]> Cypherpunks repositories - gostls13.git/log
gostls13.git
9 years agoos: skip TestHardLink on Android.
Rahul Chaudhry [Thu, 3 Dec 2015 00:42:17 +0000 (16:42 -0800)]
os: skip TestHardLink on Android.

From Android release M (Marshmallow), hard linking files is blocked
and an attempt to call link() on a file will return EACCES.
- https://code.google.com/p/android-developer-preview/issues/detail?id=3150

Change-Id: Ifdadaa31e3d5ee330553f45db6c001897dc955be
Reviewed-on: https://go-review.googlesource.com/17339
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>

9 years agodoc: correct release date for go1.5.2
Chris Broadfoot [Thu, 3 Dec 2015 00:22:53 +0000 (16:22 -0800)]
doc: correct release date for go1.5.2

Change-Id: I6bee207db7485f96a499f51b2d1346c35e086d41
Reviewed-on: https://go-review.googlesource.com/17337
Reviewed-by: Russ Cox <rsc@golang.org>
9 years agodoc: document go1.5.2
Chris Broadfoot [Tue, 24 Nov 2015 00:07:20 +0000 (16:07 -0800)]
doc: document go1.5.2

Change-Id: Ib6a9e131113523e6b1e5b7604480028b9ffbfa93
Reviewed-on: https://go-review.googlesource.com/17178
Reviewed-by: Andrew Gerrand <adg@golang.org>
9 years agonet/mail: add test index to error messages in a recent test
Alexandre Cesaro [Wed, 2 Dec 2015 22:28:49 +0000 (23:28 +0100)]
net/mail: add test index to error messages in a recent test

Add the test index so it is easier to find which test case failed.

Change-Id: Ic04682651b26b137355950ff0c51bdbdb1d85a9c
Reviewed-on: https://go-review.googlesource.com/17351
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
9 years agocmd/go: allow buildmode=c-shared for android/arm64.
Rahul Chaudhry [Wed, 2 Dec 2015 21:27:44 +0000 (13:27 -0800)]
cmd/go: allow buildmode=c-shared for android/arm64.

Also, enable test misc/cgo/testcshared for android/arm64.
c/17245 and c/17246 provide the missing pieces for making
this test work.

"androidtest.bash" now passes on a Nexus 9 (volantis)
device running Android build "LMY48T".

Change-Id: Icb9fd2d17d97e0f04cb18d0cd91640c80fbd3fb4
Reviewed-on: https://go-review.googlesource.com/17333
Reviewed-by: Ian Lance Taylor <iant@golang.org>
9 years agoruntime: set TLSG_IS_VARIABLE for android/arm64.
Rahul Chaudhry [Thu, 26 Nov 2015 01:35:44 +0000 (17:35 -0800)]
runtime: set TLSG_IS_VARIABLE for android/arm64.

On android, runtime.tls_g is a normal variable.
TLS offset is computed in x_cgo_inittls.

Change-Id: I18bc9a736d5fb2a89d0f798956c754e3c10d10e2
Reviewed-on: https://go-review.googlesource.com/17246
Reviewed-by: David Crawshaw <crawshaw@golang.org>
9 years agoruntime/cgo: define x_cgo_inittls() for android/arm64.
Rahul Chaudhry [Thu, 26 Nov 2015 01:31:57 +0000 (17:31 -0800)]
runtime/cgo: define x_cgo_inittls() for android/arm64.

On android, runtime.tls_g is a normal variable.
TLS offset is computed in x_cgo_inittls.

Change-Id: I64cfd3543040776dcdf73cad8dba54fc6aaf6f35
Reviewed-on: https://go-review.googlesource.com/17245
Reviewed-by: David Crawshaw <crawshaw@golang.org>
9 years agocmd/internal/obj: fix stack barriers in ppc64le shared libs
Michael Hudson-Doyle [Tue, 1 Dec 2015 02:39:41 +0000 (15:39 +1300)]
cmd/internal/obj: fix stack barriers in ppc64le shared libs

runtime.stackBarrier is a strange function: it is only ever "called" by
smashing its address into a LR slot on the stack. Calling it like this
certainly does not adhere to the rule that r12 is set to the global entry point
before calling it and the prologue instrutions that compute r2 from r12 in fact
just corrupt r2, which is bad because the function that stackBarrier returns to
probably uses r2 to access global data.

Fortunately stackBarrier itself does not access any global data and so does not
depend on the value of r2, meaning we can ignore the ABI rules and simply skip
inserting the prologue instructions into this specific function.

Fixes 64bit.go, append.go and fixedbugs/issue13169.go from "cd test; go run
run.go -linkshared".

Change-Id: I606864133a83935899398e2d42edd08a946aab24
Reviewed-on: https://go-review.googlesource.com/17281
Reviewed-by: Austin Clements <austin@google.com>
9 years agocmd/cgo: error, not panic, if not enough arguments to function
Ian Lance Taylor [Wed, 2 Dec 2015 20:07:57 +0000 (12:07 -0800)]
cmd/cgo: error, not panic, if not enough arguments to function

Fixes #13423.

Change-Id: I41bb45790cca36c57a107796f0eca61287acb2a9
Reviewed-on: https://go-review.googlesource.com/17332
Run-TryBot: Ian Lance Taylor <iant@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agocmd/compile: don't allow blank method declarations on builtins
Håvard Haugen [Sun, 15 Nov 2015 22:32:30 +0000 (23:32 +0100)]
cmd/compile: don't allow blank method declarations on builtins

Move test for isblank into addmethod so that most of the type checking
for methods is also performed for blank methods.

Fixes #11366.

Change-Id: I13d554723bf96d906d0b3ff390d7b7c87c1a5020
Reviewed-on: https://go-review.googlesource.com/16866
Reviewed-by: Robert Griesemer <gri@golang.org>
9 years agodatabase/sql: Add DB.SetConnMaxLifetime
INADA Naoki [Tue, 3 Mar 2015 12:27:07 +0000 (21:27 +0900)]
database/sql: Add DB.SetConnMaxLifetime

Long lived connections may make some DB operation difficult.
(e.g. retiring load balanced DB server.)
So SetConnMaxLifetime closes long lived connections.

It can be used to limit maximum idle time, too.
Closing idle connections reduces active connections while application is idle
and avoids connections are closed by server side (cause errBadConn while querying).

fixes #9851

Change-Id: I2e8e824219c1bee7f4b885d38ed96d11b7202b56
Reviewed-on: https://go-review.googlesource.com/6580
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
9 years agonet/mail: use base64 encoding when needed in Address.String()
Alexandre Cesaro [Tue, 20 Oct 2015 15:30:21 +0000 (17:30 +0200)]
net/mail: use base64 encoding when needed in Address.String()

When the name of an Address contains non-ASCII characters,
Address.String() used mime.QEncoding to encode the name.

However certain characters are forbidden when an encoded-word is
in a phrase context (see RFC 2047 section 5.3) and these
characters are not encoded by mime.QEncoding.

In this case we now use mime.BEncoding (base64 encoding) so that
forbidden characters are also encoded.

