net: run all timeout tests in parallel

For #10571.

Change-Id: I9a42226078b9c52dbe0c65cb101b5f452233e911
Reviewed-on: https://go-review.googlesource.com/18205
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Reviewed-by: Mikio Hara <mikioh.mikioh@gmail.com>
Run-TryBot: Russ Cox <rsc@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

misc/cgo/testcshared: print a little more about GOROOT on failure

For #13789.

Change-Id: I83973298a35afcf55627f0a72223098306a51f4b
Reviewed-on: https://go-review.googlesource.com/18233
Reviewed-by: Ian Lance Taylor <iant@golang.org>

cmd/dist: wait for pending tests before exiting

When 'go tool dist test' stops, it was intended that it first wait for
pending background tests, like a failed compilation waits for pending
background compiles. But these three lines prevented that.
Fix by deleting them. (The actual loop already contains the correct
logic to avoid running the others and to wait for what's left.)

Change-Id: I4e945495ada903fb0af567910626241bc1c52ba6
Reviewed-on: https://go-review.googlesource.com/18232
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

net/http: relax recently-updated rules and behavior of CloseNotifier

The CloseNotifier implementation and documentation was
substantially changed in https://golang.org/cl/17750 but it was a bit
too aggressive.

Issue #13666 highlighted that in addition to breaking external
projects, even the standard library (httputil.ReverseProxy) didn't
obey the new rules about not using CloseNotifier until the
Request.Body is fully consumed.

So, instead of fixing httputil.ReverseProxy, dial back the rules a
bit. It's now okay to call CloseNotify before consuming the request
body. The docs now say CloseNotifier may wait to fire before the
request body is fully consumed, but doesn't say that the behavior is
undefined anymore. Instead, we just wait until the request body is
consumed and start watching for EOF from the client then.

This CL also adds a test to ReverseProxy (using a POST request) that
would've caught this earlier.

Fixes #13666

Change-Id: Ib4e8c29c4bfbe7511f591cf9ffcda23a0f0b1269
Reviewed-on: https://go-review.googlesource.com/18144
Reviewed-by: Russ Cox <rsc@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>

cmd/internal/obj: fix PCSP table at runtime.morestack calls

Fixes #13346.

Change-Id: Ic903ee90575e8dbe23905d0678d3295745d1d47f
Reviewed-on: https://go-review.googlesource.com/18154
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
Reviewed-by: Russ Cox <rsc@golang.org>

os: remove unused import to fix build

Change-Id: Ia8c1c77590115a5ffda144962436d489ed77a423
Reviewed-on: https://go-review.googlesource.com/18227
Reviewed-by: Ian Lance Taylor <iant@golang.org>

os, runtime: better EPIPE behavior for command line programs

Old behavior: 10 consecutive EPIPE errors on any descriptor cause the
program to exit with a SIGPIPE signal.

New behavior: an EPIPE error on file descriptors 1 or 2 cause the
program to raise a SIGPIPE signal.  If os/signal.Notify was not used to
catch SIGPIPE signals, this will cause the program to exit with SIGPIPE.
An EPIPE error on a file descriptor other than 1 or 2 will simply be
returned from Write.

Fixes #11845.
Update #9896.

Change-Id: Ic85d77e386a8bb0255dc4be1e4b3f55875d10f18
Reviewed-on: https://go-review.googlesource.com/18151
Reviewed-by: Russ Cox <rsc@golang.org>

runtime: disable a signal by restoring the original disposition

Fixes #13034.
Fixes #13042.
Update #9896.

Change-Id: I189f381090223dd07086848aac2d69d2c00d80c4
Reviewed-on: https://go-review.googlesource.com/18062
Reviewed-by: Russ Cox <rsc@golang.org>

cmd/go: update out of date help text about vendoring

Change-Id: I2b61f3b3ecf28d8f6a8dff94d194b6d3d450ea22
Reviewed-on: https://go-review.googlesource.com/17996
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Reviewed-by: Russ Cox <rsc@golang.org>

testing: add clear panic for duplicate call to t.Parallel

Change-Id: I155633b58e1823344a26c3edf11f5626fae080ee
Reviewed-on: https://go-review.googlesource.com/18204
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>

runtime/pprof: shorten a few tests

For #10571.

Change-Id: I4bdad64e2dfd692ef2adccf2e5e82e9b1996a8ea
Reviewed-on: https://go-review.googlesource.com/18206
Reviewed-by: Austin Clements <austin@google.com>
Run-TryBot: Austin Clements <austin@google.com>

cmd/link: use current GOROOT for source file paths for standard library

This CL changes the source file information in the
standard library's .a files to say "$GOROOT/src/runtime/chan.go"
(with a literal "$GOROOT") instead of spelling out the actual directory.
The linker then substitutes the actual $GOROOT (or $GOROOT_FINAL)
as appropriate.

If people download a binary distribution to an alternate location,
following the instructions at https://golang.org/doc/install#install,
the code before this CL would end up with source paths pointing to
/usr/local/go no matter where the actual sources were.
Now the source paths for built binaries will point to the actual sources
(hopefully).

