGoLookupIPOrderFallbackToFile

If the test fails, conf.teardown wouldn't be.
It doesn't look like it matters much, but clean up anyway.

Change-Id: I45c18095abfd49422975d061be20cbd971a98f8f
Reviewed-on: https://go-review.googlesource.com/c/go/+/180780
Run-TryBot: Russ Cox <rsc@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
5 years agoruntime: use default system stack size, not 64 kB, on non-cgo macOS
Russ Cox [Wed, 5 Jun 2019 16:31:42 +0000 (12:31 -0400)]
runtime: use default system stack size, not 64 kB, on non-cgo macOS

At least one libc call we make
(res_search, which calls _mdns_query and then mdns_item_call)
pushes a 64 kB stack frame onto the stack.
Then it faults on the guard page.

Use the default system stack size, under the assumption
that the C code being called is compatible with that stack size.

For #31705.

Change-Id: I1b0bfc2e54043c49f0709255988ef920ce30ee82
Reviewed-on: https://go-review.googlesource.com/c/go/+/180779
Run-TryBot: Russ Cox <rsc@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
5 years agocmd/go: accept -Wl,-R/path/
ng0 [Wed, 5 Jun 2019 22:01:40 +0000 (22:01 +0000)]
cmd/go: accept -Wl,-R/path/

Fixes #32167

Change-Id: If1277789ac96363430045e73e49d008bdae9be79
GitHub-Last-Rev: 71c9435c8417b19faaf66dc2e08740d7a9c22e3b
GitHub-Pull-Request: golang/go#32171
Reviewed-on: https://go-review.googlesource.com/c/go/+/178397
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 agodoc: clarify safety of multiple and concurent os.(*File).Close() calls
Alex Myasoedov [Tue, 4 Jun 2019 18:07:57 +0000 (20:07 +0200)]
doc: clarify safety of multiple and concurent os.(*File).Close() calls

Fixes #32427

Change-Id: I4b863bd3836067dcc2eb3a9c3a7169656763d003
Reviewed-on: https://go-review.googlesource.com/c/go/+/180438
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Reviewed-by: Rob Pike <r@golang.org>
5 years agoRevert "cmd/compile,runtime: allocate defer records on the stack"
Keith Randall [Wed, 5 Jun 2019 18:42:31 +0000 (18:42 +0000)]
Revert "cmd/compile,runtime: allocate defer records on the stack"

This reverts commit fff4f599fe1c21e411a99de5c9b3777d06ce0ce6.

Reason for revert: Seems to still have issues around GC.

Fixes #32452

Change-Id: Ibe7af629f9ad6a3d5312acd7b066123f484da7f0
Reviewed-on: https://go-review.googlesource.com/c/go/+/180761
Run-TryBot: Keith Randall <khr@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
5 years agocmd/link: regenerate symkind_string.go
Austin Clements [Mon, 3 Jun 2019 19:20:30 +0000 (15:20 -0400)]
cmd/link: regenerate symkind_string.go

Change-Id: Ibe514058770bd5342e1f344c5b8a9ab1427d5dfe
Reviewed-on: https://go-review.googlesource.com/c/go/+/180297
Run-TryBot: Austin Clements <austin@google.com>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

5 years agodoc: remove CL 159258 from 1.13 release notes list
Ian Lance Taylor [Wed, 5 Jun 2019 17:09:08 +0000 (10:09 -0700)]
doc: remove CL 159258 from 1.13 release notes list

It was rolled back.

Change-Id: I8372bb7e11bab7be242f4af7093a73d2fa093067
Reviewed-on: https://go-review.googlesource.com/c/go/+/180760
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
5 years agodoc: update Go 1.13 release notes using relnote
Andrew Bonventre [Wed, 5 Jun 2019 15:45:01 +0000 (11:45 -0400)]
doc: update Go 1.13 release notes using relnote

The additions were generated using golang.org/x/build/cmd/relnote.

Change-Id: Ie7322f7d01a2dd4a7bca89b9ef9c1ce93bc2671a
Reviewed-on: https://go-review.googlesource.com/c/go/+/180778
Reviewed-by: Bryan C. Mills <bcmills@google.com>
5 years agocmd/go: replace uses of ioutil.ReadFile with renameio.ReadFile
Bryan C. Mills [Tue, 4 Jun 2019 20:50:59 +0000 (16:50 -0400)]
cmd/go: replace uses of ioutil.ReadFile with renameio.ReadFile

Windows does not have atomic renames; instead, it produces one of a
handful of errors in case a read races with a rename.

CL 180219 added a utility function that retries those errors in most
cases; this change updates the locations that use renameio for writes
to also use the new renameio.ReadFile function for reads.

It remains possible for a renameio.ReadFile to fail with a spurious
ERROR_FILE_NOT_FOUND, but with retries in place for the other errors
(and practical limits on write concurrency) such failures are unlikely
in practice.

Fixes #32188

Change-Id: I78c81051cc871325c1e3229e696b921b0fcd865a
Reviewed-on: https://go-review.googlesource.com/c/go/+/180517
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/internal/renameio: add a ReadFile function
Bryan C. Mills [Fri, 31 May 2019 21:22:13 +0000 (17:22 -0400)]
cmd/go/internal/renameio: add a ReadFile function

ReadFile is a drop-in replacement for ioutil.ReadFile that works
around Windows filesystem flakiness under load.

A followup CL will replace uses of ioutil.ReadFile in cmd/go with this
function.

Updates #32188

Change-Id: I232ba893b132bdc84cd7b0edde436165a69e1aa8
Reviewed-on: https://go-review.googlesource.com/c/go/+/180219
Run-TryBot: Bryan C. Mills <bcmills@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Jay Conrod <jayconrod@google.com>
5 years agomisc/cgo/test: add test for issue 31093
Ian Lance Taylor [Tue, 4 Jun 2019 02:35:40 +0000 (19:35 -0700)]
misc/cgo/test: add test for issue 31093

Updates #31093

Change-Id: I7962aaca0b012de01768b7b42dc2283d5845eeea
Reviewed-on: https://go-review.googlesource.com/c/go/+/180377
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Dmitri Shuralyov <dmitshur@golang.org>
5 years agocmd/cgo: roll back "use C exact-width integer types to represent Go types"
Ian Lance Taylor [Tue, 4 Jun 2019 01:04:58 +0000 (18:04 -0700)]
cmd/cgo: roll back "use C exact-width integer types to represent Go types"

Roll back CL 159258 and CL 168337. Those changes broke existing
code. I can't see any way to keep existing code working while also
producing good error messages for types like C.ulong (such as the ones
already tested for in misc/cgo/errors).

This is not an exact roll back because parts of the code have changed
since those CLs.

Updates #29878
Fixes #31093

Change-Id: I56fe76c167ff0ab381ed273b9ca4b952402e1434
Reviewed-on: https://go-review.googlesource.com/c/go/+/180357
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Dmitri Shuralyov <dmitshur@golang.org>
5 years agonet/http: roll back "clean the path of the stripped URL by StripPrefix"
Dmitri Shuralyov [Tue, 4 Jun 2019 21:21:33 +0000 (21:21 +0000)]
net/http: roll back "clean the path of the stripped URL by StripPrefix"

Roll back CL 161738. That fix changed StripPrefix behavior in the
general case, not just in the situation where where stripping the
prefix from path resulted in the empty string, causing issue #31622.

