]> Cypherpunks repositories - gostls13.git/log
gostls13.git
7 years ago[dev.boringcrypto] misc/boring: update VERSION
Russ Cox [Fri, 22 Sep 2017 15:59:40 +0000 (11:59 -0400)]
[dev.boringcrypto] misc/boring: update VERSION

Change-Id: I805422f3bc4a8a64e55b7453da25c9d1e18f063f
Reviewed-on: https://go-review.googlesource.com/65394
Run-TryBot: Russ Cox <rsc@golang.org>
Reviewed-by: Russ Cox <rsc@golang.org>
7 years ago[dev.boringcrypto] crypto/tls/fipsonly: new package to force FIPS-allowed TLS settings
Russ Cox [Wed, 20 Sep 2017 17:50:35 +0000 (13:50 -0400)]
[dev.boringcrypto] crypto/tls/fipsonly: new package to force FIPS-allowed TLS settings

Change-Id: I3268cab2de8aed9e2424e9c3bc7667083bc5e1ce
Reviewed-on: https://go-review.googlesource.com/65250
Run-TryBot: Russ Cox <rsc@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Adam Langley <agl@golang.org>
7 years ago[dev.boringcrypto] crypto/internal/boring: fix finalizer-induced crashes
Russ Cox [Wed, 20 Sep 2017 03:50:50 +0000 (23:50 -0400)]
[dev.boringcrypto] crypto/internal/boring: fix finalizer-induced crashes

All the finalizer-enabled C wrappers must be careful to use
runtime.KeepAlive to ensure the C wrapper object (a Go object)
lives through the end of every C call using state that the
wrapper's finalizer would free.

This CL makes the wrappers appropriately careful.

The test proves that this is the bug I was chasing in a
separate real program, and that the KeepAlives fix it.
I did not write a test of every possible operation.

Change-Id: I627007e480f16adf8396e7f796b54e5525d9ea80
Reviewed-on: https://go-review.googlesource.com/64870
Run-TryBot: Russ Cox <rsc@golang.org>
Reviewed-by: Adam Langley <agl@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
7 years ago[dev.boringcrypto] cmd/go: exclude SysoFiles when using -msan
Russ Cox [Thu, 14 Sep 2017 15:11:30 +0000 (11:11 -0400)]
[dev.boringcrypto] cmd/go: exclude SysoFiles when using -msan

There's no way for a *.syso file to be compiled to work both in
normal mode and in msan mode. Assume they are compiled in
normal mode and drop them in msan mode.

Packages with syso files currently fail in -msan mode because
the syso file calls out to a routine like memcmp which then
falsely reports uninitialized memory. After this CL, they will fail
in -msan with link errors, because the syso will not be used.
But then it will at least be possible for package authors to write
fallback code in the package that avoids the syso in -msan mode,
so that the package with the syso can at least run in both modes.
Without this CL, that's not possible.

See #21884.

Change-Id: I77340614c4711325032484e65fa9c3f8332741d5
Reviewed-on: https://go-review.googlesource.com/63917
Reviewed-by: Adam Langley <agl@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
7 years ago[dev.boringcrypto] crypto/internal/boring: fall back to standard crypto when using...
Russ Cox [Thu, 14 Sep 2017 15:10:23 +0000 (11:10 -0400)]
[dev.boringcrypto] crypto/internal/boring: fall back to standard crypto when using -msan

The syso is not compiled with -fsanitize=memory, so don't try to use it.
Otherwise the first time it calls out to memcmp, memcmp complains
that it is being asked to compare uninitialized memory.

Change-Id: I85ab707cfbe64eded8e110d4d6b40d1b75f50541
Reviewed-on: https://go-review.googlesource.com/63916
Reviewed-by: Adam Langley <agl@golang.org>
7 years ago[dev.boringcrypto] crypto/aes: panic on invalid dst, src overlap
Russ Cox [Thu, 14 Sep 2017 03:10:49 +0000 (23:10 -0400)]
[dev.boringcrypto] crypto/aes: panic on invalid dst, src overlap

I've now debugged multiple mysterious "inability to communicate"
bugs that manifest as a silent unexplained authentication failure but are
really crypto.AEAD.Open being invoked with badly aligned buffers.
In #21624 I suggested using a panic as the consequence of bad alignment,
so that this kind of failure is loud and clearly different from, say, a
corrupted or invalid message signature. Adding the panic here made
my failure very easy to track down, once I realized that was the problem.
I don't want to debug another one of these.

Also using this CL as an experiment to get data about the impact of
maybe applying this change more broadly in the master branch.

Change-Id: Id2e2d8e980439f8acacac985fc2674f7c96c5032
Reviewed-on: https://go-review.googlesource.com/63915
Reviewed-by: Adam Langley <agl@golang.org>
7 years ago[dev.boringcrypto] crypto/rsa: fix boring GenerateKey to set non-nil Precomputed...
Russ Cox [Thu, 14 Sep 2017 03:07:38 +0000 (23:07 -0400)]
[dev.boringcrypto] crypto/rsa: fix boring GenerateKey to set non-nil Precomputed.CRTValues

This matches the standard GenerateKey and more importantly Precompute,
so that if you generate a key and then store it, read it back, call Precompute
on the new copy, and then do reflect.DeepEqual on the two copies, they
will match. Before this CL, the original key had CRTValues == nil and the
reconstituted key has CRTValues != nil (but len(CRTValues) == 0).

Change-Id: I1ddc64342a50a1b65a48d827e4d564f1faab1945
Reviewed-on: https://go-review.googlesource.com/63914
Reviewed-by: Adam Langley <agl@golang.org>
7 years ago[dev.boringcrypto] crypto/internal/boring: fix detection of tests to allow *.test...
Russ Cox [Thu, 14 Sep 2017 03:06:54 +0000 (23:06 -0400)]
[dev.boringcrypto] crypto/internal/boring: fix detection of tests to allow *.test and *_test

When using the go command, test binaries end in .test,
but when using Bazel, test binaries conventionally end in _test.

Change-Id: Ic4cac8722fd93ae316169f87b321f68e0b71f0c3
Reviewed-on: https://go-review.googlesource.com/63913
Reviewed-by: Adam Langley <agl@golang.org>
7 years ago[dev.boringcrypto] crypto/rsa: add test for, fix observable reads from custom randomness
Russ Cox [Tue, 12 Sep 2017 20:19:24 +0000 (16:19 -0400)]
[dev.boringcrypto] crypto/rsa: add test for, fix observable reads from custom randomness

In routines like GenerateKey, where bits from the randomness source have a
visible effect on the output, we bypass BoringCrypto if given a non-standard
randomness source (and also assert that this happens only during tests).

In the decryption paths, the randomness source is only for blinding and has
no effect on the output, so we unconditionally invoke BoringCrypto, letting it
use its own randomness source as it sees fit. This in turn lets us verify that
the non-BoringCrypto decryption function is never called, not even in tests.

Unfortunately, while the randomness source has no visible effect on the
decrypt operation, the decrypt operation does have a visible effect on
the randomness source. If decryption doesn't use the randomness source,
and it's a synthetic stream, then a future operation will read a different
position in the stream and may produce different output. This happens
in tests more often than you'd hope.

To keep behavior of those future operations unchanged while still
ensuring that the original decrypt is never called, this CL adds a
simulation of the blinding preparation, to discard the right amount
from the random source before invoking BoringCrypto.

Change-Id: If2f87b856c811b59b536187c93efa99a97721419
Reviewed-on: https://go-review.googlesource.com/63912
Reviewed-by: Adam Langley <agl@golang.org>
7 years ago[dev.boringcrypto] crypto/hmac: add test for Write/Sum after Sum
Russ Cox [Thu, 7 Sep 2017 16:16:04 +0000 (12:16 -0400)]
[dev.boringcrypto] crypto/hmac: add test for Write/Sum after Sum

This is documented to work (in hash.Hash's definition)
and existing code assumes it works. Add a test.

Change-Id: I63546f3b2d66222683a4f268a4eaff835fd836fe
Reviewed-on: https://go-review.googlesource.com/63911
Reviewed-by: Adam Langley <agl@golang.org>
7 years ago[dev.boringcrypto] crypto/internal/boring: allow hmac operations after Sum
Russ Cox [Thu, 7 Sep 2017 16:14:31 +0000 (12:14 -0400)]
[dev.boringcrypto] crypto/internal/boring: allow hmac operations after Sum

hmac.New returns a hash.Hash, which defines Sum as:

// Sum appends the current hash to b and returns the resulting slice.
// It does not change the underlying hash state.
Sum(b []byte) []byte

I've now seen two different pieces of code that make
use of the assumption that Sum has no effect on the
internal state, so make it so.