Fixes #11292

Change-Id: I52db98b41ece439295e97d7e94c8190426f499c2
Reviewed-on: https://go-review.googlesource.com/16012
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agocmd/go: fix reading PT_NOTE segment with multiple notes
Ian Lance Taylor [Wed, 2 Dec 2015 05:16:29 +0000 (21:16 -0800)]
cmd/go: fix reading PT_NOTE segment with multiple notes

The old code was assuming that a PT_NOTE segment never had more than one
note, but there is no such requirement.

Fixes #13364.

Change-Id: I3f6b3716130bf7af6abe81b8e10571a8c7cd943c
Reviewed-on: https://go-review.googlesource.com/17331
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Reviewed-by: Russ Cox <rsc@golang.org>
9 years agocompress/bzip2/testdata: make Mark.Twain-Tom.Sawyer.txt free
Joe Tsai [Mon, 30 Nov 2015 22:43:33 +0000 (14:43 -0800)]
compress/bzip2/testdata: make Mark.Twain-Tom.Sawyer.txt free

Commit 7a1fb95d50b7a8302445ebc1296010695151ce7b strips non-free license
from Mark.Twain-Tom.Sawyer.txt, but forgot to remove it from the compressed
version of the file.

Update #13216

Change-Id: I60f53275d56ba5baa6898db47b1d41f85e985c00
Reviewed-on: https://go-review.googlesource.com/17264
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agoarchive/tar: convert Reader.Next to be loop based
Joe Tsai [Wed, 16 Sep 2015 07:58:56 +0000 (00:58 -0700)]
archive/tar: convert Reader.Next to be loop based

Motivation for change:
* Recursive logic is hard to follow, since it tends to apply
things in reverse. On the other hand, the tar formats tend to
describe meta headers as affecting the next entry.
* Recursion also applies changes in the wrong order. Two test
files are attached that use multiple headers. The previous Go
behavior differs from what GNU and BSD tar do.

Change-Id: Ic1557256fc1363c5cb26570e5d0b9f65a9e57341
Reviewed-on: https://go-review.googlesource.com/14624
Run-TryBot: Joe Tsai <joetsai@digital-static.net>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
9 years agonet/http: enable HTTP/2 on all Transports, not just the DefaultTransport
Brad Fitzpatrick [Tue, 1 Dec 2015 19:19:55 +0000 (19:19 +0000)]
net/http: enable HTTP/2 on all Transports, not just the DefaultTransport

This mirrors the same behavior and API from the server code to the
client side: if TLSNextProto is nil, HTTP/2 is on by default for
both. If it's non-nil, the user was trying to do something fancy and
step out of their way.

Updates #6891

Change-Id: Ia31808b71f336a8d5b44b985591d72113429e1d4
Reviewed-on: https://go-review.googlesource.com/17300
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agotime: fix handling of -07, handle Z07
Ian Lance Taylor [Mon, 30 Nov 2015 17:21:34 +0000 (09:21 -0800)]
time: fix handling of -07, handle Z07

The existing code has partial support for -07 (just the hours of a time
zone offset).  Complete the support, add support for Z07, and add a few
tests.

Fixes #13426.

Change-Id: Ic6377bbf3e65b4bb761b9779f7e80c07ce4f57e8
Reviewed-on: https://go-review.googlesource.com/17260
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
9 years agoarchive/tar: move parse/format methods to standalone receiver
Joe Tsai [Mon, 28 Sep 2015 20:49:35 +0000 (13:49 -0700)]
archive/tar: move parse/format methods to standalone receiver

Motivations for this change:
* It allows these functions to be used outside of Reader/Writer.
* It allows these functions to be more easily unit tested.

Change-Id: Iebe2b70bdb8744371c9ffa87c24316cbbf025b59
Reviewed-on: https://go-review.googlesource.com/15113
Reviewed-by: Russ Cox <rsc@golang.org>
Run-TryBot: Joe Tsai <joetsai@digital-static.net>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
9 years agonet/mail: better errors on non-ascii characters
Ingo Oeser [Fri, 4 Sep 2015 16:22:56 +0000 (18:22 +0200)]
net/mail: better errors on non-ascii characters

Fixes #12492

Change-Id: I8bb512027639301e2f2c41aab84e6d06ae88b137
Reviewed-on: https://go-review.googlesource.com/14312
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agonet: fall back to hosts file if DNS lookup fails, despite order
Benjamin Prosnitz [Mon, 30 Nov 2015 05:08:46 +0000 (13:08 +0800)]
net: fall back to hosts file if DNS lookup fails, despite order

Fixes #13090

Change-Id: I5612d792dabdff89bd0cec57dc2cacf9be7ebf64
Reviewed-on: https://go-review.googlesource.com/16341
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
9 years agoencoding/base64: add package-level example
Russ Cox [Wed, 25 Nov 2015 16:58:06 +0000 (11:58 -0500)]
encoding/base64: add package-level example

Fixes #13011.

Change-Id: Ia4c67880fca83f4298ff6bb1b217ec26c8c83427
Reviewed-on: https://go-review.googlesource.com/17231
Reviewed-by: David Crawshaw <crawshaw@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>

9 years agoregexp/syntax: fix handling of \Q...\E
Russ Cox [Wed, 25 Nov 2015 17:32:34 +0000 (12:32 -0500)]
regexp/syntax: fix handling of \Q...\E

It's not a group: must handle the inside as a sequence of literal chars,
not a single literal string.

That is, \Qab\E+ is the same as ab+, not (ab)+.

Fixes #11187.

Change-Id: I5406d05ccf7efff3a7f15395bdb0cfb2bd23a8ed
Reviewed-on: https://go-review.googlesource.com/17233
Reviewed-by: David Crawshaw <crawshaw@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agonet/http/httputil: add package level examples
Carlos Cirello [Tue, 1 Dec 2015 18:17:49 +0000 (19:17 +0100)]
net/http/httputil: add package level examples

Change-Id: I5be18b68602a7ebb740cf05924236ce17e386e55
Reviewed-on: https://go-review.googlesource.com/17295
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
9 years agotesting: document that T and B are safe for concurrent calls
Russ Cox [Wed, 25 Nov 2015 21:15:45 +0000 (16:15 -0500)]
testing: document that T and B are safe for concurrent calls

Fixes #13108.

Change-Id: I474cc2a3b7ced1c9eb978fc815f9c6bae9fb3ecc
Reviewed-on: https://go-review.googlesource.com/17235
Reviewed-by: Ian Lance Taylor <iant@golang.org>
9 years agonet/http: retry idempotent HTTP reqs on dead reused conns
Blake Gentry [Thu, 22 Jan 2015 23:58:25 +0000 (15:58 -0800)]
net/http: retry idempotent HTTP reqs on dead reused conns

If we try to reuse a connection that the server is in the process of
closing, we may end up successfully writing out our request (or a
portion of our request) only to find a connection error when we try to
read from (or finish writing to) the socket. This manifests as an EOF
returned from the Transport's RoundTrip.

The issue, among others, is described in #4677.

This change follows some of the Chromium guidelines for retrying
idempotent requests only when the connection has been already been used
successfully and no header data has yet been received for the response.

As part of this change, an unexported error was defined for
errMissingHost, which was previously defined inline. errMissingHost is
the only non-network error returned from a Request's Write() method.