That kind of change to StripPrefix behavior is not backwards compatible,
and there can be a smaller, more targeted fix for the original issue.

Fixes #31622
Updates #30165

Change-Id: Ie2fcfe6787a32e44f71d564d8f9c9d580fc6f704
Reviewed-on: https://go-review.googlesource.com/c/go/+/180498
Run-TryBot: Dmitri Shuralyov <dmitshur@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
5 years agocmd/go: document GO{NO}SUMDB in 'go help environment' and 'go help modules'
Tyler Bui-Palsulich [Tue, 28 May 2019 19:50:01 +0000 (15:50 -0400)]
cmd/go: document GO{NO}SUMDB in 'go help environment' and 'go help modules'

GOSUMDB and GONOSUMDB are described in detail by
'go help module-auth'. So, this change mentions the two
variables and says to see 'go help module-auth'.

This also adds GONOPROXY to 'go help environment'.

Fixes #32292 and updates #32056.

Change-Id: I2eae0f906a3fbfcb88ad5c8fb4870917d0d7bbcd
Reviewed-on: https://go-review.googlesource.com/c/go/+/179219
Reviewed-by: Russ Cox <rsc@golang.org>
5 years agocmd/compile,runtime: allocate defer records on the stack
Keith Randall [Thu, 11 Apr 2019 16:50:59 +0000 (09:50 -0700)]
cmd/compile,runtime: allocate defer records on the stack

When a defer is executed at most once in a function body,
we can allocate the defer record for it on the stack instead
of on the heap.

This should make defers like this (which are very common) faster.

This optimization applies to 363 out of the 370 static defer sites
in the cmd/go binary.

name     old time/op  new time/op  delta
Defer-4  52.2ns ± 5%  36.2ns ± 3%  -30.70%  (p=0.000 n=10+10)

Fixes #6980
Update #14939

Change-Id: I697109dd7aeef9e97a9eeba2ef65ff53d3ee1004
Reviewed-on: https://go-review.googlesource.com/c/go/+/171758
Run-TryBot: Keith Randall <khr@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Austin Clements <austin@google.com>
5 years agodoc: fix missing anchor links in contribute.html
Agniva De Sarker [Tue, 4 Jun 2019 03:56:35 +0000 (09:26 +0530)]
doc: fix missing anchor links in contribute.html

Some <h3>s were missing ids due to which the anchor links
weren't getting generated.

Fixes #32415

Change-Id: Ica21425c1a7c49735231c1de96b6c77dd594ce64
Reviewed-on: https://go-review.googlesource.com/c/go/+/180397
Reviewed-by: Ian Lance Taylor <iant@golang.org>
5 years agoerrors: fix package example
Andrew Gerrand [Fri, 31 May 2019 10:33:01 +0000 (20:33 +1000)]
errors: fix package example

The example in example_test.go requires that the whole file be
displayed; the addition of ExampleAs meant that only the body of the
package example function was shown, rather than the surrounding context.

This change moves ExampleAs to the file wrap_test.go file, restoring the
package example to its former glory.

Update #31716

Change-Id: Id0ea77bc06023b239a63c1d6a7c8b3c1dae91ce9
Reviewed-on: https://go-review.googlesource.com/c/go/+/179737
Reviewed-by: Marcel van Lohuizen <mpvl@golang.org>
Reviewed-by: Jean de Klerk <deklerk@google.com>
Run-TryBot: Benny Siegert <bsiegert@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

5 years agoruntime: fix sigpipe do not check SIGPIPE was ignored
LE Manh Cuong [Mon, 3 Jun 2019 08:57:43 +0000 (15:57 +0700)]
runtime: fix sigpipe do not check SIGPIPE was ignored

Fixes #32386

Change-Id: I29ad4113b02264336eebe7865895f07ef386f450
Reviewed-on: https://go-review.googlesource.com/c/go/+/180177
Run-TryBot: Cuong Manh Le <cuong.manhle.vn@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
5 years agoruntime: save/restore callee saved registers in arm64's sigtramp
Ben Shi [Wed, 15 May 2019 02:12:10 +0000 (02:12 +0000)]
runtime: save/restore callee saved registers in arm64's sigtramp

ARM64's R19-R29 and F8-F15 are callee saved registers, which
should be saved in the beginning of sigtramp, and restored at
the end.

fixes #31827

Change-Id: I622e03f1a13fec969d3a11b6a303a8a492e02bcd
Reviewed-on: https://go-review.googlesource.com/c/go/+/177045
Run-TryBot: Ben Shi <powerman1st@163.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
5 years agoruntime: get map of args of unstarted goroutines like we do for defers
Keith Randall [Mon, 3 Jun 2019 17:03:44 +0000 (10:03 -0700)]
runtime: get map of args of unstarted goroutines like we do for defers

Normally, reflect.makeFuncStub records the context value at a known
point in the stack frame, so that the runtime can get the argument map
for reflect.makeFuncStub from that known location.

This doesn't work for defers or goroutines that haven't started yet,
because they haven't allocated a frame or run an instruction yet. The
argument map must be extracted from the context value. We already do
this for defers (the non-nil ctxt arg to getArgInfo), we just need to
do it for unstarted goroutines as well.

When we traceback a goroutine, remember the context value from
g.sched.  Use it for the first frame we find.

(We never need it for deeper frames, because we normally don't stop at
 the start of reflect.makeFuncStub, as it is nosplit. With this CL we
 could allow makeFuncStub to no longer be nosplit.)

Fixes #25897

Change-Id: I427abf332a741a80728cdc0b8412aa8f37e7c418
Reviewed-on: https://go-review.googlesource.com/c/go/+/180258
Reviewed-by: Cherry Zhang <cherryyz@google.com>
5 years agocmd/go: document GO111MODULE=auto behavior for Go 1.13
Bryan C. Mills [Mon, 3 Jun 2019 16:13:54 +0000 (12:13 -0400)]
cmd/go: document GO111MODULE=auto behavior for Go 1.13

Fixes #31857

Change-Id: Ic4afe1262d863112db97794edc1ade20f6926d64
Reviewed-on: https://go-review.googlesource.com/c/go/+/180197
Reviewed-by: Jay Conrod <jayconrod@google.com>
5 years agocmd/cover: fix counting of blocks split by goto statements
Andrew Gerrand [Fri, 31 May 2019 06:58:44 +0000 (16:58 +1000)]
cmd/cover: fix counting of blocks split by goto statements

When adding coverage counters to a block, the block's statement list is
mutated. CL 77150 removed the part where the mutated list is assigned
back to its parent node; this was confusing ast.Walk, which would then
lose its place and stop walking the current block, dropping counters in
the process.

This change has addCounters make a copy of the list before mutating
it, so that the original list doesn't change under Walk's feet.

Fix #32200

Change-Id: Ia3b67d8cee860ceb7caf8748cb7a80ff9c6276e0
Reviewed-on: https://go-review.googlesource.com/c/go/+/179581
Reviewed-by: Rob Pike <r@golang.org>
5 years agosyscall/js: fix comments
Hajime Hoshi [Sun, 2 Jun 2019 16:52:50 +0000 (01:52 +0900)]
syscall/js: fix comments

As js.TypedArray no longer exists, the comment should be updated.