The source line information in distributed binaries is not affected:
those will still say /usr/local/go. But binaries people build themselves
(their own programs, not the go distribution programs) will be correct.

Fixing this path also fixes the lookup of the runtime-gdb.py file.

Fixes #5533.

Change-Id: I03729baae3fbd8cd636e016275ee5ad2606e4663
Reviewed-on: https://go-review.googlesource.com/18200
Run-TryBot: Russ Cox <rsc@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>

build: prevent the entire repo to be wiped out by cgo test scripts

Following the parallelization of some tests, a race condition can
occur in testcarchive, testshared and testcshared.

In some cases, it can result in the go env GOROOT command returning
corrupted data, which are then passed to a rm command.

Make the shell script more robust by not trusting the result of
the go env GOROOT command. It does not really fix the issue, but
at least prevent the entire repository to be deleted.

Updates #13789

Change-Id: Iaf04a7bd078ed3a82e724e35c4b86e6f756f2a2f
Reviewed-on: https://go-review.googlesource.com/18173
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Russ Cox <rsc@golang.org>

net/http: deflake tests in full mode after t.Parallel additions

https://golang.org/cl/18087 added a bunch of t.Parallel calls, which
aren't compatible with the afterTest func. But in short mode, afterTest
is a no-op. To keep all.bash (short mode) fast, conditionally set
t.Parallel when in short mode, but keep it unset for compatibility with
afterFunc otherwise.

Fixes #13804

Change-Id: Ie841fbc2544e1ffbee43ba1afbe895774e290da0
Reviewed-on: https://go-review.googlesource.com/18143
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

runtime: remove now-unused test string constants

Change-Id: I491197e1505d02cd107a8788e5377cf1d0a9828c
Reviewed-on: https://go-review.googlesource.com/18157
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>

runtime: implement setsigstack for Solaris

Change-Id: I0280d478b7d0a59d8e2082aa87fab6d7d9f36a18
Reviewed-on: https://go-review.googlesource.com/18158
Reviewed-by: Aram Hăvărneanu <aram@mgk.ro>

doc: fix typo in install-source.html.

Change-Id: I6ea7339ff9412111319e46f2962c6b3880987a1e
Reviewed-on: https://go-review.googlesource.com/18174
Reviewed-by: Minux Ma <minux@golang.org>

runtime: treat SIGTSTP/SIGCANCEL on Solaris as we do on GNU/Linux

Change-Id: I617abd53f5fc883b972a1ef090886b85607e00bb
Reviewed-on: https://go-review.googlesource.com/18155
Reviewed-by: Aram Hăvărneanu <aram@mgk.ro>

doc: fix typo

Fixes #13780.

Change-Id: I629e2ba79b74d693e04c3747812c9a686cae5335
Reviewed-on: https://go-review.googlesource.com/18218
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>

doc: 2016 is the Year of the Gopher.

Change-Id: I85caba8c743dcd82954de75df947053b3d206bdc
Reviewed-on: https://go-review.googlesource.com/18117
Reviewed-by: Minux Ma <minux@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
Reviewed-by: Alex Brainman <alex.brainman@gmail.com>

os: change Open(`C:`) to open current directory on C:

Open(`C:`) currently opens root directory on C:. Change that to open
current directory on C:. Just like cmd.exe's "dir C:" command does.
Just like FindFirstFile("C:*") Windows API does. It is also consistent
with what filepath.Join("C:", "a") currently does.

Fixes #13763

Change-Id: I60b6e7d80215d110bbbb6265c9f32717401638c6
Reviewed-on: https://go-review.googlesource.com/18184
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Alex Brainman <alex.brainman@gmail.com>

database/sql: fix doc typos

Change-Id: I374dabed6bf9783839d637e9d7fd6f4e61c7eecf
Reviewed-on: https://go-review.googlesource.com/18183
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>

os: fix race condition between (*os.File).Stat and os.Chdir on windows

Fixes #13752

Change-Id: I53cfc4ecae90c35b6f1074f3be08489c408a6464
Reviewed-on: https://go-review.googlesource.com/18181
Reviewed-by: Alex Brainman <alex.brainman@gmail.com>

runtime: move test programs out of source code, coalesce

Now there are just three programs to compile instead of many,
and repeated tests can reuse the compilation result instead of
rebuilding it.

Combined, these changes reduce the time spent testing runtime
during all.bash on my laptop from about 60 to about 30 seconds.
(All.bash itself runs in 5½ minutes.)

For #10571.

Change-Id: Ie2c1798b847f1a635a860d11dcdab14375319ae9
Reviewed-on: https://go-review.googlesource.com/18085
Reviewed-by: Austin Clements <austin@google.com>
Run-TryBot: Austin Clements <austin@google.com>

net/http: update docs on Request.Proto, ProtoMajor, ProtoMinor

Change-Id: I4a6928b4674b6aaab3611cad7526347923a0015f
Reviewed-on: https://go-review.googlesource.com/18153
Reviewed-by: Ian Lance Taylor <iant@golang.org>

cmd/dist: apply final edits from CL 18199.