Test in next CL in stack, so that it can be cherry-picked
to master.

Change-Id: Iad84ab3e2cc12dbecef25c3fc8f2662d157b0d0b
Reviewed-on: https://go-review.googlesource.com/63910
Reviewed-by: Adam Langley <agl@golang.org>
7 years ago[dev.boringcrypto] crypto/internal/boring: handle RSA verification of short signatures
Russ Cox [Thu, 7 Sep 2017 03:31:41 +0000 (23:31 -0400)]
[dev.boringcrypto] crypto/internal/boring: handle RSA verification of short signatures

The standard Go crypto/rsa allows signatures to be shorter
than the RSA modulus and assumes leading zeros.
BoringCrypto does not, so supply the leading zeros explicitly.

This fixes the golang.org/x/crypto/openpgp tests.

Change-Id: Ic8b18d6beb0e02992a0474f5fdb2b73ccf7098cf
Reviewed-on: https://go-review.googlesource.com/62170
Reviewed-by: Adam Langley <agl@golang.org>
7 years ago[dev.boringcrypto] cmd/compile: refine BoringCrypto kludge
Russ Cox [Thu, 7 Sep 2017 02:01:50 +0000 (22:01 -0400)]
[dev.boringcrypto] cmd/compile: refine BoringCrypto kludge

Did not consider these fields being embedded or adopted
into structs defined in other packages, but that's possible too.
Refine the import path check to account for that.

Fixes 'go test -short golang.org/x/crypto/ssh' but also
adds a new test in internal/boring for the same problem.

Change-Id: Ied2d04fe2b0ac3b0a34f07bc8dfc50fc203abb9f
Reviewed-on: https://go-review.googlesource.com/62152
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Reviewed-by: Adam Langley <agl@golang.org>
7 years ago[dev.boringcrypto] cmd/compile: hide new boring fields from reflection
Russ Cox [Wed, 30 Aug 2017 14:10:15 +0000 (10:10 -0400)]
[dev.boringcrypto] cmd/compile: hide new boring fields from reflection

This is terrible but much simpler, cleaner, and more effective
than all the alternatives I have come up with.

Lots of code assumes that reflect.DeepEqual is meaningful
on rsa.PublicKey etc, because previously they consisted only of
exported meaningful fields.

Worse, there exists code that assumes asn1.Marshal can be
passed an rsa.PublicKey, because that struct has historically
matched exactly the form that would be needed to produce
the official ASN.1 DER encoding of an RSA public key.

Instead of tracking down and fixing all of that code
(and probably more), we can limit the BoringCrypto-induced
damage by ensliting the compiler to hide the new field
from reflection. Then nothing can get at it and nothing can
be disrupted by it.

Kill two birds with one cannon ball.

I'm very sorry.

Change-Id: I0ca4d6047c7e98f880cbb81904048c1952e278cc
Reviewed-on: https://go-review.googlesource.com/60271
Reviewed-by: Adam Langley <agl@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
7 years ago[dev.boringcrypto] crypto/hmac: test empty key
Russ Cox [Mon, 28 Aug 2017 18:27:03 +0000 (14:27 -0400)]
[dev.boringcrypto] crypto/hmac: test empty key

This happens in the scrypt and pbkdf unit tests.

Change-Id: I1eda944d7c01d28c7a6dd9f428f5fdd1cbd58939
Reviewed-on: https://go-review.googlesource.com/59771
Reviewed-by: Adam Langley <agl@golang.org>
7 years ago[dev.boringcrypto] crypto/internal/boring: fix NewHMAC with empty key
Russ Cox [Mon, 28 Aug 2017 18:26:31 +0000 (14:26 -0400)]
[dev.boringcrypto] crypto/internal/boring: fix NewHMAC with empty key

Test is in a separate CL for easier cherry-picking to master branch.

Change-Id: Ia4a9032892d2896332010fe18a3216f8c4a58d1c
Reviewed-on: https://go-review.googlesource.com/59770
Reviewed-by: Adam Langley <agl@golang.org>
7 years ago[dev.boringcrypto] crypto/internal/cipherhw: fix AESGCMSupport for BoringCrypto
Russ Cox [Sat, 26 Aug 2017 01:24:15 +0000 (21:24 -0400)]
[dev.boringcrypto] crypto/internal/cipherhw: fix AESGCMSupport for BoringCrypto

The override was not commented and was in the wrong file.

Change-Id: I739db561acff6d91b0f3559c8bb45437f11c0b04
Reviewed-on: https://go-review.googlesource.com/59250
Reviewed-by: Adam Langley <agl@golang.org>
7 years ago[dev.boringcrypto] misc/boring: release packaging
Russ Cox [Tue, 22 Aug 2017 22:22:12 +0000 (18:22 -0400)]
[dev.boringcrypto] misc/boring: release packaging

Add scripts and docs for packaging releases.

Change-Id: I0682c92bbb2e229d2636762e49fe73513852d351
Reviewed-on: https://go-review.googlesource.com/57890
Reviewed-by: Adam Langley <agl@golang.org>
7 years ago[dev.boringcrypto] crypto/internal/boring: disable for android & non-cgo builds
Russ Cox [Wed, 23 Aug 2017 23:12:54 +0000 (19:12 -0400)]
[dev.boringcrypto] crypto/internal/boring: disable for android & non-cgo builds

Change-Id: Ia4458090118c4391a73cf1ae65bc8d187f03eca0
Reviewed-on: https://go-review.googlesource.com/59051
Reviewed-by: Ian Lance Taylor <iant@golang.org>
7 years ago[dev.boringcrypto] crypto/internal/boring: clear "executable stack" bit from syso
Russ Cox [Thu, 24 Aug 2017 23:24:43 +0000 (19:24 -0400)]
[dev.boringcrypto] crypto/internal/boring: clear "executable stack" bit from syso

Change-Id: Ie9dd13f3ae78a423a231f47e746a38f96768b93c
Reviewed-on: https://go-review.googlesource.com/58830
Reviewed-by: Ian Lance Taylor <iant@golang.org>
7 years ago[dev.boringcrypto] cmd/link: allow internal linking for crypto/internal/boring
Russ Cox [Wed, 23 Aug 2017 02:52:15 +0000 (22:52 -0400)]
[dev.boringcrypto] cmd/link: allow internal linking for crypto/internal/boring

Change-Id: I5b122ad23f75296dab8cec89a4e50dcca7fa9b3f
Reviewed-on: https://go-review.googlesource.com/57944
Run-TryBot: Russ Cox <rsc@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
7 years ago[dev.boringcrypto] cmd/link: work around DWARF symbol bug
Russ Cox [Wed, 23 Aug 2017 02:50:27 +0000 (22:50 -0400)]
[dev.boringcrypto] cmd/link: work around DWARF symbol bug

The DWARF code is mishandling the case when the host object files
define multiple (distinct) symbols with the same name. They are mapped
to the same DWARF debug symbol, which then appears on the dwarfp
list multiple times, which then breaks the code that processes the list.
Detect duplicates and skip them, because that's trivial, instead of fixing
the underlying problem.

See #21566.

Change-Id: Ib5a34c891d7c15f4c7bb6239d8f31a1ec767b8bc
Reviewed-on: https://go-review.googlesource.com/57943
Run-TryBot: Russ Cox <rsc@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
7 years ago[dev.boringcrypto] crypto/rsa: use BoringCrypto
Russ Cox [Thu, 3 Aug 2017 20:46:15 +0000 (16:46 -0400)]
[dev.boringcrypto] crypto/rsa: use BoringCrypto

Change-Id: Ibb92f0f8cb487f4d179b069e588e1cb266599384
Reviewed-on: https://go-review.googlesource.com/55479
Run-TryBot: Russ Cox <rsc@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Adam Langley <agl@golang.org>
7 years ago[dev.boringcrypto] crypto/ecdsa: use unsafe.Pointer instead of atomic.Value
Russ Cox [Fri, 18 Aug 2017 13:11:40 +0000 (09:11 -0400)]
[dev.boringcrypto] crypto/ecdsa: use unsafe.Pointer instead of atomic.Value

Using atomic.Value causes vet errors in code copying
PublicKey or PrivateKey structures. I don't think the errors
are accurate, but it's easier to work around them than
to change vet or change atomic.Value.