Change-Id: Idd1087c8007afc90307fdd965f28d3be8d8cd73e
Reviewed-on: https://go-review.googlesource.com/c/go/+/180097
Reviewed-by: Richard Musiol <neelance@gmail.com>
5 years agonet: fix the doc string on parseIPv6
frennkie [Sun, 2 Jun 2019 12:10:58 +0000 (12:10 +0000)]
net: fix the doc string on parseIPv6

Change-Id: Ie9fef7767f3263ef7daed8415ab2f59c5377a3cb
GitHub-Last-Rev: a4d13b7452f0fe060cb7575acfc200e3353fdde0
GitHub-Pull-Request: golang/go#32385
Reviewed-on: https://go-review.googlesource.com/c/go/+/180037
Reviewed-by: Agniva De Sarker <agniva.quicksilver@gmail.com>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
5 years agocmd/go: ignore build tags when 'go get' modifies build list
Jay Conrod [Fri, 31 May 2019 18:14:00 +0000 (14:14 -0400)]
cmd/go: ignore build tags when 'go get' modifies build list

In module mode, 'go get' should not consider build constraints when
loading packages in order to modify the module graph. With this
change, 'go get' considers all build tags to be true except for
"ignore" and malformed build constraint expressions.

When 'go get' builds packages, it still applies build constraints for
the target platform.

Fixes #32345

Change-Id: I6dceae6f10a5185870537de730b36292271ad124
Reviewed-on: https://go-review.googlesource.com/c/go/+/179898
Reviewed-by: Bryan C. Mills <bcmills@google.com>
5 years agocmd/compile: don't move nil checks across a VarDef
Keith Randall [Tue, 28 May 2019 21:59:23 +0000 (14:59 -0700)]
cmd/compile: don't move nil checks across a VarDef

We need to make sure that there's no possible faulting
instruction between a VarDef and that variable being
fully initialized. If there was, then anything scanning
the stack during the handling of that fault will see
a live but uninitialized variable on the stack.

If we have:

  NilCheck p
  VarDef x
  x = *p

We can't rewrite that to

  VarDef x
  NilCheck p
  x = *p

Particularly, even though *p faults on p==nil, we still
have to do the explicit nil check before the VarDef.

Fixes #32288

Change-Id: Ib8b88e6a5af3bf6f238ff5491ac86f53f3cf9fc9
Reviewed-on: https://go-review.googlesource.com/c/go/+/179239
Run-TryBot: Keith Randall <khr@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
Reviewed-by: Cuong Manh Le <cuong.manhle.vn@gmail.com>
5 years agocmd/compile: make sure build works when intrinsics are disabled
Cherry Zhang [Fri, 31 May 2019 21:11:50 +0000 (17:11 -0400)]
cmd/compile: make sure build works when intrinsics are disabled

Some runtime functions, like getcallerpc/sp, don't have Go or
assembly implementations and have to be intrinsified. Make sure
they are, even if intrinsics are disabled.

This makes "go build -gcflags=all=-d=ssa/intrinsics/off hello.go"
work.

Change-Id: I77caaed7715d3ca7ffef68a3cdc9357f095c6b9f
Reviewed-on: https://go-review.googlesource.com/c/go/+/179897
Run-TryBot: Cherry Zhang <cherryyz@google.com>
Run-TryBot: Austin Clements <austin@google.com>
Reviewed-by: Austin Clements <austin@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

5 years agocmd/go/internal/modfetch: use the resolved version to search for tags in (*codeRepo...
Bryan C. Mills [Fri, 31 May 2019 19:24:49 +0000 (15:24 -0400)]
cmd/go/internal/modfetch: use the resolved version to search for tags in (*codeRepo).convert

Previously, we used the passed-in statVers as the basis for tag search,
but it is not always valid.
Instead, use info.Name, which (by precondition) must be valid.

Updates #32161
Updates #27171

Change-Id: Iaecb5043bdf2fefd26fbe3f8e3714b07d22f580f
Reviewed-on: https://go-review.googlesource.com/c/go/+/179857
Run-TryBot: Bryan C. Mills <bcmills@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Jay Conrod <jayconrod@google.com>
5 years agotest: add a test for gccgo bug #32347
Cherry Zhang [Fri, 31 May 2019 16:32:30 +0000 (12:32 -0400)]
test: add a test for gccgo bug #32347

The gccgo compiler crashes with int-to-string conversion with
large integer constant operand. CL 179777 is the fix. This CL
adds a test.

Updates #32347.

Change-Id: Id1d9dbbcdd3addca4636f1b9c5fdbc450cc48c1d
Reviewed-on: https://go-review.googlesource.com/c/go/+/179797
Run-TryBot: Cherry Zhang <cherryyz@google.com>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

5 years agocmd/link: revise test case to work on pre-10.14 macos
Than McIntosh [Fri, 31 May 2019 16:05:07 +0000 (12:05 -0400)]
cmd/link: revise test case to work on pre-10.14 macos

Rework this recently introduced test case to insure that it works with
older versions of the OS. It was using a new framework library not
available on pre-10.14 to trigger the weak symbol reference; switch to
using a new symbol from an existing library. Tested on MacOS 10.14 and
10.11.

Updates #32233.

Change-Id: I1fe2a9255fca46cb7cdf33ff7fed67bba86fdc22
Reviewed-on: https://go-review.googlesource.com/c/go/+/179837
Run-TryBot: Than McIntosh <thanm@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: David Chase <drchase@google.com>
5 years agocmd/go: use the rev to resolve tags for pseudo-versions in mod init
LE Manh Cuong [Wed, 22 May 2019 10:52:24 +0000 (17:52 +0700)]
cmd/go: use the rev to resolve tags for pseudo-versions in mod init

Fixes #32161
Updates #31251

Change-Id: I22760836b48cd734b90bc43aacb23e23e38c0f4e
Reviewed-on: https://go-review.googlesource.com/c/go/+/178497
Reviewed-by: Bryan C. Mills <bcmills@google.com>
5 years agostrconv: document handling of NaN and ±Inf
Andrew Gerrand [Fri, 31 May 2019 10:59:35 +0000 (20:59 +1000)]
strconv: document handling of NaN and ±Inf

In addition to the example that was added in 203b80ab, mention these
special cases in the doc comment. This change also adjusts the example
to include "+Inf", as it was not otherwise mentioned that the plus
symbol may be present.

Fix #30990

Change-Id: I97d66f4aff6a17a6ccc0ee2e7f32e39ae91ae454
Reviewed-on: https://go-review.googlesource.com/c/go/+/179738
Reviewed-by: Alex Miasoedov <msoedov@gmail.com>
Reviewed-by: Benny Siegert <bsiegert@gmail.com>
Run-TryBot: Benny Siegert <bsiegert@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

5 years agomisc/cgo/errors: limit number of parallel executions
Ian Lance Taylor [Thu, 30 May 2019 22:32:27 +0000 (18:32 -0400)]
misc/cgo/errors: limit number of parallel executions

Fixes #32328

Change-Id: Iee71ecb247f2c439804c2ff03a6ed7b7f5a8b562
Reviewed-on: https://go-review.googlesource.com/c/go/+/179603
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
5 years agomisc/cgo/errors: remove debugging println
Ian Lance Taylor [Thu, 30 May 2019 22:33:14 +0000 (18:33 -0400)]
misc/cgo/errors: remove debugging println

Change-Id: I0c7da31e94307f939e1a2f023e00d5b24bac0ea6
Reviewed-on: https://go-review.googlesource.com/c/go/+/179604
Run-TryBot: Ian Lance Taylor <iant@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

5 years agoos: deflake TestNewFileNonBlock
Ian Lance Taylor [Thu, 30 May 2019 22:08:08 +0000 (18:08 -0400)]
os: deflake TestNewFileNonBlock

Fixes #32325

Change-Id: Ic06938c36a25ef1a6623e35e128b73729d02d955
Reviewed-on: https://go-review.googlesource.com/c/go/+/179698
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
5 years agonet/http: prevent Transport from spamming stderr on server 408 reply
Brad Fitzpatrick [Wed, 29 May 2019 21:49:20 +0000 (21:49 +0000)]
net/http: prevent Transport from spamming stderr on server 408 reply

HTTP 408 responses now exist and are seen in the wild (e.g. from
Google's GFE), so make Go's HTTP client not spam about them when seen.
They're normal (now).