I broke the rule: never click the Submit button on the web.

Change-Id: If81a5cc31c1f28664960bad124cc596f5cab1222
Reviewed-on: https://go-review.googlesource.com/18203
Reviewed-by: Russ Cox <rsc@golang.org>

cmd/dist: run shards of test dir in parallel

Saves 15 seconds from all.bash on my laptop (3:20 -> 3:05).

Change-Id: Ic5dc3c7804e78b584789dd856a3dada94000a8e2
Reviewed-on: https://go-review.googlesource.com/18199
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>

cmd/dist: run wiki, shootout etc tests only on builders

These find approximately nothing.
Takes 5% off my all.bash run time.

For #10571.

Change-Id: I21d3a844af756eb37f59bba0064f24995626da0d
Reviewed-on: https://go-review.googlesource.com/18198
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>

cmd/dist: run various one-off tests in parallel

Takes 15% off my all.bash run time
(after this and earlier CLs, now down to 3½ from 5½ minutes).

For #10571.

Change-Id: Iac316ffb730c9ff0a0faa7cc3b82ed4f7e6d4361
Reviewed-on: https://go-review.googlesource.com/18088
Reviewed-by: Ian Lance Taylor <iant@golang.org>

cmd/dist: run limited tests in race and libgcc=none modes

Mostly we just care that the test binaries link and start up.
No need to run the full test suites.

Takes 12% off my all.bash run time.

For #10571.

Change-Id: I01af618f3d51deb841ea638424e1389a2df7d746
Reviewed-on: https://go-review.googlesource.com/18086
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Russ Cox <rsc@golang.org>

build: shorten a few packages with long tests

Takes 3% off my all.bash run time.

For #10571.

Change-Id: I8f00f523d6919e87182d35722a669b0b96b8218b
Reviewed-on: https://go-review.googlesource.com/18087
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Run-TryBot: Russ Cox <rsc@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

api/go1.6.txt: add debug.SetTraceback, build.Package.InvalidGoFiles

Change-Id: Ib00e866e29681631f6fa3a14e7d81c25fc3c8500
Reviewed-on: https://go-review.googlesource.com/18052
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>

cmd/dist: make test default to --no-rebuild

I'm tired of having to remember it on every command.
Rebuilding everything is the wrong default.

This CL updates the build script, but the builders may
(or may not) need work, depending on whether they
rebuild using the test command (I doubt it).

Change-Id: I21f202a2f13e73df3f6bd54ae6a317c467b68151
Reviewed-on: https://go-review.googlesource.com/18084
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>

os: add missing copyright notice

Change-Id: I496b701e2fcc944c764002643c7b0b2ce9e08806
Reviewed-on: https://go-review.googlesource.com/18182
Reviewed-by: Dave Cheney <dave@cheney.net>

runtime: fix, simplify, and improve scan state in goroutine header

Currently goroutineheader goes through some convolutions to *almost*
print the scan state of a G. However, the code path that would print
the scan state of the G refers to gStatusStrings where it almost
certainly meant to refer to gScanStatusStrings (which is unused), so
it winds up printing the regular status string without the scan state
either way. Furthermore, if the G is in _Gwaiting, we override the
status string and lose where this would indicate the scan state if it
worked.

This commit fixes this so the runtime prints the scan state. However,
rather than using a parallel list of status strings, this simply adds
a conditional print if the scan bit is set. This lets us remove the
string list, prints the scan state even in _Gwaiting, and lets us
strip off the scan bit at the beginning of the function, which
simplifies the rest of it.

Change-Id: Ic0adbe5c05abf4adda93da59f93b578172b28e3d
Reviewed-on: https://go-review.googlesource.com/18092
Reviewed-by: Keith Randall <khr@golang.org>

runtime: fix sigaltstack test Skipf format string

Change-Id: I9485ecbd9d546893e4f0db846b08d835fa7515d7
Reviewed-on: https://go-review.googlesource.com/18140
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>

runtime: adjust gsignal stack to current signal stack

If non-Go code calls sigaltstack before a signal is received, use
sigaltstack to determine the current signal stack and set the gsignal
stack to use it.  This makes the Go runtime more robust in the face of
non-Go code.  We still can't handle a disabled signal stack or a signal
triggered with SA_ONSTACK clear, but we now give clear errors for those
cases.

Fixes #7227.
Update #9896.

Change-Id: Icb1607e01fd6461019b6d77d940e59b3aed4d258
Reviewed-on: https://go-review.googlesource.com/18102
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Mikio Hara <mikioh.mikioh@gmail.com>

image/color: have NYCbCrA.RGBA work in 16-bit color.

This makes NYCbCrA consistent with YCbCr.

Fixes #13706.