See #21504.

Change-Id: I3a3435c1fc664cc5166c81674f6f7c58dab35f21
Reviewed-on: https://go-review.googlesource.com/56671
Run-TryBot: Russ Cox <rsc@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Reviewed-by: Adam Langley <agl@golang.org>
7 years ago[dev.boringcrypto] crypto/ecdsa: use BoringCrypto
Russ Cox [Thu, 3 Aug 2017 19:49:33 +0000 (15:49 -0400)]
[dev.boringcrypto] crypto/ecdsa: use BoringCrypto

Change-Id: I108e0a527bddd673b16582d206e0697341d0a0ea
Reviewed-on: https://go-review.googlesource.com/55478
Run-TryBot: Russ Cox <rsc@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Adam Langley <agl@golang.org>
7 years ago[dev.boringcrypto] crypto/tls: use TLS-specific AES-GCM mode if available
Russ Cox [Tue, 15 Aug 2017 23:23:26 +0000 (19:23 -0400)]
[dev.boringcrypto] crypto/tls: use TLS-specific AES-GCM mode if available

Change-Id: Ide00c40c0ca8d486f3bd8968e1d301c8b0ed6d05
Reviewed-on: https://go-review.googlesource.com/56011
Run-TryBot: Russ Cox <rsc@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Adam Langley <agl@golang.org>
7 years ago[dev.boringcrypto] crypto/aes: implement TLS-specific AES-GCM mode from BoringCrypto
Russ Cox [Thu, 3 Aug 2017 15:59:56 +0000 (11:59 -0400)]
[dev.boringcrypto] crypto/aes: implement TLS-specific AES-GCM mode from BoringCrypto

Change-Id: I8407310e7d00eafe9208879228dbf4ac3d26a907
Reviewed-on: https://go-review.googlesource.com/55477
Run-TryBot: Russ Cox <rsc@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Adam Langley <agl@golang.org>
7 years ago[dev.boringcrypto] crypto/aes: use BoringCrypto
Russ Cox [Thu, 3 Aug 2017 15:08:27 +0000 (11:08 -0400)]
[dev.boringcrypto] crypto/aes: use BoringCrypto

Change-Id: If83fdeac31f65aba818bbc7edd2f215b16814021
Reviewed-on: https://go-review.googlesource.com/55476
Run-TryBot: Russ Cox <rsc@golang.org>
Reviewed-by: Adam Langley <agl@golang.org>
7 years ago[dev.boringcrypto] crypto/hmac: use BoringCrypto
Russ Cox [Thu, 3 Aug 2017 04:02:43 +0000 (00:02 -0400)]
[dev.boringcrypto] crypto/hmac: use BoringCrypto

Change-Id: Id4019d601c615b4835b0337d82be3d508292810e
Reviewed-on: https://go-review.googlesource.com/55475
Run-TryBot: Russ Cox <rsc@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Adam Langley <agl@golang.org>
7 years ago[dev.boringcrypto] crypto/sha1,sha256,sha512: use BoringCrypto
Russ Cox [Thu, 3 Aug 2017 03:36:53 +0000 (23:36 -0400)]
[dev.boringcrypto] crypto/sha1,sha256,sha512: use BoringCrypto

Change-Id: I80a764971b41f75c3b699797bfed71f509e3407d
Reviewed-on: https://go-review.googlesource.com/55474
Run-TryBot: Russ Cox <rsc@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Adam Langley <agl@golang.org>
7 years ago[dev.boringcrypto] runtime/race: move TestRaceIssue5567 from sha1 to crc32
Russ Cox [Thu, 17 Aug 2017 15:29:19 +0000 (11:29 -0400)]
[dev.boringcrypto] runtime/race: move TestRaceIssue5567 from sha1 to crc32

If we substitute a SHA1 implementation where the entirety of the
reading of the buffer is done in assembly (or C called from cgo),
then the race detector cannot observe the race.

Change to crc32 with a fake polynomial, in the hope that it will
always be handled by Go code, not optimized assembly or cgo calls.

Change-Id: I34e90b14ede6bc220ef686f6aef16b8e464b5cde
Reviewed-on: https://go-review.googlesource.com/56510
Run-TryBot: Russ Cox <rsc@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
7 years ago[dev.boringcrypto] crypto/rand: use BoringCrypto
Russ Cox [Thu, 3 Aug 2017 03:25:07 +0000 (23:25 -0400)]
[dev.boringcrypto] crypto/rand: use BoringCrypto

Change-Id: Ie630eff90f7fee9b359683930aec2daf96c1bdfe
Reviewed-on: https://go-review.googlesource.com/55473
Run-TryBot: Russ Cox <rsc@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Adam Langley <agl@golang.org>
7 years ago[dev.boringcrypto] crypto/internal/boring: add initial BoringCrypto access
Russ Cox [Thu, 3 Aug 2017 03:14:57 +0000 (23:14 -0400)]
[dev.boringcrypto] crypto/internal/boring: add initial BoringCrypto access

Right now the package doesn't do anything useful, but it will.
This CL is about the machinery for building goboringcrypto_linux_amd64.syso
and then running the self-test and checking FIPS_mode from Go init.

Change-Id: I4ec0f5efaa88ccfb506b9818d24a7f1cbcc5a7d6
Reviewed-on: https://go-review.googlesource.com/55472
Run-TryBot: Russ Cox <rsc@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Adam Langley <agl@golang.org>
7 years ago[dev.boringcrypto] add README.boringcrypto.md, update VERSION
Russ Cox [Mon, 14 Aug 2017 15:44:27 +0000 (11:44 -0400)]
[dev.boringcrypto] add README.boringcrypto.md, update VERSION

Change-Id: I415fcc23b62666d78aed3ddc7d2731127c810be3
Reviewed-on: https://go-review.googlesource.com/55471
Run-TryBot: Russ Cox <rsc@golang.org>
Reviewed-by: Adam Langley <agl@golang.org>
7 years ago[dev.boringcrypto] cmd/link: implement R_X86_64_PC64 relocations
Russ Cox [Fri, 11 Aug 2017 20:22:10 +0000 (16:22 -0400)]
[dev.boringcrypto] cmd/link: implement R_X86_64_PC64 relocations

Change-Id: I1d7bd5cff7350a4e0f78b8efc8406e79c74732d1
Reviewed-on: https://go-review.googlesource.com/55370
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Reviewed-on: https://go-review.googlesource.com/55470
Run-TryBot: Russ Cox <rsc@golang.org>

7 years ago[release-branch.go1.9] go1.9rc2 go1.9rc2
Chris Broadfoot [Mon, 7 Aug 2017 19:23:06 +0000 (12:23 -0700)]
[release-branch.go1.9] go1.9rc2

Change-Id: If95cec0ec7e32cdb450818c4c55e2d03b847ab65
Reviewed-on: https://go-review.googlesource.com/53630
Reviewed-by: Austin Clements <austin@google.com>
7 years ago[release-branch.go1.9] all: merge master into release-branch.go1.9
Chris Broadfoot [Mon, 7 Aug 2017 17:28:03 +0000 (10:28 -0700)]
[release-branch.go1.9] all: merge master into release-branch.go1.9