Fixes #32310

Change-Id: I558eb4654960c74cf20db1902ccaae13d03310f6
Reviewed-on: https://go-review.googlesource.com/c/go/+/179457
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Dmitri Shuralyov <dmitshur@golang.org>
5 years agosyscall: use Ctty before fd shuffle
Greg Thelen [Sat, 25 May 2019 18:44:44 +0000 (11:44 -0700)]
syscall: use Ctty before fd shuffle

On unix if exec.Command() is given both ExtraFiles and Ctty, and the
Ctty file descriptor overlaps the range of FDs intended for the child,
then cmd.Start() the ioctl(fd,TIOCSCTTY) call fails with an
"inappropriate ioctl for device" error.

When child file descriptors overlap the new child's ctty the ctty will
be closed in the fd shuffle before the TIOCSCTTY.  Thus TIOCSCTTY is
used on one of the ExtraFiles rather than the intended Ctty file.  Thus
the error.

exec.Command() callers can workaround this by ensuring the Ctty fd is
larger than any ExtraFiles destined for the child.

Fix this by doing the ctty ioctl before the fd shuffle.

Test for this issue by modifying TestTerminalSignal to use more
ExtraFiles.  The test fails on linux and freebsd without this change's
syscall/*.go changes.  Other platforms (e.g. darwin, aix, solaris) have
the same fd shuffle logic, so the same fix is applied to them.  However,
I was only able to test on linux (32 and 64 bit) and freebsd (64 bit).

Manual runs of the test in https://golang.org/issue/29458 start passing
with this patch:
  Before:
    % /tmp/src/go/bin/go run t
    successfully ran child process with ParentExtraFileFdNum=5, ChildExtraFileFd=6, ParentPtyFd=7

    panic: failed to run child process with ParentExtraFileFdNum=10, ChildExtraFileFd=11, ParentPtyFd=11: fork/exec /bin/true: inappropriate ioctl for device

  After:
    % /tmp/src/go/bin/go run t
    successfully ran child process with ParentExtraFileFdNum=5, ChildExtraFileFd=6, ParentPtyFd=7

    successfully ran child process with ParentExtraFileFdNum=10, ChildExtraFileFd=11, ParentPtyFd=11

Fixes #29458
Change-Id: I99513de7b6073c7eb855f1eeb4d1f9dc0454ef8b
Reviewed-on: https://go-review.googlesource.com/c/go/+/178919
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
5 years agocmd/go: clarify documentation for 'go get -u'
Jay Conrod [Mon, 13 May 2019 19:11:31 +0000 (15:11 -0400)]
cmd/go: clarify documentation for 'go get -u'

Also: add a comment in internal/get.newUpgrader as a follow-up to
CL 177677.

Updates #26902

Change-Id: Ibce2807ecb44fa21697ca04a51c44ddca0f661d8
Reviewed-on: https://go-review.googlesource.com/c/go/+/176902
Reviewed-by: Bryan C. Mills <bcmills@google.com>
Run-TryBot: Bryan C. Mills <bcmills@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

5 years agocmd/go: remove support for the 'go get -m' flag
Jay Conrod [Fri, 17 May 2019 14:59:34 +0000 (10:59 -0400)]
cmd/go: remove support for the 'go get -m' flag

Fixes #32038

Change-Id: Ib4981f76572405363f404ee5038a45cb1752a2ad
Reviewed-on: https://go-review.googlesource.com/c/go/+/177879
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/link: new test case for Darwin/DWARF
Than McIntosh [Sat, 25 May 2019 01:04:49 +0000 (21:04 -0400)]
cmd/link: new test case for Darwin/DWARF

Test case for issue 32233.

Updates #32233.

Change-Id: I0e3b4a46832f39de4ef36d8fd8c6070bf9b1a019
Reviewed-on: https://go-review.googlesource.com/c/go/+/178726
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

5 years agocmd/go: fix error message for go mod in GOPATH mode
Constantin Konstantinidis [Mon, 20 May 2019 04:58:15 +0000 (06:58 +0200)]
cmd/go: fix error message for go mod in GOPATH mode

Checks if modules are enabled in GOPATH mode for go mod [graph, verify].
Added tests for GO111MODULE=[auto, off].

Fixes: #31237
Change-Id: I91efccfa10d0b2385ec2af1ea133deaa8234ba37
Reviewed-on: https://go-review.googlesource.com/c/go/+/174697
Run-TryBot: Jay Conrod <jayconrod@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Bryan C. Mills <bcmills@google.com>
Reviewed-by: Jay Conrod <jayconrod@google.com>
5 years agonet/http: enable WASM fetch where supported
Johan Brandhorst [Tue, 28 May 2019 20:51:54 +0000 (21:51 +0100)]
net/http: enable WASM fetch where supported

The existing check was introduced to allow tests to pass
on WASM without an environment where the fetch RoundTripper
could run. However, the check now prohibits the use of the
Fetch RoundTripper in all WASM tests, even where the
RoundTripper could run. The new change should only disable
the RoundTripper when used in an environment without fetch.

Fixes #32289

Change-Id: I30d2e0dbcb0e64d4b1a46b583f7e984c2a57d5e5
Reviewed-on: https://go-review.googlesource.com/c/go/+/179118
Run-TryBot: Agniva De Sarker <agniva.quicksilver@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
5 years agocmd/go: add two more uses of lazyregexp
Daniel Martí [Thu, 30 May 2019 09:23:23 +0000 (10:23 +0100)]
cmd/go: add two more uses of lazyregexp

A couple of regexp.MustCompile globals have made their way in since we
introduced this package for cmd/go.

Convert the declarations. It's just two of them, so the cmd/go exec
benchmark isn't affected noticeably.

Change-Id: Ibd0615c99b6a049124a952c59978fd714c1b9615
Reviewed-on: https://go-review.googlesource.com/c/go/+/179358
Run-TryBot: Daniel Martí <mvdan@mvdan.cc>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Bryan C. Mills <bcmills@google.com>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
5 years agocmd/link/internal/ld: ensure mach-o debug symbols are visible to App Store
Aman Gupta [Wed, 3 Apr 2019 08:49:24 +0000 (01:49 -0700)]
cmd/link/internal/ld: ensure mach-o debug symbols are visible to App Store

Passing test that shows Apple's symbols utility can now read
DWARF data in go.o, after the fix in CL174538

Updates #31022 #22716 #31459

Change-Id: I56c3517ad6d0a9f39537182f63cef56bb198aa83
Reviewed-on: https://go-review.googlesource.com/c/go/+/170451
Reviewed-by: Than McIntosh <thanm@google.com>
5 years agocmd/link/internal/ld: improve messages for testDWARF failures
Aman Gupta [Thu, 11 Apr 2019 23:20:27 +0000 (16:20 -0700)]
cmd/link/internal/ld: improve messages for testDWARF failures

Change-Id: I60af7a6477d1b25c43b311246ae4a79bf691460e
Reviewed-on: https://go-review.googlesource.com/c/go/+/171819
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

5 years agovendor: update golang.org/x/sys for aix/ppc64 fix
Tobias Klauser [Wed, 29 May 2019 16:40:06 +0000 (18:40 +0200)]
vendor: update golang.org/x/sys for aix/ppc64 fix

This pulls in CL 179317 which fixes the build on aix/ppc64.