Change-Id: Ifced84372e4865925fa6efef9ca2f1de43da70e0
Reviewed-on: https://go-review.googlesource.com/18115
Reviewed-by: Rob Pike <r@golang.org>

net/http: fix typo in docstring

s/activitiy/activity

Change-Id: Ib2bbc929b38b1993000da57daed2d795f4a93997
Reviewed-on: https://go-review.googlesource.com/18131
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>

crypto/cipher: fix typo from last change

s/encrypt/decrypt/

The text is unsafe to cut and paste...

Change-Id: Iab19ddf8182d087e9a4b4d34a9eeabd1d2aa02d6
Reviewed-on: https://go-review.googlesource.com/18104
Reviewed-by: Rob Pike <r@golang.org>

crypto/cipher: improve documentation for AEAD

Give a link to the wikipedia page describing the mechanism and
explain better how to use the same buffer for input and output.

Change-Id: If6dfd6cf9c6dff0517cb715f60a11349dbdd91e0
Reviewed-on: https://go-review.googlesource.com/18103
Reviewed-by: Russ Cox <rsc@golang.org>

misc/cgo/testcarchive: skip test4 on darwin/{arm,arm64}

Fixes #13701.

Change-Id: I9825864d23aeba1971cf5f581e1e59ac4c9b87fd
Reviewed-on: https://go-review.googlesource.com/18090
Reviewed-by: David Crawshaw <crawshaw@golang.org>

net: fix race in TestTCPStress

Fixes #13704.

Change-Id: I7afef5058fa88b0de41213cf46219b684369f47f
Reviewed-on: https://go-review.googlesource.com/18111
Reviewed-by: Ian Lance Taylor <iant@golang.org>

net/internal/socktest: simplify log message format

This change replaces the existing log format separated by commas and
spaces with space-separated one.

Change-Id: I9a4b38669025430190c9a1a6b5c82b862866559d
Reviewed-on: https://go-review.googlesource.com/17999
Reviewed-by: Ian Lance Taylor <iant@golang.org>

net/http: add new HTTP 451 status code, Unavailable For Legal Reasons.

Approved by the IETF.

https://datatracker.ietf.org/doc/draft-ietf-httpbis-legally-restricted-status/

Change-Id: I688597bb5f7ef7c7a9be660a4fcd2ef02d9dc9f4
Reviewed-on: https://go-review.googlesource.com/18112
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Run-TryBot: David Symonds <dsymonds@golang.org>

test: add test that gccgo miscompiled

Change-Id: Ic3f006f86a86de628e14b107f88a5923ea856a58
Reviewed-on: https://go-review.googlesource.com/18093
Reviewed-by: David Symonds <dsymonds@golang.org>

cmd/compile: const name and label name may match

Fixes #13684.

Change-Id: I3977119b6eb1d6b7dc2ea1e7d6656a8f0d421bc1
Reviewed-on: https://go-review.googlesource.com/18060
Run-TryBot: Robert Griesemer <gri@golang.org>
Reviewed-by: Rob Pike <r@golang.org>

net: make use of IPv4 for parsing routing information on windows

In general the package net deals IPv4 addresses as IPv6 IPv4-mapped
addresses internally for the dual stack era, when we need to support
various techniques on IPv4/IPv6 translation.

This change makes windows implementation follow the same pattern which
BSD variants and Linux do.

Updates #13544.

Also fixes an unintentionally formatted line by accident by gofmt.

Change-Id: I4953796e751fd8050c73094468a0d7b0d33f5516
Reviewed-on: https://go-review.googlesource.com/17992
Reviewed-by: Alex Brainman <alex.brainman@gmail.com>

net: adjust TestInterfaceHardwareAddrWithGetmac

CL skips interfaces that are not listed on getmac output.

Fixes #13606

Change-Id: Ic25c9dc95e8eeff4d84b78e99131a4f97020164c
Reviewed-on: https://go-review.googlesource.com/17994
Reviewed-by: Mikio Hara <mikioh.mikioh@gmail.com>
Run-TryBot: Mikio Hara <mikioh.mikioh@gmail.com>

api: fix typo in README

Fixes #13690.

Change-Id: I3b9b993a2e7ecf07bab7d1935d4c83a86bc6ba3a
Reviewed-on: https://go-review.googlesource.com/18054
Reviewed-by: Alex Brainman <alex.brainman@gmail.com>

runtime: for c-archive/c-shared, don't install unnecessary signal handlers

Only install signal handlers for synchronous signals that become
run-time panics.  Set the SA_ONSTACK flag for other signal handlers as
needed.

Fixes #13028.
Update #12465.
Update #13034.
Update #13042.

Change-Id: I28375e70641f60630e10f3c86e24b6e4f8a35cc9
Reviewed-on: https://go-review.googlesource.com/17903
Reviewed-by: Russ Cox <rsc@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

doc: change go1.6 http heading

Reapply golang.org/cl/17918

Change-Id: I0df40585cdd4dae8d365ed9860a81e0cb23f21b9
Reviewed-on: https://go-review.googlesource.com/18032
Reviewed-by: Russ Cox <rsc@golang.org>

runtime: write sigsetstack for Darwin, fix sigaction arg

It turns out that the second argument for sigaction on Darwin has a
different type than the first argument.  The second argument is the user
visible sigaction struct, and does not have the sa_tramp field.