579120323f runtime: mapassign_* should use typedmemmove to update keys
380525598c all: remove some manual hyphenation
f096b5b340 runtime: mark activeModules nosplit/nowritebarrier
3e3da54633 math/bits: fix example for OnesCount64
9b1e7cf2ac math/bits: add examples for OnesCount functions
b01db023b1 misc/cgo/testsanitizers: also skip tsan11/tsan12 when using GCC
a279b53a18 reflect: document how DeepEqual handles cycles
909f409a8d doc: mention handling of moved GOROOT in 1.9 release notes
58ad0176ca doc: use better wording to explain type-aware completion
92dac21d29 doc: replace paid with commercial
9bb98e02de doc/1.9: add CL 43712, ReverseProxy of HTTP/2 trailers to the release notes.
78d74fc2cd doc: clarify that Gogland is for paid IntelliJ platform IDEs
5495047223 doc/1.9: fix broken html link in CL 53030/53210
890e0e862f doc: fix bad link in go1.9 release notes
be596f049a doc/1.9: fix stray html in CL 53030
0173631d53 encoding/binary: add examples for varint functions
ac0ccf3cd2 doc/1.9: add CL 36696 for crypto/x509 to the release notes
cc402c2c4d doc: hide blog content for golang.google.cn
f396fa4285 internal/poll: don't add non-sockets to runtime poller
664cd26c89 cmd/vet: don't exit with failure on type checking error
a8730cd93a doc: hide video and share if being served from CN
b63db76c4a testsanitizers: check that tsan program runs, skip tsan10 on gcc
193eda7291 time: skip ZoneAbbr test in timezones with no abbreviation
6f08c935a9 cmd/go: show examples with empty output in go test -list
f20944de78 cmd/compile: set/unset base register for better assembly print
623e2c4603 runtime: map bitmap and spans during heap initialization
780249eed4 runtime: fall back to small mmaps if we fail to grow reservation
31b2c4cc25 .github: add .md extension to SUPPORT file
ac29f30dbb plugin: mention that there are known bugs with plugins
45a4609c0a cmd/dist: skip moved GOROOT on Go's Windows builders when not sharding tests
e157fac02d test: add README
835dfef939 runtime/pprof: prevent a deadlock that SIGPROF might create on mips{,le}
df91b8044d doc: list editor options by name, not plugin name
3d9475c04b doc: cleanup editor page
b9661a14ea doc: add Atom to editor guide
ee392ac10c cmd/compile: consider exported flag in namedata

Change-Id: I3a48493e8c05d97cb3b61635503ef0ccd646e5cb

7 years agoruntime: mapassign_* should use typedmemmove to update keys
Keith Randall [Sat, 5 Aug 2017 16:58:41 +0000 (09:58 -0700)]
runtime: mapassign_* should use typedmemmove to update keys

We need to make sure that when the key contains a pointer, we use
a write barrier to update the key.

Also mapdelete_* should use typedmemclr.

Fixes #21297

Change-Id: I63dc90bec1cb909c2c6e08676c9ec853d736cdf8
Reviewed-on: https://go-review.googlesource.com/53414
Run-TryBot: Keith Randall <khr@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Austin Clements <austin@google.com>
Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
7 years agoall: remove some manual hyphenation
Josh Bleecher Snyder [Sun, 6 Aug 2017 13:17:42 +0000 (06:17 -0700)]
all: remove some manual hyphenation

Manual hyphenation doesn't work well when text gets reflown,
for example by godoc.

There are a few other manual hyphenations in the tree,
but they are in local comments or comments for unexported functions.

Change-Id: I17c9b1fee1def650da48903b3aae2fa1e1119a65
Reviewed-on: https://go-review.googlesource.com/53510
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
7 years agoruntime: mark activeModules nosplit/nowritebarrier
Ian Lance Taylor [Fri, 4 Aug 2017 17:36:40 +0000 (10:36 -0700)]
runtime: mark activeModules nosplit/nowritebarrier

The activeModules function is called by the cgo pointer checking code,
which is called by the write barrier (when GODEBUG=cgocheck=2), and as
such must be nosplit/nowritebarrier.

Fixes #21306

Change-Id: I57f2124f14de7f3872b2de9532abab15df95d45a
Reviewed-on: https://go-review.googlesource.com/53352
Reviewed-by: Austin Clements <austin@google.com>
7 years agomath/bits: fix example for OnesCount64
Francesc Campoy Flores [Sat, 5 Aug 2017 00:03:33 +0000 (17:03 -0700)]
math/bits: fix example for OnesCount64

Erroneously called OnesCount instead of OnesCount64

Change-Id: Ie877e43f213253e45d31f64931c4a15915849586
Reviewed-on: https://go-review.googlesource.com/53410
Reviewed-by: Chris Broadfoot <cbro@golang.org>
7 years agomath/bits: add examples for OnesCount functions
Francesc Campoy [Fri, 4 Aug 2017 22:41:57 +0000 (15:41 -0700)]
math/bits: add examples for OnesCount functions

Change-Id: Ie673f9665825a40281c2584d478ba1260f725856
Reviewed-on: https://go-review.googlesource.com/53357
Run-TryBot: Chris Broadfoot <cbro@golang.org>
Reviewed-by: Chris Broadfoot <cbro@golang.org>
7 years agomisc/cgo/testsanitizers: also skip tsan11/tsan12 when using GCC
Ian Lance Taylor [Fri, 4 Aug 2017 22:33:48 +0000 (15:33 -0700)]
misc/cgo/testsanitizers: also skip tsan11/tsan12 when using GCC

Updates #21196

Change-Id: I307cacc963448b90a23f633bec15498ba7bf1937
Reviewed-on: https://go-review.googlesource.com/53356
Run-TryBot: Ian Lance Taylor <iant@golang.org>
Reviewed-by: Bryan Mills <bcmills@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

7 years agoreflect: document how DeepEqual handles cycles
Ian Lance Taylor [Thu, 3 Aug 2017 00:52:06 +0000 (17:52 -0700)]
reflect: document how DeepEqual handles cycles

Fixes #20428

Change-Id: Ia450e615728efd4ccb6e42117b547cac162f13a3
Reviewed-on: https://go-review.googlesource.com/52931
Reviewed-by: Joe Tsai <thebrokentoaster@gmail.com>
7 years agodoc: mention handling of moved GOROOT in 1.9 release notes
Ian Lance Taylor [Fri, 4 Aug 2017 18:23:25 +0000 (11:23 -0700)]
doc: mention handling of moved GOROOT in 1.9 release notes

Updates #20587

Change-Id: Ia131b9a4dc4986950d9ecbfcbd6b026ade234fc0
Reviewed-on: https://go-review.googlesource.com/53370
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
7 years agodoc: use better wording to explain type-aware completion
Jaana Burcu Dogan [Fri, 4 Aug 2017 19:59:55 +0000 (12:59 -0700)]
doc: use better wording to explain type-aware completion

Some editors can filter the autocompletion suggestions based on
whether the code will compile once autocompleted. Explain this
feature with better wording.

Change-Id: I29e4b0396878f18c79208915402c0a209a813b04
Reviewed-on: https://go-review.googlesource.com/53355
Reviewed-by: Florin Patan <florinpatan@gmail.com>
Reviewed-by: Andrew Bonventre <andybons@golang.org>
7 years agodoc: replace paid with commercial
Jaana Burcu Dogan [Fri, 4 Aug 2017 19:44:18 +0000 (12:44 -0700)]
doc: replace paid with commercial

Change-Id: I3e6de4da0648f61e254c7597f316df3e5791321a
Reviewed-on: https://go-review.googlesource.com/53354
Reviewed-by: Andrew Bonventre <andybons@golang.org>
7 years agodoc/1.9: add CL 43712, ReverseProxy of HTTP/2 trailers to the release notes.
Tristan Colgate [Fri, 4 Aug 2017 07:27:56 +0000 (08:27 +0100)]
doc/1.9: add CL 43712, ReverseProxy of HTTP/2 trailers to the release notes.

Add https://go-review.googlesource.com/c/43712,
"net/http/httputil: ReverseProxy should pass on unannounced Trailers"
to the relase notes.

Fixes #21307

Change-Id: I52c126987a5d0abc4153c0e71b535529c46cd457
Reviewed-on: https://go-review.googlesource.com/53290
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
7 years agodoc: clarify that Gogland is for paid IntelliJ platform IDEs
Jaana Burcu Dogan [Fri, 4 Aug 2017 16:47:06 +0000 (09:47 -0700)]
doc: clarify that Gogland is for paid IntelliJ platform IDEs

Fixes #21213.