Change-Id: I7d90950800627904503db97cfb0bc670a45b701e
Reviewed-on: https://go-review.googlesource.com/c/go/+/179357
Run-TryBot: Tobias Klauser <tobias.klauser@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
5 years agocmd/compile: fix fmt_test.go after CL 170062
Matthew Dempsky [Wed, 29 May 2019 21:30:23 +0000 (14:30 -0700)]
cmd/compile: fix fmt_test.go after CL 170062

Updates #22326.

Change-Id: Ia9173b6eb29b2a4f90f4ba39bf53b6e9b7a6d6bf
Reviewed-on: https://go-review.googlesource.com/c/go/+/179398
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
5 years agocmd/compile: sort OAS2* declarations
Matthew Dempsky [Tue, 28 May 2019 21:50:53 +0000 (14:50 -0700)]
cmd/compile: sort OAS2* declarations

Change-Id: Idd3acf5f808705c608cd4e5877bc93e1626d9a58
Reviewed-on: https://go-review.googlesource.com/c/go/+/179238
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Robert Griesemer <gri@golang.org>
5 years agocmd/compile: fix package initialization ordering
Matthew Dempsky [Thu, 28 Mar 2019 21:35:49 +0000 (14:35 -0700)]
cmd/compile: fix package initialization ordering

This CL rewrites cmd/compile's package-level initialization ordering
algorithm to be compliant with the Go spec. See documentation in
initorder.go for details.

Incidentally, this CL also improves fidelity of initialization loop
diagnostics by including referenced functions in the emitted output
like go/types does.

Fixes #22326.

Change-Id: I7c9ac47ff563df4d4f700cf6195387a0f372cc7b
Reviewed-on: https://go-review.googlesource.com/c/go/+/170062
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Robert Griesemer <gri@golang.org>
5 years agocmd/link: do not generate NT 4 compatibility binaries
Jason A. Donenfeld [Sun, 26 May 2019 13:28:05 +0000 (15:28 +0200)]
cmd/link: do not generate NT 4 compatibility binaries

Incredibly, the subsystem version numbers in the PE header influence how
win32k handles various syscalls. The first time a win32k syscall is
invoked and the kernel upgrades the thread object to a tagTHREADINFO
with all of the lovely undocumented UI members and such, it sets the
dwExpWinVer member (offset 624 in Windows 10 build 1809) to the result
of RtlGetExpWinVer(PsGetProcessSectionBaseAddress(proc)).
RtlGetExpWinVer, also undocumented, then calls into the undocumented
RtlImageNtHeader function, which returns a fortunately documented
IMAGE_NT_HEADERS structure. It uses the subsystem members in there to
set the dwExpWinVer member of our newly minted tagTHREADINFO object.
Later, functions like SendInput consult this to vary their behaviors and
return values. In fact, littered through out win32k are checks like `if
(gsti->dwExpWinVer >= 0x501) { ... }`.

I don't think Go ever supported NT 4.0. These days the minimum version
is Windows 7, which is 6.1. So, let's set the version numbers in the PE
header at that, which should give us the behavior that MSDN advertises
for various functions, as opposed to bizarre archeological remnants.

Interestingly, I suspect that most people never noticed the brokenness,
because most people trying to do serious Win32 UI stuff wind up linking
in cgo, if not for actually using C, then just to have a larger system
stack so that the stack doesn't get corrupted by various UI functions.
When MingW is used, the PE header gets a later version. But recently
there's been a bug report of some people trying to do more modest UI
manipulation using SendInput in a setting where this cgo hack probably
isn't required, so they ran into the weird historical compatibility
stuff.

Fixes #31685

Change-Id: I54461ce820f6e9df349e37be5ecc5a44c04a3e26
Reviewed-on: https://go-review.googlesource.com/c/go/+/178977
Run-TryBot: Jason Donenfeld <Jason@zx2c4.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Alex Brainman <alex.brainman@gmail.com>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
5 years agocmd/go: update docs for GOPROXY default
Russ Cox [Fri, 24 May 2019 21:42:41 +0000 (17:42 -0400)]
cmd/go: update docs for GOPROXY default

The default is now "https://proxy.golang.org,direct" not just "https://proxy.golang.org"
(as of CL 178720).

Change-Id: I47e67602e994332b988ed1b75fa57319f3876cc6
Reviewed-on: https://go-review.googlesource.com/c/go/+/178725
Run-TryBot: Russ Cox <rsc@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Bryan C. Mills <bcmills@google.com>
5 years agocmd/compile: add GCNode to VarDef in tests
Keith Randall [Wed, 29 May 2019 00:17:40 +0000 (17:17 -0700)]
cmd/compile: add GCNode to VarDef in tests

Change-Id: I0f79d317ae7ac46c7e824bd382a4f51d53adaaa6
Reviewed-on: https://go-review.googlesource.com/c/go/+/179257
Run-TryBot: Keith Randall <khr@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
5 years agosrc/vendor: update golang.org/x/sys to remove x/sys/unix dep
Brad Fitzpatrick [Tue, 28 May 2019 18:37:55 +0000 (18:37 +0000)]
src/vendor: update golang.org/x/sys to remove x/sys/unix dep

Fixes #32102

Change-Id: I3022104035a12bf31a7cd214aede58c024effc1b
Reviewed-on: https://go-review.googlesource.com/c/go/+/179180
Reviewed-by: Bryan C. Mills <bcmills@google.com>
5 years agonet/http: quiet some log spam in tests
Brad Fitzpatrick [Tue, 28 May 2019 17:45:22 +0000 (17:45 +0000)]
net/http: quiet some log spam in tests

One of these tests creates a bunch of connections concurrently, then
discovers it doesn't need them all, which then makes the server log
that the client went away midway through the TLS handshake. Perhaps
the server should recognize that as a case not worthy of logging
about, but this is a safer way to eliminate the stderr spam during go
test for now.

The other test's client gives up on its connection and closes it,
similarly confusing the server.

Change-Id: I49ce442c9a63fc437e58ca79f044aa76e8c317b5
Reviewed-on: https://go-review.googlesource.com/c/go/+/179179
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Dmitri Shuralyov <dmitshur@golang.org>
5 years agocmd/compile: process blocks containing only dead values in fuseIf
Josh Bleecher Snyder [Tue, 14 May 2019 17:11:23 +0000 (10:11 -0700)]
cmd/compile: process blocks containing only dead values in fuseIf

The code in #29218 resulted in an If block containing only its control.
That block was then converted by fuseIf into a plain block;
as a result, that control value was dead.
However, the control value was still present in b.Values.
This prevented further fusing of that block.

This change beefs up the check in fuseIf to allow fusing
blocks that contain only dead values (if any).
In the case of #29218, this enables enough extra
fusing that the control value could be eliminated,
allowing all values in turn to be eliminated.

This change also fuses 34 new blocks during make.bash.