I base this on
  http://www.opensource.apple.com/source/Libc/Libc-1081.1.3/sys/sigaction.c
not to mention actual testing.

While I was at it I removed a useless memclr in setsig, a relic of the C
code.

This CL is Darwin-specific changes.  The tests for this CL are in
https://golang.org/cl/17903 .

Change-Id: I61fe305c72311df6a589b49ad7b6e49b6960ca24
Reviewed-on: https://go-review.googlesource.com/18015
Reviewed-by: David Crawshaw <crawshaw@golang.org>

doc: add mention of debug.SetTraceback

Change-Id: I59829029769ae08c6c54208a1e38a0794868c5db
Reviewed-on: https://go-review.googlesource.com/18045
Reviewed-by: Ian Lance Taylor <iant@golang.org>

testing: use debug.SetTraceback("all") to show all goroutines at test timeout

Fixes #13681.

Change-Id: I308930f4d9200fbe0f09cd08c38392ca1bb0db67
Reviewed-on: https://go-review.googlesource.com/18044
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Reviewed-by: Austin Clements <austin@google.com>

runtime/debug: add SetTraceback

Programs that call panic to crash after detecting a serious problem
may wish to use SetTraceback to force printing of all goroutines first.

Change-Id: Ib23ad9336f405485aabb642ca73f454a14c8baf3
Reviewed-on: https://go-review.googlesource.com/18043
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>

debug/elf: rename Chdr64.Reserved to _

This future-proofs the Chdr64 structure against later versions of ELF
defining this field and declutters the documentation without changing
the layout of the struct.

This structure does not exist in the current release, so this change
is safe.

Change-Id: I239aad7243ddaf063a1f8cd521d8a50b30413281
Reviewed-on: https://go-review.googlesource.com/18028
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>

cmd/go: ignore XML errors after Go <meta> tags

Fixes #13683.

Change-Id: I26afb3ac346beb95624f9032d94a29b5bc7853ef
Reviewed-on: https://go-review.googlesource.com/18051
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>

bytes: document that buffer is reused in Truncate and Reset

Fixes #13671

Change-Id: Ic752de6a3030ff25474717505fa05895054217e7
Reviewed-on: https://go-review.googlesource.com/18029
Reviewed-by: Russ Cox <rsc@golang.org>

net/http: document that Server.TLSNextProto has automatic HTTP/2 also

Copy the same sentence from Transport.TLSNextProto.

Change-Id: Ib67bf054e891a68be8ba466a8c52968363374d16
Reviewed-on: https://go-review.googlesource.com/18031
Reviewed-by: Russ Cox <rsc@golang.org>

cmd/go: do not skip dirs with syntax errors in wildcard matching (like ./...)

Fixes #11407.

Change-Id: If35a8e04a3abf8acf955250c909dde57131b6bb8
Reviewed-on: https://go-review.googlesource.com/17971
Reviewed-by: Ian Lance Taylor <iant@golang.org>

runtime: document stack barrier synchronization rules

Change-Id: I545e53561f37bceabd26d814d272cecc3ff19847
Reviewed-on: https://go-review.googlesource.com/18024
Reviewed-by: Russ Cox <rsc@golang.org>

runtime: require the stack barrier lock to traceback cgo and libcalls

Currently, if sigprof determines that the G is in user code (not cgo
or libcall code), it will only traceback the G stack if it can acquire
the stack barrier lock. However, it has no such restriction if the G
is in cgo or libcall code. Because cgo calls count as syscalls, stack
scanning and stack barrier installation can occur during a cgo call,
which means sigprof could attempt to traceback a G in a cgo call while
scanstack is installing stack barriers in that G's stack. As a result,
the following sequence of events can cause the sigprof traceback to
panic with "missed stack barrier":

1. M1: G1 performs a Cgo call (which, on Windows, is any system call,
   which could explain why this is easier to reproduce on Windows).

2. M1: The Cgo call puts G1 into _Gsyscall state.

3. M2: GC starts a scan of G1's stack. It puts G1 in to _Gscansyscall
   and acquires the stack barrier lock.

4. M3: A profiling signal comes in. On Windows this is a global
   (though I don't think this matters), so the runtime stops M1 and
   calls sigprof for G1.

5. M3: sigprof fails to acquire the stack barrier lock (because the
   GC's stack scan holds it).

6. M3: sigprof observes that G1 is in a Cgo call, so it calls
   gentraceback on G1 with its Cgo transition point.

7. M3: gentraceback on G1 grabs the currently empty g.stkbar slice.

8. M2: GC finishes scanning G1's stack and installing stack barriers.

9. M3: gentraceback encounters one of the just-installed stack
   barriers and panics.

This commit fixes this by only allowing cgo tracebacks if sigprof can
acquire the stack barrier lock, just like in the regular user
traceback case.