Additionally, this breaks TestLinuxSendfile because its test server
explicitly triggers the type of scenario this change is meant to retry
on. Because that test server stops accepting conns on the test listener
before the retry, the test would time out. To fix this, the test was
altered to use a non-idempotent test type (POST).

Change-Id: I1ca630b944f0ed7ec1d3d46056a50fb959481a16
Reviewed-on: https://go-review.googlesource.com/3210
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>

9 years agoencoding/asn1: Reject invalid INTEGERs.
David Benjamin [Wed, 18 Nov 2015 04:24:36 +0000 (23:24 -0500)]
encoding/asn1: Reject invalid INTEGERs.

The empty string is not a valid DER integer. DER also requires that values be
minimally-encoded, so excess padding with leading 0s (0xff for negative
numbers) is forbidden. (These rules also apply to BER, incidentally.)

Fixes #12622.

Change-Id: I041f94e34a8afa29dbf94dd8fc450944bc91c9c3
Reviewed-on: https://go-review.googlesource.com/17008
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agoarchive/tar: fix issues with readGNUSparseMap1x0
Joe Tsai [Thu, 1 Oct 2015 08:35:15 +0000 (01:35 -0700)]
archive/tar: fix issues with readGNUSparseMap1x0

Motivations:
* Use of strconv.ParseInt does not properly treat integers as 64bit,
preventing this function from working properly on 32bit machines.
* Use of io.ReadFull does not properly detect truncated streams
when the file suddenly ends on a block boundary.
* The function blindly trusts user input for numEntries and allocates
memory accordingly.
* The function does not validate that numEntries is not negative,
allowing a malicious sparse file to cause a panic during make.

In general, this function was overly complicated for what it was
accomplishing and it was hard to reason that it was free from
bounds errors. Instead, it has been rewritten and relies on
bytes.Buffer.ReadString to do the main work. So long as invariants
about the number of '\n' in the buffer are maintained, it is much
easier to see why this approach is correct.

Change-Id: Ibb12c4126c26e0ea460ea063cd17af68e3cf609e
Reviewed-on: https://go-review.googlesource.com/15174
Reviewed-by: Russ Cox <rsc@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agoarchive/tar: properly handle header-only "files" in Reader
Joe Tsai [Wed, 4 Nov 2015 02:12:31 +0000 (18:12 -0800)]
archive/tar: properly handle header-only "files" in Reader

Certain special type-flags, specifically 1, 2, 3, 4, 5, 6,
do not have a data section. Thus, regardless of what the size field
says, we should not attempt to read any data for these special types.

The relevant PAX and USTAR specification says:
<<<
If the typeflag field is set to specify a file to be of type 1 (a link)
or 2 (a symbolic link), the size field shall be specified as zero.
If the typeflag field is set to specify a file of type 5 (directory),
the size field shall be interpreted as described under the definition
of that record type. No data logical records are stored for types 1, 2, or 5.
If the typeflag field is set to 3 (character special file),
4 (block special file), or 6 (FIFO), the meaning of the size field is
unspecified by this volume of POSIX.1-2008, and no data logical records shall
be stored on the medium.
Additionally, for type 6, the size field shall be ignored when reading.
If the typeflag field is set to any other value, the number of logical
records written following the header shall be (size+511)/512, ignoring
any fraction in the result of the division.
>>>

Contrary to the specification, we do not assert that the size field
is zero for type 1 and 2 since we liberally accept non-conforming formats.

Change-Id: I666b601597cb9d7a50caa081813d90ca9cfc52ed
Reviewed-on: https://go-review.googlesource.com/16614
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agobufio: clarify Read docs
Russ Cox [Thu, 26 Nov 2015 16:45:50 +0000 (11:45 -0500)]
bufio: clarify Read docs

Or at least make them true.

Fixes #12237.

Change-Id: I3c92a07233b2174c5731d6fa7fbb9ca3a97beb6b
Reviewed-on: https://go-review.googlesource.com/17237
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
9 years agoarchive/zip: enable overriding (de)compressors per file
Colin Cross [Thu, 5 Nov 2015 23:47:20 +0000 (15:47 -0800)]
archive/zip: enable overriding (de)compressors per file

Implement setting the compression level for a zip archive by registering
a per-Writer compressor through Writer.RegisterCompressor.  If no
compressors are registered, fall back to the ones registered at the
package level.  Also implements per-Reader decompressors.

Fixes #8359

Change-Id: I93b27c81947b0f817b42e0067aa610ff267fdb21
Reviewed-on: https://go-review.googlesource.com/16669
Reviewed-by: Joe Tsai <joetsai@digital-static.net>
Run-TryBot: Joe Tsai <joetsai@digital-static.net>
Reviewed-by: Klaus Post <klauspost@gmail.com>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
9 years agogo/types: fix race-y initialization of Struct.offsets
Aleksandr Demakin [Tue, 20 Oct 2015 20:05:50 +0000 (23:05 +0300)]
go/types: fix race-y initialization of Struct.offsets

Use sync.Once to ensure, that 'offsets' field is initialized
once only in a threadsafe way.

Fixes #12887

Change-Id: I90ef929c421ccd3094339c67a39b02d8f2e47211
Reviewed-on: https://go-review.googlesource.com/16013
Reviewed-by: Robert Griesemer <gri@golang.org>
9 years agomime: let FormatMediaType format slash-less media types, to mirror ParseMediaType.
Brad Fitzpatrick [Mon, 30 Nov 2015 19:33:49 +0000 (19:33 +0000)]
mime: let FormatMediaType format slash-less media types, to mirror ParseMediaType.

A Content-Type always has a slash (type/subtype)
A Content-Disposition does not (e.g. "attachment" or "line").
A "media type" is either one of those, plus optional parameters afterwards.

Our ParseMediaType and FormatMediaType weren't consistent in whether
they permitted Content-Dispositions. Now they both do.

Fixes #11289

Change-Id: Ia75723c9d7adb7f4de0f65482780f823cdadb5bd
Reviewed-on: https://go-review.googlesource.com/17135
Reviewed-by: Russ Cox <rsc@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agomime: don't accept single-quoted strings in media type parameter values
Brad Fitzpatrick [Mon, 30 Nov 2015 20:28:19 +0000 (20:28 +0000)]
mime: don't accept single-quoted strings in media type parameter values

Fix an old bug where media type parameter values could be escaped by
either double quotes (per the spec) or single quotes (due to my bug).

The original bug was introduced by me in git rev 90e4ece3
(https://golang.org/cl/4430049) in April 2011 when adding more tests
from http://greenbytes.de/tech/tc2231/ and misinterpreting the
expected value of test "attwithfntokensq" and not apparently thinking
about it enough.

No known spec or existing software produces or expects single quotes
around values. In fact, it would have be a parsing ambiguity if it
were allowed: the string `a=', b='` could parse as two keys "a" and
"b" both with value "'", or it could be parse as a single key "a" with
value "', b=".