Change-Id: I7b8a84de92bbd1d3f78f8a9612f3af8cd092cb94
Reviewed-on: https://go-review.googlesource.com/53351
Reviewed-by: Andrew Bonventre <andybons@golang.org>
7 years agodoc/1.9: fix broken html link in CL 53030/53210
Dmitry Savintsev [Fri, 4 Aug 2017 08:12:21 +0000 (10:12 +0200)]
doc/1.9: fix broken html link in CL 53030/53210

Change-Id: I7176becd10ad84cbfc3fb9427e190028626e5baf
Reviewed-on: https://go-review.googlesource.com/53291
Reviewed-by: Ian Lance Taylor <iant@golang.org>
7 years agodoc: fix bad link in go1.9 release notes
Alberto Donizetti [Fri, 4 Aug 2017 10:13:33 +0000 (12:13 +0200)]
doc: fix bad link in go1.9 release notes

Change-Id: I64ba37428f5cc560f0f20fe039feaecf5fcda93e
Reviewed-on: https://go-review.googlesource.com/53330
Reviewed-by: Ian Lance Taylor <iant@golang.org>
7 years agodoc/1.9: fix stray html in CL 53030
Dmitry Savintsev [Thu, 3 Aug 2017 21:25:59 +0000 (23:25 +0200)]
doc/1.9: fix stray html in CL 53030

Change-Id: Ib4102b1e2a8863712f725c4d1e37fdbe3dfe3c07
Reviewed-on: https://go-review.googlesource.com/53210
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
7 years agoencoding/binary: add examples for varint functions
Axel Wagner [Sat, 15 Jul 2017 17:43:22 +0000 (11:43 -0600)]
encoding/binary: add examples for varint functions

Change-Id: I191f6e46b452fadde9f641140445d843b0c7d534
Reviewed-on: https://go-review.googlesource.com/48604
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
7 years agodoc/1.9: add CL 36696 for crypto/x509 to the release notes
Dmitry Savintsev [Thu, 3 Aug 2017 14:51:32 +0000 (16:51 +0200)]
doc/1.9: add CL 36696 for crypto/x509 to the release notes

add https://go-review.googlesource.com/c/36696
"crypto/x509: ignore CN if SAN extension present"
to the release notes.

Fixes #21289

Change-Id: Ifa184d3816806a8da3c67b68476c923329acf13e
Reviewed-on: https://go-review.googlesource.com/53030
Reviewed-by: Ian Lance Taylor <iant@golang.org>
7 years agodoc: hide blog content for golang.google.cn
Andrew Bonventre [Thu, 3 Aug 2017 16:33:26 +0000 (12:33 -0400)]
doc: hide blog content for golang.google.cn

/blog redirects to blog.golang.org (currently blocked in China)
unless there is a local checkout of golang.org/x/blog, which is
not possible on App Engine Classic.

Change-Id: Ia695e663c9bebcc6c3bedea324c630299eaad4dc
Reviewed-on: https://go-review.googlesource.com/53051
Reviewed-by: Chris Broadfoot <cbro@golang.org>
7 years agointernal/poll: don't add non-sockets to runtime poller
Ian Lance Taylor [Tue, 1 Aug 2017 23:05:17 +0000 (16:05 -0700)]
internal/poll: don't add non-sockets to runtime poller

Updates #21172

Change-Id: I0fec6e645328bbc85f3e47f4f71dd8d1d68c75ab
Reviewed-on: https://go-review.googlesource.com/52551
Reviewed-by: Alex Brainman <alex.brainman@gmail.com>
7 years agocmd/vet: don't exit with failure on type checking error
Ian Lance Taylor [Wed, 2 Aug 2017 19:06:58 +0000 (12:06 -0700)]
cmd/vet: don't exit with failure on type checking error

The vet tool only reports a type checking error when invoked with -v.
Don't let that by itself cause vet to exit with an error exit status.

Updates #21188

Change-Id: I172c13d46c35d49e229e96e833683d8c82a77de7
Reviewed-on: https://go-review.googlesource.com/52851
Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
Reviewed-by: Rob Pike <r@golang.org>
7 years agodoc: hide video and share if being served from CN
Andrew Bonventre [Wed, 2 Aug 2017 20:31:46 +0000 (16:31 -0400)]
doc: hide video and share if being served from CN

In the case where requests are coming from mainland China, hide
links to locations that are blocked and functionality that is
not permitted.

Additionally, some very small cleanup of the JS.

This change requires https://go-review.googlesource.com/c/52873

Change-Id: I7fc68748e629dbe5b966d6bf117e7f7b546966eb
Reviewed-on: https://go-review.googlesource.com/52872
Reviewed-by: Chris Broadfoot <cbro@golang.org>
7 years agotestsanitizers: check that tsan program runs, skip tsan10 on gcc
Ian Lance Taylor [Wed, 2 Aug 2017 17:49:20 +0000 (10:49 -0700)]
testsanitizers: check that tsan program runs, skip tsan10 on gcc

Check not only that a tsan program can be built, but also that it runs.
This fails with some installations of GCC 7.

Skip the tsan10 program when using GCC, as it reportedly hangs.

This is a patch to help people build 1.9; we may be able to do a
better fix for 1.10.

Updates #21196

Change-Id: Icd1ffbd018dc65a97ff45cab1264b9b0c7fa0ab2
Reviewed-on: https://go-review.googlesource.com/52790
Run-TryBot: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Bryan Mills <bcmills@google.com>
Reviewed-by: Bryan Mills <bcmills@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

7 years agotime: skip ZoneAbbr test in timezones with no abbreviation
Alberto Donizetti [Tue, 1 Aug 2017 15:41:51 +0000 (17:41 +0200)]
time: skip ZoneAbbr test in timezones with no abbreviation

The testZoneAbbr assumes that

  Parse(RFC1123, t1.Format(RFC1123))

will always succeed. This is not true because Format will fall back to
the numeric zone (ex. -07) for timezones with no abbreviation, but
Parse won't accept the numeric zone when the layout specifies 'MST'
(an abbreviation).

Skip the zone abbreviation test in timezones with no abbreviation.

Fixes #21183

Change-Id: If04691cc23ae1075d8a953733024e17f5a7646de
Reviewed-on: https://go-review.googlesource.com/52430
Run-TryBot: Alberto Donizetti <alb.donizetti@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
7 years agocmd/go: show examples with empty output in go test -list
Seiji Takahashi [Sun, 30 Jul 2017 12:00:09 +0000 (21:00 +0900)]
cmd/go: show examples with empty output in go test -list

Fixes #21205

Change-Id: I81b001eb42cbf2a5d5b7b82eb63548b22f501be5
Reviewed-on: https://go-review.googlesource.com/52110
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>

7 years agocmd/compile: set/unset base register for better assembly print
Cherry Zhang [Tue, 18 Jul 2017 12:35:00 +0000 (08:35 -0400)]
cmd/compile: set/unset base register for better assembly print

For address of an auto or arg, on all non-x86 architectures
the assembler backend encodes the actual SP offset in the
instruction but leaves the offset in Prog unchanged. When the
assembly is printed in compile -S, it shows an offset
relative to pseudo FP/SP with an actual hardware SP base
register (e.g. R13 on ARM). This is confusing. Unset the
base register if it is indeed SP, so the assembly output is
consistent. If the base register isn't SP, it should be an
error and the error output contains the actual base register.

For address loading instructions, the base register isn't set
in the compiler on non-x86 architectures. Set it. Normally it
is SP and will be unset in the change mentioned above for
printing. If it is not, it will be an error and the error
output contains the actual base register.

No change in generated binary, only printed assembly. Passes
"go build -a -toolexec 'toolstash -cmp' std cmd" on all
architectures.

Fixes #21064.

Change-Id: Ifafe8d5f9b437efbe824b63b3cbc2f5f6cdc1fd5
Reviewed-on: https://go-review.googlesource.com/49432
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: David Chase <drchase@google.com>
7 years ago[release-branch.go1.9] runtime: map bitmap and spans during heap initialization
Austin Clements [Fri, 28 Jul 2017 21:44:09 +0000 (17:44 -0400)]
[release-branch.go1.9] runtime: map bitmap and spans during heap initialization

We lazily map the bitmap and spans areas as the heap grows. However,
right now we're very slightly too lazy. Specifically, the following
can happen on 32-bit:

1. mallocinit fails to allocate any heap arena, so
   arena_used == arena_alloc == arena_end == bitmap.

2. There's less than 256MB between the end of the bitmap mapping and
   the next mapping.

3. On the first allocation, mheap.sysAlloc sees that there's not
   enough room in [arena_alloc, arena_end) because there's no room at
   all. It gets a 256MB mapping from somewhere *lower* in the address
   space than arena_used and sets arena_alloc and arena_end to this
   hole.

4. Since the new arena_alloc is lower than arena_used, mheap.sysAlloc
   doesn't bother to call mheap.setArenaUsed, so we still don't have a
   bitmap mapping or a spans array mapping.

5. mheap.grow, which called mheap.sysAlloc, attempts to fill in the
   spans array and crashes.

Fix this by mapping the metadata regions for the initial arena_used
when the heap is initialized, rather than trying to wait for an
allocation. This maintains the intended invariant that the structures
are always mapped for [arena_start, arena_used).