For good measure, we put the same constraint on libcall tracebacks.
This case is probably already safe because, unlike cgo calls, libcalls
leave the G in _Grunning and prevent reaching a safe point, so
scanstack cannot run during a libcall. However, this also means that
sigprof will always acquire the stack barrier lock without contention,
so there's no cost to adding this constraint to libcall tracebacks.

Fixes #12528. For 1.5.3 (will require some backporting).

Change-Id: Ia5a4b8e3d66b23b02ffcd54c6315c81055c0cec2
Reviewed-on: https://go-review.googlesource.com/18023
Run-TryBot: Austin Clements <austin@google.com>
Reviewed-by: Russ Cox <rsc@golang.org>

runtime: prevent race between setNextBarrierPC and sigprof

Currently, setNextBarrierPC manipulates the stack barriers without
acquiring the stack barrier lock. This is mostly okay because
setNextBarrierPC also runs synchronously on the G and prevents safe
points, but this doesn't prevent a sigprof from occurring during a
setNextBarrierPC and performing a traceback.

Given that setNextBarrierPC simply sets one entry in the stack barrier
array, this is almost certainly safe in reality. However, given that
this depends on a subtle argument, which may not hold in the future,
and that setNextBarrierPC almost never happens, making it nowhere near
performance-critical, we can simply acquire the stack barrier lock and
be sure that the synchronization will work.

Updates #12528. For 1.5.3.

Change-Id: Ife696e10d969f190157eb1cbe762a2de2ebce079
Reviewed-on: https://go-review.googlesource.com/18022
Run-TryBot: Austin Clements <austin@google.com>
Reviewed-by: Russ Cox <rsc@golang.org>

doc: fix typo in go1.6.html

Change-Id: I7405cf6f65bccbb07a27f2dc2e3802cab591e296
Reviewed-on: https://go-review.googlesource.com/18030
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>

cmd/vet: add some checks for invalid pointer passing using cgo

Update #12416.

Change-Id: I21d97cbe211ccc8048e5a78ea4d89664f4d195ba
Reviewed-on: https://go-review.googlesource.com/17041
Reviewed-by: Russ Cox <rsc@golang.org>

test: fix linkmain test

Change-Id: Ie8ec4cfc68abef51e52090a75245f96af874c74a
Reviewed-on: https://go-review.googlesource.com/18000
Reviewed-by: Ian Lance Taylor <iant@golang.org>

go/types: add link to tutorial

Change-Id: Ic4f4bc7ea7478908716b951815280e394c55310b
Reviewed-on: https://go-review.googlesource.com/17975
Reviewed-by: Robert Griesemer <gri@golang.org>

api: create go1.6.txt

Change-Id: If2b30ab412d6799c8be01eb007462d6b58660ece
Reviewed-on: https://go-review.googlesource.com/18014
Reviewed-by: Chris Broadfoot <cbro@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

src: remove reference to since-deleted dist -s flag

(Found while debugging release problems with go1.6beta1)

Updates #12002

Change-Id: Iec197a754205e7fd28be154f27f17f3315886364
Reviewed-on: https://go-review.googlesource.com/18011
Reviewed-by: Chris Broadfoot <cbro@golang.org>

doc: change go1.6 http heading

Change-Id: Iae05082530891175e9c86da244e610bc92759561
Reviewed-on: https://go-review.googlesource.com/17918
Reviewed-by: Chris Broadfoot <cbro@golang.org>

doc: Americanise spelling of wilful

Fixes #13660

Change-Id: I05bcb4efcb865192a1ef6756e9dccef83505934c
Reviewed-on: https://go-review.googlesource.com/17990
Reviewed-by: Chris Broadfoot <cbro@golang.org>

net/http: document ResponseWriter and Handler more; add test

Update docs on ResponseWriter and Handler around concurrency.

Also add a test.

The Handler docs were old and used "object" a lot. It was also too
ServeMux-centric.

Fixes #13050
Updates #13659 (new issue found in http2 while writing the test)

Change-Id: I25f53d5fa54f1c9d579d3d0f191bf3d94b1a251b
Reviewed-on: https://go-review.googlesource.com/17982
Reviewed-by: Russ Cox <rsc@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>

cmd/link: reject non-package main toplevel.a file, remove dead code

The test for non-package main top-level inputs is done while parsing
the export data. Issue #13468 happened because we were not parsing
the export data when using compiler-generated archives
(that is, when using go tool compile -pack).

Fix this by parsing the export data even for archives.

However, that turns up a different problem: the export data check
reports (one assumes spurious) skew errors now, because it has
not been run since Go 1.2.
(Go 1.3 was the first release to use go tool compile -pack.)

Since the code hasn't run since Go 1.2, it can't be that important.
Since it doesn't work today, just delete it.

Figuring out how to make this code work with Robert's export
format was one of the largest remaining TODOs for that format.
Now we don't have to.

Fixes #13468 and makes the world a better place.