Fixes #11291

Change-Id: I6de58009dd47dcabb120b017245d237cb7b1e89a
Reviewed-on: https://go-review.googlesource.com/17136
Reviewed-by: Russ Cox <rsc@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agounicode/utf8: add test for FullRune
Marcel van Lohuizen [Thu, 26 Nov 2015 09:33:45 +0000 (10:33 +0100)]
unicode/utf8: add test for FullRune

Check that it now properly handles \xC0 and \xC1.

Fixes #11733.

Change-Id: I66cfe0d43f9d123d4c4509a3fa18b9b6380dfc39
Reviewed-on: https://go-review.googlesource.com/17225
Reviewed-by: Russ Cox <rsc@golang.org>
9 years agoruntime: set r12 to sigpanic before jumping to it in sighandler
Michael Hudson-Doyle [Tue, 1 Dec 2015 01:47:22 +0000 (14:47 +1300)]
runtime: set r12 to sigpanic before jumping to it in sighandler

The ppc64le shared library ABI demands that r12 is set to a function's global
entrypoint before jumping to the global entrypoint. Not doing so means that
handling signals that usually panic actually crashes (and so, e.g. can't be
recovered). Fixes several failures of "cd test; go run run.go -linkshared".

Change-Id: Ia4d0da4c13efda68340d38c045a52b37c2f90796
Reviewed-on: https://go-review.googlesource.com/17280
Reviewed-by: Ian Lance Taylor <iant@golang.org>
9 years agospec: clarify examples for struct field tags
Robert Griesemer [Mon, 30 Nov 2015 22:18:09 +0000 (14:18 -0800)]
spec: clarify examples for struct field tags

Fixes #13420.

Change-Id: Id64ebd0527881450fdaffbb66d8b1831a6b7c43c
Reviewed-on: https://go-review.googlesource.com/17263
Reviewed-by: Rob Pike <r@golang.org>
9 years agocmd/compile: remove unused global variable loophack
Robert Griesemer [Thu, 26 Nov 2015 01:17:41 +0000 (17:17 -0800)]
cmd/compile: remove unused global variable loophack

Old parser remains.

Change-Id: I05ef1737802e23afc2c2129f58cb66feef8e3425
Reviewed-on: https://go-review.googlesource.com/17244
Reviewed-by: David Crawshaw <crawshaw@golang.org>
9 years agocmd/compile: document parsing decision for imported interfaces
Robert Griesemer [Tue, 24 Nov 2015 23:36:35 +0000 (15:36 -0800)]
cmd/compile: document parsing decision for imported interfaces

Fixes #13245.

Change-Id: I87be63cc7b27f70ca2f9fb6bc9908b9061fe3d9d
Reviewed-on: https://go-review.googlesource.com/17203
Reviewed-by: Chris Manghane <cmang@golang.org>
9 years agonet/http: update h2_bundle.go
Brad Fitzpatrick [Mon, 30 Nov 2015 20:07:50 +0000 (12:07 -0800)]
net/http: update h2_bundle.go

Updates to x/net rev 195180cf
(golang.org/cl/17134, http2: merge duplicate Transport dials)

Change-Id: I50b9c73b30c6a21e725aad80126b713d8b0fa362
Reviewed-on: https://go-review.googlesource.com/17261
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agonet/http: define File as an io.Seeker
Burcu Dogan [Sat, 28 Nov 2015 17:47:07 +0000 (11:47 -0600)]
net/http: define File as an io.Seeker

Fixes #13421.

Change-Id: Ic0fc0d0a8e86d2be23db12ddd1a8a8105e26ee56
Reviewed-on: https://go-review.googlesource.com/17250
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Run-TryBot: Burcu Dogan <jbd@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agoio/ioutil: add examples to functions
Carlos C [Sun, 9 Aug 2015 08:12:38 +0000 (10:12 +0200)]
io/ioutil: add examples to functions

Change-Id: Ica0cf1ba01ba73bd7e2fe38d653503738a962720
Reviewed-on: https://go-review.googlesource.com/13450
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agoio: add examples to functions
Carlos C [Mon, 27 Jul 2015 08:11:43 +0000 (10:11 +0200)]
io: add examples to functions

Change-Id: Id72cf4daceb7669cf9c7247df8e3ba086ccfe96e
Reviewed-on: https://go-review.googlesource.com/12730
Reviewed-by: Russ Cox <rsc@golang.org>
9 years agonet: fix typo in test
Mikio Hara [Sun, 29 Nov 2015 09:06:23 +0000 (18:06 +0900)]
net: fix typo in test

Change-Id: Ibc55e98c43cebe5a21ab3b280489e1bce6405b8b
Reviewed-on: https://go-review.googlesource.com/17218
Reviewed-by: Alex Brainman <alex.brainman@gmail.com>
9 years agocmd/compile: add 386 special case for testing first field of struct variable
Ian Lance Taylor [Mon, 16 Nov 2015 18:57:07 +0000 (10:57 -0800)]
cmd/compile: add 386 special case for testing first field of struct variable

This is the 386 version of the amd64-specific https://golang.org/cl/16933.

Update #12416.

Change-Id: Ibc3a99dcc753d6281839d8b61016d6c21dbd9649
Reviewed-on: https://go-review.googlesource.com/16970
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Russ Cox <rsc@golang.org>
9 years agocmd/cgo: make the char * pointer in GoString const
Ian Lance Taylor [Fri, 20 Nov 2015 01:03:01 +0000 (17:03 -0800)]
cmd/cgo: make the char * pointer in GoString const

This makes it more convenient for C code to use GoString with string
constants.  Since Go string values are immutable, the const qualifier is
appropriate in C.

Change-Id: I5fb3cdce2ce5079f1f0467a1544bb3a1eb27b811
Reviewed-on: https://go-review.googlesource.com/17067
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Russ Cox <rsc@golang.org>
9 years agonet: force LookupAddr results to be rooted DNS paths even in the case of local source
Mikio Hara [Fri, 27 Nov 2015 03:09:14 +0000 (12:09 +0900)]
net: force LookupAddr results to be rooted DNS paths even in the case of local source

The builtin name resolver using various resolution techniques is a bit
complicated and we sometimes fotget to take care of all the go and cgo
code paths and exchanging information to local and remote sources. This
change makes LookupAddr return absolute domain names even in the case of
local source.

Updates #12189.
Fixes #12240.

Change-Id: Icdd3375bcddc7f5d4d3b24f134d93815073736fc
Reviewed-on: https://go-review.googlesource.com/17216
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
9 years agopath: fix typo
Shenghou Ma [Sat, 28 Nov 2015 04:10:29 +0000 (23:10 -0500)]
path: fix typo

Fixes #13419.

Change-Id: I530d0b714ea0743c72eacd07b390c3f8cc556e21
Reviewed-on: https://go-review.googlesource.com/17239
Reviewed-by: Robert Griesemer <gri@golang.org>
9 years agocmd/compile: remove gratuituous type conversions
Robert Griesemer [Sat, 28 Nov 2015 00:11:05 +0000 (16:11 -0800)]
cmd/compile: remove gratuituous type conversions

Follow-up cleanup for https://go-review.googlesource.com/17248:
Use properly typed local variable op now that that variable use
is not overloaded anymore.

Also: Remove unnecessary if stmt from common lexical path.