Fixes #21044.

Cherry-pick of CL 51714. Fixes #21234.

Change-Id: I4422375a6e234b9f979d22135fc63ae3395946b0
Reviewed-on: https://go-review.googlesource.com/52191
Run-TryBot: Austin Clements <austin@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
7 years ago[release-branch.go1.9] runtime: fall back to small mmaps if we fail to grow reservation
Austin Clements [Fri, 28 Jul 2017 22:06:03 +0000 (18:06 -0400)]
[release-branch.go1.9] runtime: fall back to small mmaps if we fail to grow reservation

Right now, if it's possible to grow the arena reservation but
mheap.sysAlloc fails to get 256MB more of memory, it simply fails.
However, on 32-bit we have a fallback path that uses much smaller
mmaps that could take in this situation, but fail to.

This commit fixes mheap.sysAlloc to use a common failure path in case
it can't grow the reservation. On 32-bit, this path includes the
fallback.

Ideally, mheap.sysAlloc would attempt smaller reservation growths
first, but taking the fallback path is a simple change for Go 1.9.

Updates #21044 (fixes one of two issues).

Cherry-pick of CL 51713. Updates #21234.

Change-Id: I1e0035ffba986c3551479d5742809e43da5e7c73
Reviewed-on: https://go-review.googlesource.com/52190
Run-TryBot: Austin Clements <austin@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
7 years agoruntime: map bitmap and spans during heap initialization
Austin Clements [Fri, 28 Jul 2017 21:44:09 +0000 (17:44 -0400)]
runtime: map bitmap and spans during heap initialization

We lazily map the bitmap and spans areas as the heap grows. However,
right now we're very slightly too lazy. Specifically, the following
can happen on 32-bit:

1. mallocinit fails to allocate any heap arena, so
   arena_used == arena_alloc == arena_end == bitmap.

2. There's less than 256MB between the end of the bitmap mapping and
   the next mapping.

3. On the first allocation, mheap.sysAlloc sees that there's not
   enough room in [arena_alloc, arena_end) because there's no room at
   all. It gets a 256MB mapping from somewhere *lower* in the address
   space than arena_used and sets arena_alloc and arena_end to this
   hole.

4. Since the new arena_alloc is lower than arena_used, mheap.sysAlloc
   doesn't bother to call mheap.setArenaUsed, so we still don't have a
   bitmap mapping or a spans array mapping.

5. mheap.grow, which called mheap.sysAlloc, attempts to fill in the
   spans array and crashes.

Fix this by mapping the metadata regions for the initial arena_used
when the heap is initialized, rather than trying to wait for an
allocation. This maintains the intended invariant that the structures
are always mapped for [arena_start, arena_used).

Fixes #21044.

Change-Id: I4422375a6e234b9f979d22135fc63ae3395946b0
Reviewed-on: https://go-review.googlesource.com/51714
Run-TryBot: Austin Clements <austin@google.com>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
7 years agoruntime: fall back to small mmaps if we fail to grow reservation
Austin Clements [Fri, 28 Jul 2017 22:06:03 +0000 (18:06 -0400)]
runtime: fall back to small mmaps if we fail to grow reservation

Right now, if it's possible to grow the arena reservation but
mheap.sysAlloc fails to get 256MB more of memory, it simply fails.
However, on 32-bit we have a fallback path that uses much smaller
mmaps that could take in this situation, but fail to.

This commit fixes mheap.sysAlloc to use a common failure path in case
it can't grow the reservation. On 32-bit, this path includes the
fallback.

Ideally, mheap.sysAlloc would attempt smaller reservation growths
first, but taking the fallback path is a simple change for Go 1.9.

Updates #21044 (fixes one of two issues).

Change-Id: I1e0035ffba986c3551479d5742809e43da5e7c73
Reviewed-on: https://go-review.googlesource.com/51713
Run-TryBot: Austin Clements <austin@google.com>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

7 years ago.github: add .md extension to SUPPORT file
Gustav Westling [Sun, 30 Jul 2017 20:14:22 +0000 (22:14 +0200)]
.github: add .md extension to SUPPORT file

This makes GitHub render the markdown file automatically
on their web UI.

SUPPORT.md is the recommended file name according to the GitHub
documentation:
https://help.github.com/articles/adding-support-resources-to-your-project/

Fixes #21223

Change-Id: I9f9b9daced9c29a16850e8c446656f353f50b1ae
Reviewed-on: https://go-review.googlesource.com/52013
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
7 years agoplugin: mention that there are known bugs with plugins
Ian Lance Taylor [Wed, 26 Jul 2017 16:46:58 +0000 (09:46 -0700)]
plugin: mention that there are known bugs with plugins

Change-Id: I9e63661cac2bebc41d7aa3cd80e1920eec22b894
Reviewed-on: https://go-review.googlesource.com/51250
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Reviewed-by: David Crawshaw <crawshaw@golang.org>
7 years ago[release-branch.go1.9] cmd/dist: skip moved GOROOT on Go's Windows builders when...
Brad Fitzpatrick [Thu, 27 Jul 2017 04:33:18 +0000 (04:33 +0000)]
[release-branch.go1.9] cmd/dist: skip moved GOROOT on Go's Windows builders when not sharding tests

Change-Id: I0bcae339624e7d61037d9ea0885b7bd07491bbb6
Reviewed-on: https://go-review.googlesource.com/51430
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
(cherry picked from commit 4833e920c1d7f6b23458e6ff3c73951fcf754219)
Reviewed-on: https://go-review.googlesource.com/51450
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
7 years agocmd/dist: skip moved GOROOT on Go's Windows builders when not sharding tests
Brad Fitzpatrick [Thu, 27 Jul 2017 04:33:18 +0000 (04:33 +0000)]
cmd/dist: skip moved GOROOT on Go's Windows builders when not sharding tests

Change-Id: I0bcae339624e7d61037d9ea0885b7bd07491bbb6
Reviewed-on: https://go-review.googlesource.com/51430
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
7 years agotest: add README
Josh Bleecher Snyder [Mon, 24 Jul 2017 17:17:15 +0000 (10:17 -0700)]
test: add README

Updates #21034

Change-Id: I951fb48ab3b9ed54d225c11879db8f09048a36a3
Reviewed-on: https://go-review.googlesource.com/50950
Reviewed-by: Rob Pike <r@golang.org>
7 years agoruntime/pprof: prevent a deadlock that SIGPROF might create on mips{,le}
Vladimir Stefanovic [Thu, 4 May 2017 14:45:29 +0000 (16:45 +0200)]
runtime/pprof: prevent a deadlock that SIGPROF might create on mips{,le}

64bit atomics on mips/mipsle are implemented using spinlocks. If SIGPROF
is received while the program is in the critical section, it will try to
write the sample using the same spinlock, creating a deadloop.
Prevent it by creating a counter of SIGPROFs during atomic64 and
postpone writing the sample(s) until called from elsewhere, with
pc set to _LostSIGPROFDuringAtomic64.

Added a test case, per Cherry's suggestion. Works around #20146.

Change-Id: Icff504180bae4ee83d78b19c0d9d6a80097087f9
Reviewed-on: https://go-review.googlesource.com/42652
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
7 years agodoc: list editor options by name, not plugin name
Jaana Burcu Dogan [Tue, 25 Jul 2017 20:03:48 +0000 (13:03 -0700)]
doc: list editor options by name, not plugin name

So the users can recognize their option by their editor's name.

Fixes #20398.