Change-Id: I40a4b284cf140d49d48b714bd80762d6889acdb9
Reviewed-on: https://go-review.googlesource.com/17976
Reviewed-by: Robert Griesemer <gri@golang.org>

cmd/go: for go get -insecure, do not fall back to HTTP for non-200 responses

Since we allow non-200 responses from HTTPS in normal operation,
it seems odd to reject them in -insecure operation.

Fixes #13037 (again).

Change-Id: Ie232f7544ab192addfad407525888db6b967befe
Reviewed-on: https://go-review.googlesource.com/17945
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>

cmd/go: fix processing of HTTPS 404 without -insecure

The change here is to move the closeBody call into the if block.
The logging adjustments are just arranging to tell the truth:
in particular if we're not in insecure mode and we get a non-200
error then we do not actually ignore the response
(except as caused by closing the body incorrectly).

As the comment below the change indicates, it is intentional that
we process non-200 pages. The code does process them, because
the if err != nil || status != 200 block does not return.
But that block does close the body, which depending on timing
can apparently poison the later read from the body.

See #13037's initial report:

$ go get -v bosun.org/cmd/bosun/cache
Fetching https://bosun.org/cmd/bosun/cache?go-get=1
ignoring https fetch with status code 404
Parsing meta tags from https://bosun.org/cmd/bosun/cache?go-get=1 (status code 404)
import "bosun.org/cmd/bosun/cache": parsing bosun.org/cmd/bosun/cache: http: read on closed response body
package bosun.org/cmd/bosun/cache: unrecognized import path "bosun.org/cmd/bosun/cache"

The log print about ignoring the https fetch is not strictly true,
since the next thing that happened was parsing the body of that fetch.
But the read on the closed response body failed during parsing.

Moving the closeBody to happen only when we're about to discard the
result and start over (that is, only in -insecure mode) fixes the parse.

At least it should fix the parse. I can't seem to break the parse anymore,
because of #13648 (close not barring future reads anymore),
but this way is clearly better than the old way. If nothing else the old code
closed the body twice when err != nil and -insecure was not given.

Fixes #13037.

Change-Id: Idf57eceb6d5518341a2f7f75eb8f8ab27ed4e0b4
Reviewed-on: https://go-review.googlesource.com/17944
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>

Revert "cmd/go: allow omitted user name in git ssh repo syntax"

This caused #13657.
Reverting fixes #13657.

I was trying to be helpful by fixing #12313,
but I don't need the fix myself.
Will leave for someone with more motivation.

This reverts commit 3e9f0636709de7cd86a08785f2e64487580bf5ea.

Change-Id: Ifc78a6196f23e0f58e3b9ad7340e207a2d5de0a6
Reviewed-on: https://go-review.googlesource.com/17977
Reviewed-by: Russ Cox <rsc@golang.org>

net/http: don't accept invalid bytes in server request headers

Fixes #11207

Change-Id: I7f00b638e749fbc7907dc1597347ea426367d13e
Reviewed-on: https://go-review.googlesource.com/17980
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Russ Cox <rsc@golang.org>

net/http: be more consistent about Request.Method "" vs "GET"

Patch from Russ.

No bug identified, but I didn't search exhaustively. The new code is
easier to read.

Fixes #13621

Change-Id: Ifda936e4101116fa254ead950b5fe06adb14e977
Reviewed-on: https://go-review.googlesource.com/17981
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Russ Cox <rsc@golang.org>

cmd/go: fix git submodule fetch

Thanks to @toxeus on GitHub for the test case.

Fixes #12612.

Change-Id: I0c32fbe5044f3552053460a5347c062568093dff
Reviewed-on: https://go-review.googlesource.com/17974
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>

net/http: update bundled http2, add tests reading response Body after Close

Updates to golang.org/x/net/http2 git rev 28273ec9 for
https://golang.org/cl/17937

Fixes #13648

Change-Id: I27c77524b2e4a172c5f8be08f6fbb0f2e2e4b200
Reviewed-on: https://go-review.googlesource.com/17938
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>

cmd/go: runtime and runtime/internal packages depend on runtime/internal/sys

Fixes #13655.

Change-Id: I764019aecdd59743baa436b7339499e6c2126268
Reviewed-on: https://go-review.googlesource.com/17916
Run-TryBot: Ian Lance Taylor <iant@golang.org>
Reviewed-by: Russ Cox <rsc@golang.org>

cmd/go: respect umask when creating executables

Also update many call sites where I forgot that the permission
argument is going to be masked by umask.

Fixes #12692.

Change-Id: I52b315b06236122ca020950447863fa396b68abd
Reviewed-on: https://go-review.googlesource.com/17950
Reviewed-by: Ian Lance Taylor <iant@golang.org>

net: fix the build even harder

Should fix nacl.

Follow-up to
https://golang.org/cl/17936 (fix race) and
https://golang.org/cl/17914 (fix build) for
https://golang.org/cl/16953 (broke the build)

Third time's a charm.

Change-Id: I23930d5cff4235209546952ce2231f165ab5bf8a
Reviewed-on: https://go-review.googlesource.com/17939
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>

runtime: disable gdb test when a GOROOT_FINAL move is pending

Fixes #13577.