Change-Id: I984b0b346f3fdccd5aedc937330c0a5f99acf324
Reviewed-on: https://go-review.googlesource.com/17249
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
9 years agocmd/compile: use correct line number for := (LCOLAS)
Robert Griesemer [Fri, 27 Nov 2015 23:27:14 +0000 (15:27 -0800)]
cmd/compile: use correct line number for := (LCOLAS)

- use same local variable name (lno) for line number for LCOLAS everywhere
- remove now unneeded assignment of line number to yylval.i in lexer

Fix per suggestion of mdempsky.

Fixes #13415.

Change-Id: Ie3c7f5681615042a12b81b26724b3a5d8a979c25
Reviewed-on: https://go-review.googlesource.com/17248
Run-TryBot: Robert Griesemer <gri@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
9 years agocrypto/x509: permit serial numbers to be negative.
Adam Langley [Fri, 27 Nov 2015 21:50:36 +0000 (13:50 -0800)]
crypto/x509: permit serial numbers to be negative.

Some software that produces certificates doesn't encode integers
correctly and, about half the time, ends up producing certificates with
serial numbers that are actually negative.

This buggy software, sadly, appears to be common enough that we should
let these errors pass. This change allows a Certificate.SerialNumber to
be negative.

Fixes #8265.

Change-Id: Ief35dae23988fb6d5e2873e3c521366fb03c6af4
Reviewed-on: https://go-review.googlesource.com/17247
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
9 years agonet: fix case insensitivity lookup for local database such as /etc/hosts
Mikio Hara [Fri, 27 Nov 2015 08:17:18 +0000 (17:17 +0900)]
net: fix case insensitivity lookup for local database such as /etc/hosts

The previous change for #12806 modified internal lookup tables and made
LookupAddr return forcibly lowercased host names by accident.

This change fixes the issue again without any behavioral change for
LookupAddr and adds missing test cases for lookupStaticHost and
lookupStaticAddr.

Updates #12806.
Fixes #13359.

Change-Id: Ifff4741cd79eb8b320b1b0f8c5e02b3a167c9fa8
Reviewed-on: https://go-review.googlesource.com/17217
Run-TryBot: Mikio Hara <mikioh.mikioh@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
9 years agocmd/cover: allow part selection to be retained across page refreshes
Shenghou Ma [Thu, 26 Nov 2015 22:45:09 +0000 (17:45 -0500)]
cmd/cover: allow part selection to be retained across page refreshes

Usually, you are primarily interested to see the coverage of a particular
file (e.g. when you're changing tests that affects a given source file),
it is very valuable if you can just refresh the page and immediately see
changes to the part you're already looking at (without selecting from the
selector again.)

Change-Id: I615207c9be6713f436e444771134fceaf4600ff3
Reviewed-on: https://go-review.googlesource.com/17238
Run-TryBot: Minux Ma <minux@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Andrew Gerrand <adg@golang.org>
9 years agocmd/yacc: memory allocation improvements
Tamir Duberstein [Mon, 9 Nov 2015 15:34:02 +0000 (10:34 -0500)]
cmd/yacc: memory allocation improvements

Places a fixed size initial stack and the lval inside the parser
struct so that they are allocated together. Places $$char inside the
parser struct to avoid allocating the closure used in Lookahead().

Change-Id: I0de664a6d612279fdc3255633e2dff904030bc36
Reviewed-on: https://go-review.googlesource.com/16705
Reviewed-by: Russ Cox <rsc@golang.org>
9 years agocrypto/tls: Server can specify an unadvertised cipher suite
Ralph Corderoy [Sat, 7 Nov 2015 12:18:36 +0000 (12:18 +0000)]
crypto/tls: Server can specify an unadvertised cipher suite

During the TLS handshake, check the cipher suite the server selects is
one of those offered in the ClientHello.  The code was checking it was
in the larger list that was sometimes whittled down for the ClientHello.

Fixes #13174

Change-Id: Iad8eebbcfa5027f30403b9700c43cfa949e135bb
Reviewed-on: https://go-review.googlesource.com/16698
Reviewed-by: Adam Langley <agl@golang.org>
Run-TryBot: Adam Langley <agl@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Russ Cox <rsc@golang.org>
9 years agocompress/flate: tweak offsetCode so that it can be inlined
Joe Tsai [Sun, 22 Nov 2015 10:41:49 +0000 (02:41 -0800)]
compress/flate: tweak offsetCode so that it can be inlined

Functions with switches (#13071) cannot be inlined.
Functions with consts (#7655) cannot be inlined.

benchmark                              old MB/s     new MB/s     speedup
BenchmarkEncodeDigitsSpeed1e4-4        10.25        10.20        1.00x
BenchmarkEncodeDigitsSpeed1e5-4        26.44        27.22        1.03x
BenchmarkEncodeDigitsSpeed1e6-4        32.28        33.51        1.04x
BenchmarkEncodeDigitsDefault1e4-4      8.61         8.74         1.02x
BenchmarkEncodeDigitsDefault1e5-4      7.03         6.98         0.99x
BenchmarkEncodeDigitsDefault1e6-4      6.47         6.46         1.00x
BenchmarkEncodeDigitsCompress1e4-4     8.62         8.73         1.01x
BenchmarkEncodeDigitsCompress1e5-4     7.01         6.98         1.00x
BenchmarkEncodeDigitsCompress1e6-4     6.43         6.53         1.02x
BenchmarkEncodeTwainSpeed1e4-4         9.67         10.16        1.05x
BenchmarkEncodeTwainSpeed1e5-4         26.46        26.94        1.02x
BenchmarkEncodeTwainSpeed1e6-4         33.19        34.02        1.03x
BenchmarkEncodeTwainDefault1e4-4       8.12         8.37         1.03x
BenchmarkEncodeTwainDefault1e5-4       8.22         8.21         1.00x
BenchmarkEncodeTwainDefault1e6-4       8.10         8.13         1.00x
BenchmarkEncodeTwainCompress1e4-4      8.24         8.39         1.02x
BenchmarkEncodeTwainCompress1e5-4      6.51         6.58         1.01x
BenchmarkEncodeTwainCompress1e6-4      6.16         6.13         1.00x

Change-Id: Ibafa5e3e2de0529853b5b3180e6fd6cb7090b76f
Reviewed-on: https://go-review.googlesource.com/17171
Reviewed-by: Russ Cox <rsc@golang.org>
9 years agonet/http: add test that automatic gzip works for HTTP2's Transport
Brad Fitzpatrick [Thu, 26 Nov 2015 00:08:32 +0000 (16:08 -0800)]
net/http: add test that automatic gzip works for HTTP2's Transport

And updates h2_bundle.go with the fix from x/net/http2.

Fixes #13298
Updates #6891

Change-Id: Ia25f22fa10e2a64b9d59211269882681aa18c101
Reviewed-on: https://go-review.googlesource.com/17241
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
Reviewed-by: Andrew Gerrand <adg@golang.org>
9 years agospec: clarify that iota is incremented even if not used in a const spec
Robert Griesemer [Wed, 25 Nov 2015 21:18:49 +0000 (13:18 -0800)]
spec: clarify that iota is incremented even if not used in a const spec

Slightly modified an example.