It is not clear that this fixes every variant of #29218,
but it is a reasonable standalone change.
And code like #29218 is rare and fundamentally buggy,
so we can handle new instances if/when they actually occur.

Fixes #29218

Negligible toolspeed impact.

name        old time/op       new time/op       delta
Template          213ms ± 3%        213ms ± 2%    ~     (p=0.914 n=97+88)
Unicode          89.8ms ± 2%       89.6ms ± 2%  -0.22%  (p=0.045 n=93+95)
GoTypes           712ms ± 3%        709ms ± 2%  -0.35%  (p=0.023 n=95+95)
Compiler          3.24s ± 2%        3.23s ± 2%  -0.30%  (p=0.020 n=98+97)
SSA               10.0s ± 1%        10.0s ± 1%    ~     (p=0.382 n=98+99)
Flate             135ms ± 3%        135ms ± 2%    ~     (p=0.983 n=98+98)
GoParser          158ms ± 2%        158ms ± 2%    ~     (p=0.170 n=99+99)
Reflect           447ms ± 3%        447ms ± 2%    ~     (p=0.538 n=98+89)
Tar               189ms ± 2%        189ms ± 3%    ~     (p=0.874 n=95+96)
XML               251ms ± 2%        251ms ± 2%    ~     (p=0.434 n=94+96)
[Geo mean]        427ms             426ms       -0.15%

name        old user-time/op  new user-time/op  delta
Template          264ms ± 2%        265ms ± 2%    ~     (p=0.075 n=96+90)
Unicode           119ms ± 6%        119ms ± 7%    ~     (p=0.864 n=99+98)
GoTypes           926ms ± 2%        924ms ± 2%    ~     (p=0.071 n=94+94)
Compiler          4.38s ± 2%        4.37s ± 2%  -0.34%  (p=0.001 n=98+97)
SSA               13.4s ± 1%        13.4s ± 1%    ~     (p=0.693 n=90+93)
Flate             162ms ± 3%        161ms ± 2%    ~     (p=0.163 n=99+99)
GoParser          186ms ± 2%        186ms ± 3%    ~     (p=0.130 n=96+100)
Reflect           572ms ± 3%        572ms ± 2%    ~     (p=0.608 n=97+97)
Tar               239ms ± 2%        239ms ± 3%    ~     (p=0.999 n=93+91)
XML               302ms ± 2%        302ms ± 2%    ~     (p=0.627 n=91+97)
[Geo mean]        540ms             540ms       -0.08%

file    before    after     Δ       %
asm     4862704   4858608   -4096   -0.084%
compile 24001568  24001680  +112    +0.000%
total   132520780 132516796 -3984   -0.003%

file                       before    after     Δ       %
cmd/compile/internal/gc.a  8887638   8887596   -42     -0.000%
cmd/compile/internal/ssa.a 29995056  29998986  +3930   +0.013%
cmd/internal/obj/wasm.a    209444    203652    -5792   -2.765%
total                      129471798 129469894 -1904   -0.001%

Change-Id: I2d18f9278e68b9766058ae8ca621e844f9d89dd8
Reviewed-on: https://go-review.googlesource.com/c/go/+/177140
Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
Reviewed-by: Cuong Manh Le <cuong.manhle.vn@gmail.com>
5 years agotesting/internal/testdeps: delete didSetLogger variable
Dmitri Shuralyov [Tue, 28 May 2019 15:37:12 +0000 (11:37 -0400)]
testing/internal/testdeps: delete didSetLogger variable

It was added in CL 83956 but never used.

Updates #23129

Change-Id: I70b50e974a56620069a77658386722af314cc857
Reviewed-on: https://go-review.googlesource.com/c/go/+/179138
Run-TryBot: Dmitri Shuralyov <dmitshur@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

5 years agobytes: remove obsolete comment
Maxim Eryomenko [Tue, 28 May 2019 02:45:21 +0000 (02:45 +0000)]
bytes: remove obsolete comment

Change-Id: Iec4348eb674c8c64497c82e5d7cd656736cfcebd
GitHub-Last-Rev: 462c323960be0ca1c126c051ab97f72949bdc2b6
GitHub-Pull-Request: golang/go#32225
Reviewed-on: https://go-review.googlesource.com/c/go/+/178797
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
5 years agonet/http: fix TestTransportServerClosingUnexpectedly flake
Brad Fitzpatrick [Sat, 25 May 2019 18:47:22 +0000 (18:47 +0000)]
net/http: fix TestTransportServerClosingUnexpectedly flake

Fixes #32119

Change-Id: I8cf2e2e69737e2485568af91ab75149f3cf66781
Reviewed-on: https://go-review.googlesource.com/c/go/+/178918
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
5 years agocmd/doc: always print package clause except for commands
Rob Pike [Fri, 17 May 2019 03:43:51 +0000 (13:43 +1000)]
cmd/doc: always print package clause except for commands

There was an implicit heuristic before about when to print the
package clause or omit it, but it was undocumented and confusing.
Get rid of it and print it always unless asking for the package
docs for a command, which are more of a usage message than a
programming question. This simplifies the processing.

There are several paths to the output, so to put the fix in one
place we place a wrapper before the output buffer than adds the
clause when Write is first called.

The tests don't verify this behavior, but they didn't before either.
Unsure what the right approach is but this will do for now.

Fixes #31457

Change-Id: Ia6a9e740d556f45265c55f06b5306621c7a40ea9
Reviewed-on: https://go-review.googlesource.com/c/go/+/177797
Reviewed-by: Russ Cox <rsc@golang.org>
5 years agoruntime: remove VDSO fallback test and benchmarks
Ian Lance Taylor [Wed, 1 May 2019 04:06:01 +0000 (21:06 -0700)]
runtime: remove VDSO fallback test and benchmarks

These tests assume that it is OK to switch between time implementations,
but the clock_gettime call uses CLOCK_MONOTONIC and the fallback call,
gettimeofday, uses CLOCK_REALTIME. Disabling the clock_gettime call means
that calls to nanotime will start returning very different values.
This breaks the new timer code, which assumes that nanotime will return
a consistently increasing value.

This test is not very useful in any case as it doesn't check the results.

Removing this file also removes BenchmarkTimeNow, which is a duplicate
of BenchmarkNow in the time package.

Updates #27707
Fixes #32109

Change-Id: I6a884af07f75822d724193c5eed94742f524f07d
Reviewed-on: https://go-review.googlesource.com/c/go/+/174679
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
5 years agocmd/doc: display a proper error when no packages are found
Agniva De Sarker [Sat, 25 May 2019 18:02:28 +0000 (20:02 +0200)]
cmd/doc: display a proper error when no packages are found

Fixes #32044

Change-Id: Ia220dbbe4d6851befe28dd2d4b8a1a7bc2757460
Reviewed-on: https://go-review.googlesource.com/c/go/+/178878
Run-TryBot: Agniva De Sarker <agniva.quicksilver@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Rob Pike <r@golang.org>
5 years agocmd/link: fix incorrect constant for macho-o load command
Than McIntosh [Fri, 24 May 2019 19:38:47 +0000 (15:38 -0400)]
cmd/link: fix incorrect constant for macho-o load command

The constant defined in macho.go for LC_LOAD_WEAK_DYLIB was
not correct, was 0x18 should have been 0x80000018. Switch
to the correct definition.

Fixes #32233.