Change-Id: I0bb8157d6210b0c7c09380c2163b7d7349495732
Reviewed-on: https://go-review.googlesource.com/17970
Reviewed-by: Ian Lance Taylor <iant@golang.org>

net/http: updated bundled http2 to finish trailer support

This updates the bundled copy of x/net/http2 to git rev d2ecd08
for https://golang.org/cl/17912 (http2: send client trailers)
and enables the final Trailer test for http2.

Fixes #13557

Change-Id: Iaa15552b82bf7a2cb01b7787a2e1ec5ee680a9d3
Reviewed-on: https://go-review.googlesource.com/17935
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
Reviewed-by: Russ Cox <rsc@golang.org>

net: include both ipv4 and ipv6 netsh output in TestInterfacesWithNetsh

Also include test for interface state (up or down).

Updates #13606

Change-Id: I03538d65525ddd9c2d0254761861c2df7fc5bd5a
Reviewed-on: https://go-review.googlesource.com/17850
Reviewed-by: Russ Cox <rsc@golang.org>
Reviewed-by: Mikio Hara <mikioh.mikioh@gmail.com>
Run-TryBot: Russ Cox <rsc@golang.org>

net: fix race in test

Fixes race builders, broken in https://golang.org/cl/16953

Change-Id: Id61171672b69d0ca412de4b44bf2c598fe557906
Reviewed-on: https://go-review.googlesource.com/17936
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
Reviewed-by: Russ Cox <rsc@golang.org>

cmd/go: allow ~ as first char of path element in general git import paths

This makes go get gitserver/~user/repo.git/foo work.

Fixes #9193.

Change-Id: I8c9d4096903288f7f0e82d6ed1aa78bf038fb81a
Reviewed-on: https://go-review.googlesource.com/17952
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>

cmd/go: don't be clever about mtime precision in test

This doesn't happen enough in the tests to be worth debugging.
Empirically, I expect this to add 5 seconds to the overall 'go test -short cmd/go'
on systems with precise file systems, and nothing on systems without them
(like my Mac).

Fixes #12205.

Change-Id: I0a17cb37bdedcfc0f921c5ee658737f1698c153b
Reviewed-on: https://go-review.googlesource.com/17953
Reviewed-by: Ian Lance Taylor <iant@golang.org>

cmd/go: document that package documentation is ignored

Fixes #11801.

Change-Id: I2caeac7fdddc7f29015d6db8d4b3e296c8b9c423
Reviewed-on: https://go-review.googlesource.com/17954
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>

net/http: document a few ServeMux behaviors

Fixes #13639.
Fixes #11757.

Change-Id: Iecf9ebcd652c23c96477305a41082e5b63b41d83
Reviewed-on: https://go-review.googlesource.com/17955
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>

cmd/go: allow omitted user name in git ssh repo syntax

No test because the code has no test.

Fixes #12313.

Change-Id: I2cfd0a0422c0cd76f0371c2d3bbbdf5bb3b3f1eb
Reviewed-on: https://go-review.googlesource.com/17951
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>

cmd/go: document that tests run in the source directory

Fixes #13538.

Change-Id: I621bbe2befe838d16d3664d7a5e30d5d7cceae33
Reviewed-on: https://go-review.googlesource.com/17949
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>

cmd/go: document where get writes, and not vendor directories

Fixes #12260.

Change-Id: I95c27aad6de8064b9a205d4ee507bce75926f16d
Reviewed-on: https://go-review.googlesource.com/17948
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>

cmd/go: document that go generate accepts all build flags

Fixes #12544.

Change-Id: I5e2fd1fbb21816e9f6fb022e2664484a71093b04
Reviewed-on: https://go-review.googlesource.com/17947
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>

os/signal: document signal handling

This is an attempt to document the current state of signal handling.
It's not intended to describe the best way to handle signals.  Future
changes to signal handling should update these docs as appropriate.

update #9896.

Change-Id: I3c50af5cc641357b57dfe90ae1c7883a7e1ec059
Reviewed-on: https://go-review.googlesource.com/17877
Reviewed-by: Russ Cox <rsc@golang.org>

cmd/go: document that -p applies to test binary execution

Fixes #11521.

Change-Id: I73615b881df4a0d5e2f5bc5059359d150ca8c105
Reviewed-on: https://go-review.googlesource.com/17946
Reviewed-by: Joe Shaw <joe@joeshaw.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>

cmd/pprof: fix scaling of "gigabyte" unit

Fixes #13654

Change-Id: Id2ce32c52efcfdbd66630725d62d2ca6bf0916d5
Reviewed-on: https://go-review.googlesource.com/17934
Reviewed-by: Russ Cox <rsc@golang.org>

os,internal/syscall/windows: use ReadFile/MultiByteToWideChar to read from console

Fixes #6303

Change-Id: Ib2cd15ac6106ef8e6b975943db8efc8d8ab21052
Reviewed-on: https://go-review.googlesource.com/4310
Reviewed-by: Russ Cox <rsc@golang.org>