Fixes #13371.

Change-Id: I25d260d4200086a0ef9725950132b760657610c5
Reviewed-on: https://go-review.googlesource.com/17209
Reviewed-by: Rob Pike <r@golang.org>
9 years agogo/build: fix cgo ${SRCDIR} substitution when that variable contains spaces
Didier Spezia [Sun, 25 Oct 2015 20:18:17 +0000 (20:18 +0000)]
go/build: fix cgo ${SRCDIR} substitution when that variable contains spaces

When the source directory path contains spaces, cgo directives
cannot be properly validated:

$ pwd
/root/src/issue 11868

$ cat main.go
package main
//#cgo CFLAGS: -I${SRCDIR}/../../include
import "C"
func main() {
}

$ go build
can't load package: package issue 11868: /root/src/issue 11868/main.go:
 malformed #cgo argument: -I/root/src/issue 11868/../../include

Make sure spaces are tolerated in ${SRCDIR} when this variable
is expanded. This applies to ${SRCDIR} only. Shell safety
checks are still done in the same exact way for anything else.

Fixes #11868

Change-Id: I93d1d2b5ab167caa7ae353fe46fb8f69f1f06969
Reviewed-on: https://go-review.googlesource.com/16302
Reviewed-by: Russ Cox <rsc@golang.org>
9 years agointernal/race: add package
Dmitry Vyukov [Wed, 23 Sep 2015 08:03:54 +0000 (10:03 +0200)]
internal/race: add package

Factor out duplicated race thunks from sync, syscall net
and fmt packages into a separate package and use it.

Fixes #8593

Change-Id: I156869c50946277809f6b509463752e7f7d28cdb
Reviewed-on: https://go-review.googlesource.com/14870
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Run-TryBot: Dmitry Vyukov <dvyukov@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agocmd/internal/obj/x86: add support for TSX instructions
sergey.arseev [Mon, 2 Nov 2015 11:53:31 +0000 (14:53 +0300)]
cmd/internal/obj/x86: add support for TSX instructions

Transactional memory, will later be used for semaphore implementation.
Nacl not supported yet.

Change-Id: Ic18453dcaa08d07bb217c0b95461584f007d518b
Reviewed-on: https://go-review.googlesource.com/16479
Reviewed-by: Russ Cox <rsc@golang.org>
9 years agosyscall: skip tests that create a user namespace when chrooted
Michael Hudson-Doyle [Wed, 25 Nov 2015 22:47:32 +0000 (11:47 +1300)]
syscall: skip tests that create a user namespace when chrooted

The kernel rejects attempts to create user namespaces when in a chroot.

Change-Id: I6548302732c8f5be52f4167cd7233aea16839ad8
Reviewed-on: https://go-review.googlesource.com/17214
Reviewed-by: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agodoc: update go1.6.txt for cmd/cgo's C.complexfloat and C.complexdouble fix
Matthew Dempsky [Wed, 25 Nov 2015 23:08:00 +0000 (15:08 -0800)]
doc: update go1.6.txt for cmd/cgo's C.complexfloat and C.complexdouble fix

Updates #13402.

Change-Id: Ia7b729d81fb78206d214444911f2e6573b88717a
Reviewed-on: https://go-review.googlesource.com/17240
Reviewed-by: Ian Lance Taylor <iant@golang.org>
9 years agocmd/cgo: fix C.complexfloat and C.complexdouble
Matthew Dempsky [Wed, 25 Nov 2015 21:09:14 +0000 (13:09 -0800)]
cmd/cgo: fix C.complexfloat and C.complexdouble

This also fixes an unintended behavior where C's "complex float" and
"complex double" types were interchangeable with Go's "complex64" and
"complex128" types.

Fixes #13402.

Change-Id: I73f96d9a4772088d495073783c6982e9634430e8
Reviewed-on: https://go-review.googlesource.com/17208
Reviewed-by: Ian Lance Taylor <iant@golang.org>
9 years agocmd/cgo: handle another Clang DWARF quirk
Matthew Dempsky [Wed, 25 Nov 2015 20:40:43 +0000 (12:40 -0800)]
cmd/cgo: handle another Clang DWARF quirk

Without the fix:

    $ CC=clang-3.5 ./test.bash
    misc/cgo/errors/test.bash: BUG: expected error output to contain "C.ushort" but saw:
    # command-line-arguments
    ./issue13129.go:13: cannot use int(0) (type int) as type C.unsignedshort in assignment

Fixes #13129.

Change-Id: I2c019d2d000f5bfa3e33c477e533aff97031a84f
Reviewed-on: https://go-review.googlesource.com/17207
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agocmd/compile: clearer error for invalid array/slice literal elements
Robert Griesemer [Wed, 25 Nov 2015 01:48:52 +0000 (17:48 -0800)]
cmd/compile: clearer error for invalid array/slice literal elements

Fixes #13365.

Change-Id: I5a447ff806dbbb11c8c75e2b5cfa7fd4a845fb92
Reviewed-on: https://go-review.googlesource.com/17206
Run-TryBot: Robert Griesemer <gri@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agoregexp: fix one-pass compilation
Caleb Spare [Tue, 24 Nov 2015 00:16:42 +0000 (16:16 -0800)]
regexp: fix one-pass compilation

The one-pass transformation is structured as a search over the input
machine for conditions that violate the one-pass requisites. At each
iteration, we should fully explore all non-input paths that proceed from
the current instruction; this is implemented via recursive check calls.
But when we reach instructions that demand input (InstRune*), these
should be put onto the search queue.

Instead of searching this way, the routine previously (effectively)
proceeded through the machine one instruction at a time until finding an
Inst{Match,Fail,Rune*}, calling check on each instruction. This caused
bug #11905, where the transformation stopped before rewriting all
InstAlts as InstAltMatches.

Further, the check function unnecessarily recurred on InstRune*
instructions. (I believe this helps to mask the above bug.)

This change also deletes some unused functions and duplicate test cases.

Fixes #11905.

Change-Id: I5b0b26efea3d3bd01c7479a518b5ed1b886701cd
Reviewed-on: https://go-review.googlesource.com/17195
Reviewed-by: Russ Cox <rsc@golang.org>
9 years agoregexp: fix LiteralPrefix for certain onepass progs
Caleb Spare [Thu, 22 Oct 2015 06:03:54 +0000 (23:03 -0700)]
regexp: fix LiteralPrefix for certain onepass progs

The prefix computation for onepass was incorrectly returning
complete=true when it encountered a beginning-of-text empty width match
(^) in the middle of an expression.

Fix by returning complete only when the prefix is followed by $ and then
an accepting state.

Fixes #11175.

Change-Id: Ie9c4cf5f76c1d2c904a6fb2f016cedb265c19fde
Reviewed-on: https://go-review.googlesource.com/16200
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Russ Cox <rsc@golang.org>
9 years agoregexp: add Copy method to Regexp
Caleb Spare [Tue, 20 Oct 2015 08:37:06 +0000 (01:37 -0700)]
regexp: add Copy method to Regexp

This helps users who wish to use separate Regexps in each goroutine to
avoid lock contention. Previously they had to parse the expression
multiple times to achieve this.