Change-Id: I9fb660a3cfd5e8c451a64947258f7ead76d98c79
Reviewed-on: https://go-review.googlesource.com/c/go/+/178723
Run-TryBot: Than McIntosh <thanm@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Reviewed-by: David Chase <drchase@google.com>
5 years agocmd/go: respect default proxy setting, add direct fallback
Russ Cox [Fri, 24 May 2019 12:55:30 +0000 (08:55 -0400)]
cmd/go: respect default proxy setting, add direct fallback

Getenv("GOPROXY") says what the environment variable is
(including looking in the go env file), but it doesn't include
the default setting. This code needs to use cfg.GOPROXY
to get the actual default. Fix and test that.

Also, we forgot to include the fallback to direct for when
the proxy serves a 404. Add and test that too.

Also add HTTP fetch information to -x build flag output.
(It does not belong in the -v output, despite the GOPATH go get
command doing this.)

Change-Id: Ieab7ef13cda3e1ad041dbe04921af206e2232c9c
Reviewed-on: https://go-review.googlesource.com/c/go/+/178720
Run-TryBot: Russ Cox <rsc@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Bryan C. Mills <bcmills@google.com>
Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
5 years agocmd/go: when resolving packages, try all module paths before falling back to the...
Bryan C. Mills [Thu, 16 May 2019 13:21:49 +0000 (09:21 -0400)]
cmd/go: when resolving packages, try all module paths before falling back to the next proxy

Since we're mucking with error-propagation in modload.Query* anyway,
simplify the classification logic. Ensure that “module not found”
errors are reported as such in various places, since non-“not found”
errors terminate the module search.

Fixes #31785

Change-Id: Ie3ca5f4eec10a5f2a6037ec7e1c2cf47bd37a232
Reviewed-on: https://go-review.googlesource.com/c/go/+/177958
Run-TryBot: Russ Cox <rsc@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Russ Cox <rsc@golang.org>
5 years agocmd, runtime: remove PC_F & PC_B globals on Wasm
Cherry Zhang [Thu, 25 Apr 2019 21:25:54 +0000 (17:25 -0400)]
cmd, runtime: remove PC_F & PC_B globals on Wasm

Following the previous CL, this removes more global variables on
Wasm.

PC_B is used mostly for intra-function jumps, and for a function
telling its callee where to start or resume. This usage can be
served by a parameter. The top level loop (wasm_pc_f_loop) uses
PC_B for resuming a function. This value is either set by gogo,
or loaded from the Go stack at function return. Instead of
loading PC_B at each function return, we could make gogo stores
PC_B at the same stack location, and let the top level loop do
the load. This way, we don't need to use global PC_B to
communicate with the top level loop, and we can replace global
PC_B with a parameter.

PC_F is similar. It is even more so in that the only reader is
the top level loop. Let the top level loop read it from the stack,
and we can get rid of PC_F entirely.

PC_F and PC_B are used less entensively as SP, so this CL has
smaller performance gain.

Running on Chrome 74.0.3729.108 on Linux/AMD64,

name                   old time/op    new time/op    delta
BinaryTree17              16.6s ± 0%     16.2s ± 1%   -2.59%  (p=0.016 n=4+5)
Fannkuch11                11.1s ± 1%     10.8s ± 0%   -2.65%  (p=0.008 n=5+5)
FmtFprintfEmpty           231ns ± 1%     217ns ± 0%   -6.06%  (p=0.008 n=5+5)
FmtFprintfString          407ns ± 3%     375ns ± 2%   -7.81%  (p=0.008 n=5+5)
FmtFprintfInt             466ns ± 2%     430ns ± 0%   -7.79%  (p=0.016 n=5+4)
FmtFprintfIntInt          719ns ± 2%     673ns ± 2%   -6.37%  (p=0.008 n=5+5)
FmtFprintfPrefixedInt     706ns ± 1%     676ns ± 3%   -4.31%  (p=0.008 n=5+5)
FmtFprintfFloat          1.01µs ± 1%    0.97µs ± 1%   -4.30%  (p=0.008 n=5+5)
FmtManyArgs              2.67µs ± 1%    2.51µs ± 1%   -5.95%  (p=0.008 n=5+5)
GobDecode                30.7ms ± 9%    31.3ms ±34%     ~     (p=0.222 n=5+5)
GobEncode                24.2ms ±23%    20.2ms ± 0%  -16.36%  (p=0.016 n=5+4)
Gzip                      852ms ± 0%     823ms ± 0%   -3.38%  (p=0.016 n=4+5)
Gunzip                    160ms ± 1%     151ms ± 1%   -5.37%  (p=0.008 n=5+5)
JSONEncode               35.7ms ± 1%    34.3ms ± 1%   -3.81%  (p=0.008 n=5+5)
JSONDecode                247ms ± 8%     254ms ± 7%     ~     (p=0.548 n=5+5)
Mandelbrot200            5.39ms ± 0%    5.41ms ± 0%   +0.42%  (p=0.008 n=5+5)
GoParse                  18.5ms ± 1%    18.3ms ± 2%     ~     (p=0.343 n=4+4)
RegexpMatchEasy0_32       424ns ± 2%     397ns ± 0%   -6.23%  (p=0.008 n=5+5)
RegexpMatchEasy0_1K      2.88µs ± 0%    2.86µs ± 1%     ~     (p=0.079 n=5+5)
RegexpMatchEasy1_32       395ns ± 2%     370ns ± 1%   -6.23%  (p=0.008 n=5+5)
RegexpMatchEasy1_1K      3.26µs ± 0%    3.19µs ± 1%   -2.06%  (p=0.008 n=5+5)
RegexpMatchMedium_32      564ns ± 1%     532ns ± 0%   -5.71%  (p=0.008 n=5+5)
RegexpMatchMedium_1K      146µs ± 2%     140µs ± 1%   -4.62%  (p=0.008 n=5+5)
RegexpMatchHard_32       8.47µs ± 1%    7.91µs ± 1%   -6.65%  (p=0.008 n=5+5)
RegexpMatchHard_1K        253µs ± 1%     236µs ± 2%   -6.66%  (p=0.008 n=5+5)
Revcomp                   1.78s ± 4%     1.76s ± 5%     ~     (p=1.000 n=5+5)
Template                  292ms ±29%     269ms ± 5%     ~     (p=0.690 n=5+5)
TimeParse                1.61µs ± 4%    1.54µs ± 1%   -4.42%  (p=0.008 n=5+5)
TimeFormat               1.66µs ± 3%    1.58µs ± 1%   -5.22%  (p=0.008 n=5+5)
[Geo mean]                232µs          221µs        -4.54%