Change-Id: Id314d4dbe26f40231a479b179620d7e66512b506
Reviewed-on: https://go-review.googlesource.com/51114
Reviewed-by: Chris Broadfoot <cbro@golang.org>
7 years agodoc: cleanup editor page
Zac Bergquist [Tue, 25 Jul 2017 18:08:47 +0000 (14:08 -0400)]
doc: cleanup editor page

Fix some UI issues introduced with CL50952:
- increase header colspan to account for additional column
- remove ':' character from footnotes

Change-Id: I56f59b8e4b2852612b3c6c7c0dfe99125dd8b57b
Reviewed-on: https://go-review.googlesource.com/51113
Reviewed-by: Jaana Burcu Dogan <jbd@google.com>
7 years agodoc: add Atom to editor guide
Zac Bergquist [Mon, 24 Jul 2017 23:05:43 +0000 (19:05 -0400)]
doc: add Atom to editor guide

Fixes #20569

Change-Id: I752a49ed50c1567f8db7112859ac073f37dd77dc
Reviewed-on: https://go-review.googlesource.com/50952
Reviewed-by: Jaana Burcu Dogan <jbd@google.com>
7 years ago[release-branch.go1.9] go1.9rc1 go1.9rc1
Chris Broadfoot [Mon, 24 Jul 2017 17:28:28 +0000 (10:28 -0700)]
[release-branch.go1.9] go1.9rc1

Change-Id: I6251fc128990f9494a0b037cabb24ada5eabb6b5
Reviewed-on: https://go-review.googlesource.com/50951
Run-TryBot: Chris Broadfoot <cbro@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Reviewed-by: Chris Broadfoot <cbro@golang.org>
7 years ago[release-branch.go1.9] cmd/compile: consider exported flag in namedata
Ian Lance Taylor [Fri, 21 Jul 2017 23:37:40 +0000 (16:37 -0700)]
[release-branch.go1.9] cmd/compile: consider exported flag in namedata

It is possible to have an unexported name with a nil package,
for an embedded field whose type is a pointer to an unexported type.
We must encode that fact in the type..namedata symbol name,
to avoid incorrectly merging an unexported name with an exported name.

Fixes #21120

Change-Id: I2e3879d77fa15c05ad92e0bf8e55f74082db5111
Reviewed-on: https://go-review.googlesource.com/50710
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: David Crawshaw <crawshaw@golang.org>
Reviewed-on: https://go-review.googlesource.com/50970
Reviewed-by: Chris Broadfoot <cbro@golang.org>
7 years agocmd/compile: consider exported flag in namedata
Ian Lance Taylor [Fri, 21 Jul 2017 23:37:40 +0000 (16:37 -0700)]
cmd/compile: consider exported flag in namedata

It is possible to have an unexported name with a nil package,
for an embedded field whose type is a pointer to an unexported type.
We must encode that fact in the type..namedata symbol name,
to avoid incorrectly merging an unexported name with an exported name.

Fixes #21120

Change-Id: I2e3879d77fa15c05ad92e0bf8e55f74082db5111
Reviewed-on: https://go-review.googlesource.com/50710
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: David Crawshaw <crawshaw@golang.org>
7 years agocmd/link: use full link, not compile, to test for -no-?pie
Ian Lance Taylor [Mon, 24 Jul 2017 17:13:25 +0000 (10:13 -0700)]
cmd/link: use full link, not compile, to test for -no-?pie

This avoids an error from clang when using -nopie during compilation,
and permits us to check that the entire build succeeds.

Updates #21042

Change-Id: I2e6c7d5c97a85c223ed3288622bbb58ce33b8774
Reviewed-on: https://go-review.googlesource.com/50874
Run-TryBot: Ian Lance Taylor <iant@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
7 years ago.github: recommend 'go bug' when filing an issue
Josh Bleecher Snyder [Mon, 24 Jul 2017 16:54:13 +0000 (09:54 -0700)]
.github: recommend 'go bug' when filing an issue

It was released with 1.8, so it should be widely available.

Change-Id: I7dcf205009b528071ea63f99cb5da0db183df341
Reviewed-on: https://go-review.googlesource.com/50932
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
7 years ago.github: add SUPPORT file
Josh Bleecher Snyder [Mon, 24 Jul 2017 15:18:06 +0000 (08:18 -0700)]
.github: add SUPPORT file

GitHub recently added support for a SUPPORT file:

https://github.com/blog/2400-support-file-support

This SUPPORT file is a very lightly edited copy
of the wiki entry on asking questions:

https://github.com/golang/go/wiki/Questions

Change-Id: Ic1eb74d985ea30862defb99750fb42da84e492de
Reviewed-on: https://go-review.googlesource.com/50930
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
7 years agoencoding/json: ignore embedded fields of pointers to unexported non-structs
Joe Tsai [Fri, 21 Jul 2017 23:53:54 +0000 (16:53 -0700)]
encoding/json: ignore embedded fields of pointers to unexported non-structs

https://golang.org/cl/33773 fixes the JSON marshaler to avoid serializing
embedded fields on unexported types of non-struct types. However, Go allows
embedding pointer to types, so the check for whether the field is a non-struct
type must first dereference the pointer to get at the underlying type.

Furthermore, due to a edge-case in the behavior of StructField.PkgPath not
being a reliable indicator of whether the field is unexported (see #21122),
we use our own logic to determine whether the field is exported or not.

The logic in this CL may be simplified depending on what happens in #21122.

Fixes #21121
Updates #21122

Change-Id: I8dfd1cdfac8a87950df294a566fb96dfd04fd749
Reviewed-on: https://go-review.googlesource.com/50711
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

7 years agosync: delete a sentence from the Map docs
Brad Fitzpatrick [Fri, 21 Jul 2017 17:32:32 +0000 (17:32 +0000)]
sync: delete a sentence from the Map docs

From Josh's comments on https://golang.org/cl/50310

Once I removed the "from the Go standard library" bit, the beginning
wasn't worth keeping. It also wasn't clear whether what it meant by
"cache contention". Processor caches, or user-level caches built with
sync.Map? It didn't seem worth clarifying and didn't convey any useful
information, so deleted.

Change-Id: Id1d76105a3081d0855f6a64540700932bb83d98e
Reviewed-on: https://go-review.googlesource.com/50632
Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
Reviewed-by: Bryan Mills <bcmills@google.com>
7 years agonet/http: document that after Hijack, Request.Body is invalid
Brad Fitzpatrick [Fri, 21 Jul 2017 18:52:38 +0000 (18:52 +0000)]
net/http: document that after Hijack, Request.Body is invalid

We can make it panic with a more explicit and readable error message
during Go 1.10, but document it for now. This has always been the
case; it's not a new rule.

Updates #20933

Change-Id: I53c1fefb47a8f4aae0bb32fa742afa3a2ed20e8a
Reviewed-on: https://go-review.googlesource.com/50634
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Reviewed-by: Emmanuel Odeke <emm.odeke@gmail.com>
7 years agosync: update Map documentation with usage rule of thumb
Michael Stapelberg [Thu, 20 Jul 2017 15:16:36 +0000 (08:16 -0700)]
sync: update Map documentation with usage rule of thumb

As per bcmills’s lightning talk at GopherCon 2017:
https://github.com/gophercon/2017-talks/tree/master/lightningtalks/BryanCMills-AnOverviewOfSyncMap

Change-Id: I12dd0daa608af175d110298780f32c6dc5e1e0a0
Reviewed-on: https://go-review.googlesource.com/50310
Reviewed-by: Bryan Mills <bcmills@google.com>
7 years agodoc: minor fixes to the editor guide
Jaana Burcu Dogan [Fri, 21 Jul 2017 01:16:03 +0000 (18:16 -0700)]
doc: minor fixes to the editor guide

Change-Id: I8f6bcfab27251ef6962306d56e40d306ef85fe60
Reviewed-on: https://go-review.googlesource.com/50472
Reviewed-by: Rob Pike <r@golang.org>
7 years agoruntime: improve out-of-memory message when VirtualAlloc fails
Austin Clements [Tue, 18 Jul 2017 15:14:28 +0000 (11:14 -0400)]
runtime: improve out-of-memory message when VirtualAlloc fails

Fixes #19514.

Change-Id: I93600d5c3d11ecab5a47dd4cd55ed3aea05e221e
Reviewed-on: https://go-review.googlesource.com/49611
Run-TryBot: Austin Clements <austin@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
7 years agonet/http: improve signature of Redirect, NewRequest
Dmitri Shuralyov [Wed, 19 Jul 2017 04:41:13 +0000 (00:41 -0400)]
net/http: improve signature of Redirect, NewRequest

In CL https://golang.org/cl/4893043 (6 years ago), a new package named
"url" was created (it is currently known as "net/url"). During that
change, some identifier name collisions were introduced, and two
parameters in net/http were renamed to "urlStr".

Since that time, Go has continued to put high emphasis on the quality
and readability of the documentation. Sometimes, that means making small
sacrifices in the implementation details of a package to ensure that
the godoc reads better, since that's what the majority of users interact
with. See https://golang.org/s/style#named-result-parameters:

> Clarity of docs is always more important than saving a line or two
> in your function.

I think the "urlStr" parameter name is suboptimal for godoc purposes,
and just "url" would be better.

During the review of https://golang.org/cl/4893043, it was also noted
by @rsc that having to rename parameters named "url" was suboptimal:

> It's unfortunate that naming the package url means
> you can't have a parameter or variable named url.