I used variants of the included benchmark to evaluate this change. I
used the arguments -benchtime 20s -cpu 1,2,4,8,16 on a machine with 16
hardware cores.

Comparing a single shared Regexp vs. copied Regexps, we can see that
lock contention causes huge slowdowns at higher levels of parallelism.
The copied version shows the expected linear speedup.

name              old time/op  new time/op  delta
MatchParallel      366ns ± 0%   370ns ± 0%   +1.09%   (p=0.000 n=10+8)
MatchParallel-2    324ns ±28%   184ns ± 1%  -43.37%  (p=0.000 n=10+10)
MatchParallel-4    352ns ± 5%    93ns ± 1%  -73.70%   (p=0.000 n=9+10)
MatchParallel-8    480ns ± 3%    46ns ± 0%  -90.33%    (p=0.000 n=9+8)
MatchParallel-16   510ns ± 8%    24ns ± 6%  -95.36%   (p=0.000 n=10+8)

I also compared a modified version of Regexp that has no mutex and a
single machine (the "RegexpForSingleGoroutine" rsc mentioned in
https://github.com/golang/go/issues/8232#issuecomment-66096128).

In this next test, I compared using N copied Regexps vs. N separate
RegexpForSingleGoroutines. This shows that, even for this relatively
simple regex, avoiding the lock entirely would only buy about 10-12%
further improvement.

name              old time/op  new time/op  delta
MatchParallel      370ns ± 0%   322ns ± 0%  -12.97%    (p=0.000 n=8+8)
MatchParallel-2    184ns ± 1%   162ns ± 1%  -11.60%  (p=0.000 n=10+10)
MatchParallel-4   92.7ns ± 1%  81.1ns ± 2%  -12.43%  (p=0.000 n=10+10)
MatchParallel-8   46.4ns ± 0%  41.8ns ±10%   -9.78%   (p=0.000 n=8+10)
MatchParallel-16  23.7ns ± 6%  20.6ns ± 1%  -13.14%    (p=0.000 n=8+8)

Updates #8232.

Change-Id: I15201a080c363d1b44104eafed46d8df5e311902
Reviewed-on: https://go-review.googlesource.com/16110
Reviewed-by: Russ Cox <rsc@golang.org>
9 years agoregexp/syntax: correctly print `^` BOL and `$` EOL
Tamir Duberstein [Tue, 20 Oct 2015 15:21:21 +0000 (11:21 -0400)]
regexp/syntax: correctly print `^` BOL and `$` EOL

Fixes #12980.

Change-Id: I936db2f57f7c4dc80bb8ec32715c4c6b7bf0d708
Reviewed-on: https://go-review.googlesource.com/16112
Reviewed-by: Russ Cox <rsc@golang.org>
9 years agoencoding/xml: case-insensitive encoding recognition
Giulio Iotti [Mon, 31 Aug 2015 16:08:49 +0000 (19:08 +0300)]
encoding/xml: case-insensitive encoding recognition

From the XML spec: "XML processors should match character encoding
names in a case-insensitive way"

Fixes #12417.

Change-Id: I678c50152a49c14364be62b3f21ab9b9b009b24b
Reviewed-on: https://go-review.googlesource.com/14084
Reviewed-by: Russ Cox <rsc@golang.org>
9 years agoencoding/xml: reject invalid comments
Michal Bohuslávek [Wed, 2 Sep 2015 17:05:22 +0000 (19:05 +0200)]
encoding/xml: reject invalid comments

Fixes #11112.

Change-Id: I16e7363549a0dec8c61addfa14af0866c1fd7c40
Reviewed-on: https://go-review.googlesource.com/14173
Reviewed-by: Russ Cox <rsc@golang.org>
9 years agoencoding/xml: Add CDATA-wrapper output support to xml.Marshal.
Charles Weill [Fri, 23 Oct 2015 20:08:20 +0000 (16:08 -0400)]
encoding/xml: Add CDATA-wrapper output support to xml.Marshal.

Fixes #12963

Change-Id: Icc50dfb6130fe1e189d45f923c2f7408d3cf9401
Reviewed-on: https://go-review.googlesource.com/16047
Reviewed-by: Russ Cox <rsc@golang.org>
9 years agoencoding/gob: reduce the amount of memory allocations.
Aliaksandr Valialkin [Sat, 15 Aug 2015 10:04:46 +0000 (13:04 +0300)]
encoding/gob: reduce the amount of memory allocations.

Benchmark results:

benchmark                              old ns/op     new ns/op     delta
BenchmarkEndToEndPipe-4                7547          7294          -3.35%
BenchmarkEndToEndByteBuffer-4          5146          5092          -1.05%
BenchmarkEndToEndSliceByteBuffer-4     552779        439768        -20.44%
BenchmarkEncodeComplex128Slice-4       266370        266184        -0.07%
BenchmarkEncodeFloat64Slice-4          111891        110258        -1.46%
BenchmarkEncodeInt32Slice-4            74482         74080         -0.54%
BenchmarkEncodeStringSlice-4           84404         84279         -0.15%
BenchmarkEncodeInterfaceSlice-4        3942925       3045995       -22.75%
BenchmarkDecodeComplex128Slice-4       451837        415282        -8.09%
BenchmarkDecodeFloat64Slice-4          283584        262558        -7.41%
BenchmarkDecodeInt32Slice-4            246571        237383        -3.73%
BenchmarkDecodeStringSlice-4           734210        479625        -34.67%
BenchmarkDecodeInterfaceSlice-4        4778225       4160935       -12.92%

benchmark                              old allocs     new allocs     delta
BenchmarkEndToEndPipe-4                3              2              -33.33%
BenchmarkEndToEndByteBuffer-4          3              2              -33.33%
BenchmarkEndToEndSliceByteBuffer-4     1002           402            -59.88%
BenchmarkEncodeComplex128Slice-4       1              1              +0.00%
BenchmarkEncodeFloat64Slice-4          1              1              +0.00%
BenchmarkEncodeInt32Slice-4            1              1              +0.00%
BenchmarkEncodeStringSlice-4           1              1              +0.00%
BenchmarkEncodeInterfaceSlice-4        3001           1              -99.97%
BenchmarkDecodeComplex128Slice-4       188            185            -1.60%
BenchmarkDecodeFloat64Slice-4          188            185            -1.60%
BenchmarkDecodeInt32Slice-4            188            185            -1.60%
BenchmarkDecodeStringSlice-4           2188           1185           -45.84%
BenchmarkDecodeInterfaceSlice-4        6197           4194           -32.32%

benchmark                              old bytes     new bytes     delta
BenchmarkEndToEndPipe-4                64            48            -25.00%
BenchmarkEndToEndByteBuffer-4          64            48            -25.00%
BenchmarkEndToEndSliceByteBuffer-4     34551         10554         -69.45%
BenchmarkEncodeComplex128Slice-4       55            55            +0.00%
BenchmarkEncodeFloat64Slice-4          33            33            +0.00%
BenchmarkEncodeInt32Slice-4            32            32            +0.00%
BenchmarkEncodeStringSlice-4           36            36            +0.00%
BenchmarkEncodeInterfaceSlice-4        144555        347           -99.76%
BenchmarkDecodeComplex128Slice-4       28240         28097         -0.51%
BenchmarkDecodeFloat64Slice-4          11840         11697         -1.21%
BenchmarkDecodeInt32Slice-4            10817         10673         -1.33%
BenchmarkDecodeStringSlice-4           56128         39985         -28.76%
BenchmarkDecodeInterfaceSlice-4        132565        100421        -24.25%

Change-Id: Ief7c7706b1f2916486ab7190b81aafbb16b70f1e
Reviewed-on: https://go-review.googlesource.com/13660
Reviewed-by: Russ Cox <rsc@golang.org>
9 years agoencoding/json: check if Number is valid
Erik Dubbelboer [Wed, 15 Jul 2015 14:12:05 +0000 (16:12 +0200)]
encoding/json: check if Number is valid

json.Number is a special case which didn't have any checks and could result in invalid JSON.