name                   old speed      new speed      delta
GobDecode              25.0MB/s ± 8%  25.1MB/s ±27%     ~     (p=0.222 n=5+5)
GobEncode              32.8MB/s ±21%  38.0MB/s ± 0%  +15.84%  (p=0.016 n=5+4)
Gzip                   22.8MB/s ± 0%  23.6MB/s ± 0%   +3.49%  (p=0.016 n=4+5)
Gunzip                  121MB/s ± 1%   128MB/s ± 1%   +5.68%  (p=0.008 n=5+5)
JSONEncode             54.4MB/s ± 1%  56.5MB/s ± 1%   +3.97%  (p=0.008 n=5+5)
JSONDecode             7.88MB/s ± 8%  7.65MB/s ± 8%     ~     (p=0.548 n=5+5)
GoParse                3.07MB/s ± 8%  3.00MB/s ±22%     ~     (p=0.579 n=5+5)
RegexpMatchEasy0_32    75.6MB/s ± 2%  80.5MB/s ± 0%   +6.58%  (p=0.008 n=5+5)
RegexpMatchEasy0_1K     356MB/s ± 0%   358MB/s ± 1%     ~     (p=0.095 n=5+5)
RegexpMatchEasy1_32    81.1MB/s ± 2%  86.5MB/s ± 1%   +6.69%  (p=0.008 n=5+5)
RegexpMatchEasy1_1K     314MB/s ± 0%   320MB/s ± 0%   +2.10%  (p=0.008 n=5+5)
RegexpMatchMedium_32   1.77MB/s ± 1%  1.88MB/s ± 0%   +6.09%  (p=0.016 n=5+4)
RegexpMatchMedium_1K   6.99MB/s ± 2%  7.33MB/s ± 1%   +4.83%  (p=0.008 n=5+5)
RegexpMatchHard_32     3.78MB/s ± 1%  4.04MB/s ± 1%   +7.04%  (p=0.008 n=5+5)
RegexpMatchHard_1K     4.04MB/s ± 1%  4.33MB/s ± 2%   +7.17%  (p=0.008 n=5+5)
Revcomp                 143MB/s ± 4%   145MB/s ± 5%     ~     (p=1.000 n=5+5)
Template               6.77MB/s ±24%  7.22MB/s ± 5%     ~     (p=0.690 n=5+5)
[Geo mean]             24.4MB/s       25.4MB/s        +4.18%

Change-Id: Ib80716e62992aec28b2c4a96af280c278f83aa49
Reviewed-on: https://go-review.googlesource.com/c/go/+/173980
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Richard Musiol <neelance@gmail.com>
5 years agoruntime: ensure mheap lock stack growth invariant is maintained
Michael Anthony Knyszek [Fri, 17 May 2019 14:48:04 +0000 (14:48 +0000)]
runtime: ensure mheap lock stack growth invariant is maintained

Currently there's an invariant in the runtime wherein the heap lock
can only be acquired on the system stack, otherwise a self-deadlock
could occur if the stack grows while the lock is held.

This invariant is upheld and documented in a number of situations (e.g.
allocManual, freeManual) but there are other places where the invariant
is either not maintained at all which risks self-deadlock (e.g.
setGCPercent, gcResetMarkState, allocmcache) or is maintained but
undocumented (e.g. gcSweep, readGCStats_m).

This change adds go:systemstack to any function that acquires the heap
lock or adds a systemstack(func() { ... }) around the critical section,
where appropriate. It also documents the invariant on (*mheap).lock
directly and updates repetitive documentation to refer to that comment.

Fixes #32105.

Change-Id: I702b1290709c118b837389c78efde25c51a2cafb
Reviewed-on: https://go-review.googlesource.com/c/go/+/177857
Run-TryBot: Michael Knyszek <mknyszek@google.com>
Reviewed-by: Austin Clements <austin@google.com>
5 years agostrings: clarify example of ContainsAny
Martin Sucha [Thu, 23 May 2019 18:34:17 +0000 (20:34 +0200)]
strings: clarify example of ContainsAny

I have seen code that literally copied the example like this:

    if strings.ContainsAny(s, "1 & 2 & 3") {

The developer apparently thought that this is the way to
specify multiple characters and I noticed this pattern
being used in the example. Let's update the example so
that it's clear how multiple Unicode code points should
be specified.

Change-Id: Id4d780555e521af62fb787a7950be1e60848cd95
Reviewed-on: https://go-review.googlesource.com/c/go/+/178737
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

5 years agoimage/draw: change argument type to be consistent with other args
Leon Klingele [Fri, 24 May 2019 15:04:28 +0000 (15:04 +0000)]
image/draw: change argument type to be consistent with other args

Change-Id: Ic5b01c0a6912d87e812331d6525939df8f525cca
GitHub-Last-Rev: b6ffefcf8f2ec1ed7ee13f36e8fef6af965ba332
GitHub-Pull-Request: golang/go#29993
Reviewed-on: https://go-review.googlesource.com/c/go/+/160417
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>

5 years agocmd/vendor: sync github.com/google/pprof@v0.0.0-20190515194954-54271f7e092f
Hana (Hyang-Ah) Kim [Fri, 24 May 2019 15:01:01 +0000 (11:01 -0400)]
cmd/vendor: sync github.com/google/pprof@v0.0.0-20190515194954-54271f7e092f

Change-Id: If001d8f7e657e01711653827a170904932839e34
Reviewed-on: https://go-review.googlesource.com/c/go/+/178721
Run-TryBot: Hyang-Ah Hana Kim <hyangah@gmail.com>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

5 years agocompress/gzip: add missing error check in test
Leon Klingele [Mon, 4 Feb 2019 12:43:05 +0000 (12:43 +0000)]
compress/gzip: add missing error check in test

Change-Id: Ib7be503a46f3609ea29464569e57e1f620e97a53
GitHub-Last-Rev: a43cffe45f3b9d208028f7682088cbe537d6deca
GitHub-Pull-Request: golang/go#30007
Reviewed-on: https://go-review.googlesource.com/c/go/+/160431
Run-TryBot: Leon Klingele <git@leonklingele.de>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
5 years agosyscall/js: replace TypedArrayOf with CopyBytesToGo/CopyBytesToJS
Richard Musiol [Wed, 15 May 2019 23:03:10 +0000 (01:03 +0200)]
syscall/js: replace TypedArrayOf with CopyBytesToGo/CopyBytesToJS

The typed arrays returned by TypedArrayOf were backed by WebAssembly
memory. They became invalid each time we grow the WebAssembly memory.
This made them very error prone and hard to use correctly.

This change removes TypedArrayOf completely and instead introduces
CopyBytesToGo and CopyBytesToJS for copying bytes between a byte
slice and an Uint8Array. This breaking change is still allowed for
the syscall/js package.

Fixes #31980.
Fixes #31812.

Change-Id: I14c76fdd60b48dd517c1593972a56d04965cb272
Reviewed-on: https://go-review.googlesource.com/c/go/+/177537
Run-TryBot: Richard Musiol <neelance@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
5 years agocrypto/ecdsa: implement ecdsa on s390x for P256/P384/P521 using KDSA instruction
bill_ofarrell [Thu, 16 May 2019 16:45:52 +0000 (12:45 -0400)]
crypto/ecdsa: implement ecdsa on s390x for P256/P384/P521 using KDSA instruction

Utilize KDSA when available. This guarantees constant time operation on all three curves mentioned,
and is faster than conventional assembly. The IBM Z model(s) that support KDSA as used in this CL
are not yet publicly available, and so we are unable to release performance data at this time.

Change-Id: I85360dcf90fe42d2bf32afe3f638e282de10a518
Reviewed-on: https://go-review.googlesource.com/c/go/+/174437
Run-TryBot: Michael Munday <mike.munday@ibm.com>
Reviewed-by: Michael Munday <mike.munday@ibm.com>
5 years agoSECURITY.md: add security file
Brad Fitzpatrick [Thu, 23 May 2019 14:01:16 +0000 (14:01 +0000)]
SECURITY.md: add security file

This is now recognized and recommended by GitHub.

Fixes #32201

Change-Id: Iafb5ef1b2bee5f021a711b0b758aaf6a74758c5d
Reviewed-on: https://go-review.googlesource.com/c/go/+/178697
Reviewed-by: Ian Lance Taylor <iant@golang.org>