However, at the time, the name of the url package was still being
decided, and uri was an alternative name under consideration.
The reason urlStr was chosen is because it was a lesser evil
compared to naming the url package uri instead:

> Let's not get hung up on URI vs. URL, but I'd like s/uri/urlStr/ even for just
> that the "i" in "uri" looks very similar to the "l" in "url" in many fonts.

> Please let's go with urlStr instead of uri.

Now that we have the Go 1 compatibility guarantee, the name of the
net/url package is fixed. However, it's possible to improve the
signature of Redirect, NewRequest functions in net/http package
for godoc purposes by creating a package global alias to url.Parse,
and renaming urlStr parameter to url in the exported funcs. This CL
does so.

Updates #21077.

Change-Id: Ibcc10e3825863a663e6ad91b6eb47b1862a299a6
Reviewed-on: https://go-review.googlesource.com/49930
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
7 years agocmd/go, cmd/link: if -no-pie doesn't work, try -nopie
Ian Lance Taylor [Tue, 18 Jul 2017 20:15:05 +0000 (13:15 -0700)]
cmd/go, cmd/link: if -no-pie doesn't work, try -nopie

GCC says -no-pie, clang says -nopie.

Fixes #21042

Change-Id: Iadc83ea7a48ea0debc5064c1ee8da4ebff752044
Reviewed-on: https://go-review.googlesource.com/49710
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
7 years agosync: release m.mu during (*RWMutexMap).Range callbacks in sync_test
Bryan C. Mills [Tue, 9 May 2017 20:12:57 +0000 (16:12 -0400)]
sync: release m.mu during (*RWMutexMap).Range callbacks in sync_test

The mainline sync.Map has allowed mutations within Range callbacks
since https://golang.org/cl/37342. The reference implementations need
to do the same.

This change integrates https://go-review.googlesource.com/c/42956/
from x/sync.

Change-Id: I6b58cf874bb31cd4f6fdb8bfa8278888ed617a5a
Reviewed-on: https://go-review.googlesource.com/42957
Run-TryBot: Bryan Mills <bcmills@google.com>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

7 years agomisc/cgo/testsanitizers: add regression test for change 50251
Bryan C. Mills [Thu, 20 Jul 2017 17:47:57 +0000 (13:47 -0400)]
misc/cgo/testsanitizers: add regression test for change 50251

http://golang.org/cl/50251 fixed a regression under TSAN.
This change adds a minimal reproducer for the observed symptom.

Change-Id: Ib9ad01b458b7fdec14d6c2fe3c243f9c64b3dcf2
Reviewed-on: https://go-review.googlesource.com/50371
Run-TryBot: Bryan Mills <bcmills@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
7 years agocmd/go: fix test when go source tree has POSIX ACL
Jess Frazelle [Thu, 20 Jul 2017 16:29:15 +0000 (12:29 -0400)]
cmd/go: fix test when go source tree has POSIX ACL

Fixes TestGoBuildUmask when the user has a POSIX ACL on the Go source tree.

Fixes #17909.

Change-Id: I5bc19099af8353afd41071258f4f317612b4c8c1
Reviewed-on: https://go-review.googlesource.com/50370
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

7 years agoruntime: use SIGKILL if SIGQUIT is blocked; skip tests that need SIGQUIT
Austin Clements [Thu, 20 Jul 2017 15:12:37 +0000 (11:12 -0400)]
runtime: use SIGKILL if SIGQUIT is blocked; skip tests that need SIGQUIT

The runtime tests may be invoked from a parent that has SIGQUIT
blocked. For example, Java invokes subprocesses this way. In this
situation, TestCrashDumpsAllThreads and TestPanicSystemstack will fail
because they depend on SIGQUIT to get tracebacks, and any subprocess
test that times out will fail to kill the subprocess.

Fix this by detecting if SIGQUIT is blocked and, if so, skipping tests
that depend on it and using SIGKILL to kill timed-out subprocesses.

Based on a fix by Carl Henrik Lunde in
https://golang.org/issue/19196#issuecomment-316145733

Fixes #19196.

Change-Id: Ia20bf15b96086487d0ef6b75239dcc260c21714c
Reviewed-on: https://go-review.googlesource.com/50330
Run-TryBot: Austin Clements <austin@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
7 years agoruntime: don't call libc sigaction function in forked child
Ian Lance Taylor [Thu, 20 Jul 2017 06:42:27 +0000 (23:42 -0700)]
runtime: don't call libc sigaction function in forked child

If we are using vfork, and if something (such as TSAN) is intercepting
the sigaction function, then we must call the system call, not the
libc function. Otherwise the intercepted sigaction call in the child
may trash the data structures in the parent.

Change-Id: Id9588bfeaa934f32c920bf829c5839be5cacf243
Reviewed-on: https://go-review.googlesource.com/50251
Reviewed-by: Joe Tsai <thebrokentoaster@gmail.com>
Reviewed-by: Austin Clements <austin@google.com>
7 years agoos: remove duplicate check from windows os.Stat
Johnny Luo [Thu, 20 Jul 2017 07:20:46 +0000 (17:20 +1000)]
os: remove duplicate check from windows os.Stat

Fixes #21075

Change-Id: Idfe5002dfe17943844d9427e27f82ce894b92e80
Reviewed-on: https://go-review.googlesource.com/50270
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

7 years agonet/mail: Fix typo
yansal [Wed, 19 Jul 2017 21:04:36 +0000 (23:04 +0200)]
net/mail: Fix typo

Fixes #21089

Change-Id: Idd65c7185b3e19f33958eb165cb5b09c06db3d56
Reviewed-on: https://go-review.googlesource.com/50110
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
7 years agodoc: add an editor guide
Jaana Burcu Dogan [Wed, 14 Jun 2017 19:41:14 +0000 (12:41 -0700)]
doc: add an editor guide

The Go ecosystem provides many tools to make Go
development more productive and seamless. Document
the availability of the editor plugins and IDEs,
add an overview of feature support and screencasts.

Updates #20398.
Updates #20402.
Updates #20399.
Updates #20401.
Updates #20569.

Change-Id: I0c6cb48eb4e3848807aaad78390493e14f097916
Reviewed-on: https://go-review.googlesource.com/45772
Reviewed-by: Steve Francia <spf@golang.org>
7 years agoruntime: only trace mark assists that do work
Austin Clements [Wed, 19 Jul 2017 18:18:08 +0000 (14:18 -0400)]
runtime: only trace mark assists that do work

Currently we trace mark assists even if they're satisfied entirely by
stealing. This means even if background marking is keeping up with
allocation, we'll still emit a trace event every N bytes of
allocation. The event will be a few microseconds, if that, but they're
frequent enough that, when zoomed out in the trace view, it looks like
all of the time is spent in mark assists even if almost none is.

Change this so we only emit a trace event if the assist actually has
to do assisting. This makes the traces of these events far more
useful.

Change-Id: If4aed1c413b814341ef2fba61d2f10751d00451b
Reviewed-on: https://go-review.googlesource.com/50030
Run-TryBot: Austin Clements <austin@google.com>
Reviewed-by: Rick Hudson <rlh@golang.org>
7 years agocmd/compile: fix unaligned loads/stores to global variables on s390x
Michael Munday [Mon, 17 Jul 2017 11:07:28 +0000 (07:07 -0400)]
cmd/compile: fix unaligned loads/stores to global variables on s390x

Load/store-merging and move optimizations can result in unaligned
memory accesses. This is fine so long as the load/store instruction
used does not take a relative offset. In the SSA rules this means we
must not merge (MOVDaddr (SB)) ops into loads/stores unless we can
guarantee the alignment of the target.

Fixes #21048.

Change-Id: I70f13a62a148d5f0a56e704e8f76e36b4a4226d9
Reviewed-on: https://go-review.googlesource.com/49250
Run-TryBot: Michael Munday <mike.munday@ibm.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
7 years agonet/http: fix parameter name in comment
Ian Lance Taylor [Tue, 18 Jul 2017 20:29:47 +0000 (13:29 -0700)]
net/http: fix parameter name in comment

Fixes #21077

Change-Id: Ic61d7313907f58ff4027fd2eee1ddb8c1656304d
Reviewed-on: https://go-review.googlesource.com/49712
Reviewed-by: Dmitri Shuralyov <shurcool@gmail.com>