Fixes #10281

Change-Id: Ie3e726e4d6bf6a6aba535d36f6107013ceac913a
Reviewed-on: https://go-review.googlesource.com/12250
Reviewed-by: Russ Cox <rsc@golang.org>
9 years agohash/crc32: add noescape tags to assembly functions
Joe Tsai [Fri, 20 Nov 2015 01:41:57 +0000 (17:41 -0800)]
hash/crc32: add noescape tags to assembly functions

CRC-32 computation is stateless and the p slice does not get stored
anywhere. Thus, we mark the assembly functions as noescape so that
it doesn't believe that p leaks in:
func Update(crc uint32, tab *Table, p []byte) uint32

Before:
./crc32.go:153: leaking param: p

After:
./crc32.go:153: Update p does not escape

Change-Id: I52ba35b6cc544fff724327140e0c27898431d1dc
Reviewed-on: https://go-review.googlesource.com/17069
Reviewed-by: Russ Cox <rsc@golang.org>
9 years agonet/http: more HTTP/2 tests and fixes
Brad Fitzpatrick [Tue, 24 Nov 2015 16:55:43 +0000 (08:55 -0800)]
net/http: more HTTP/2 tests and fixes

This compares the behavior of server handlers and the net/http
Transport in both HTTP/1 and HTTP/2 mode and verifies they're the
same.

This also moves some client<->server tests into clientserver_test.go.
Many of them were in serve_test.go or transport_test.go but were
basically testing both.

h2_bundle.go is an update of the golang.org/x/net/http2 code
from https://golang.org/cl/17204 (x/net git rev c745c36eab10)

Fixes #13315
Fixes #13316
Fixes #13317
Fixes other stuff found in the process too
Updates #6891 (http2 support in general)

Change-Id: Id9c45fad44cdf70ac95d2b89e578d66e882d3cc2
Reviewed-on: https://go-review.googlesource.com/17205
Reviewed-by: Andrew Gerrand <adg@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>

9 years agoruntime: fix conflict resolution in golang.org/cl/14207
Michael Hudson-Doyle [Wed, 25 Nov 2015 01:27:26 +0000 (14:27 +1300)]
runtime: fix conflict resolution in golang.org/cl/14207

Fixes testshared on arm64 and ppc64le.

Change-Id: Ie94bc0c85c7666fbb5ab6fc6d3dbb180407a9955
Reviewed-on: https://go-review.googlesource.com/17212
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
9 years agoruntime: check that masks and shifts are correct aligned
Shenghou Ma [Thu, 3 Sep 2015 06:44:26 +0000 (02:44 -0400)]
runtime: check that masks and shifts are correct aligned

We need a runtime check because the original issue is encountered
when running cross compiled windows program from linux. It's better
to give a meaningful crash message earlier than to segfault later.

The added test should not impose any measurable overhead to Go
programs.

For #12415.

Change-Id: Ib4a24ef560c09c0585b351d62eefd157b6b7f04c
Reviewed-on: https://go-review.googlesource.com/14207
Reviewed-by: Keith Randall <khr@golang.org>
Run-TryBot: Minux Ma <minux@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agocmd/compile: remove references to go.y and update documentation
Robert Griesemer [Tue, 24 Nov 2015 22:49:10 +0000 (14:49 -0800)]
cmd/compile: remove references to go.y and update documentation

This is a comment/documentation change only but for a minor
code change in the file and package_ methods (move recognition
of semi to match grammar better).

Per request from r.

Change-Id: I81ec985cc5831074d9eb5e8ffbf7e59466284819
Reviewed-on: https://go-review.googlesource.com/17202
Reviewed-by: Rob Pike <r@golang.org>
9 years agotext/template,html/template: correct comment in DefinedTemplates
Rob Pike [Tue, 24 Nov 2015 18:17:59 +0000 (10:17 -0800)]
text/template,html/template: correct comment in DefinedTemplates

The prefix includes a semicolon.

Change-Id: I4bdb79aa9931e835e297f3ea2c46a001cd123d56
Reviewed-on: https://go-review.googlesource.com/17200
Reviewed-by: Andrew Gerrand <adg@golang.org>
9 years agocompress: make Mark.Twain-Tom.Sawyer.txt licensed under non-free license free again
Jakub Čajka [Tue, 24 Nov 2015 14:36:32 +0000 (15:36 +0100)]
compress: make Mark.Twain-Tom.Sawyer.txt licensed under non-free license free again

This change strips non-free license from Mark.Twain-Tom.Sawyer.txt along with all reference to Project Gutenberg in the file and the whole source tree. Making the file public domain again.

Fixes #13216

Change-Id: I2f41b0de225f627dde152efe93c006a4c24be668
Reviewed-on: https://go-review.googlesource.com/17196
Reviewed-by: Andrew Gerrand <adg@golang.org>
9 years agoruntime: make gcFlushBgCredit go:nowritebarrierrec
Austin Clements [Thu, 19 Nov 2015 16:25:55 +0000 (11:25 -0500)]
runtime: make gcFlushBgCredit go:nowritebarrierrec

Write barriers in gcFlushBgCredit lead to very subtle bugs because it
executes after the getfull barrier. I tracked some bugs of this form
down before go:nowritebarrierrec was implemented. Ensure that they
don't reappear by making gcFlushBgCredit go:nowritebarrierrec.

Change-Id: Ia5ca2dc59e6268bce8d8b4c87055bd0f6e19bed2
Reviewed-on: https://go-review.googlesource.com/17052
Reviewed-by: Russ Cox <rsc@golang.org>
9 years agoruntime: recursively disallow write barriers in sighandler
Austin Clements [Tue, 17 Nov 2015 22:20:38 +0000 (17:20 -0500)]
runtime: recursively disallow write barriers in sighandler

sighandler may run during STW, so write barriers are not allowed.

Change-Id: Icdf46be10ea296fd87e73ab56ebb718c5d3c97ac
Reviewed-on: https://go-review.googlesource.com/17007
Reviewed-by: Russ Cox <rsc@golang.org>
9 years agoencoding/gob: document that zero elements of arrays and slices are sent
Rob Pike [Tue, 24 Nov 2015 18:32:37 +0000 (10:32 -0800)]
encoding/gob: document that zero elements of arrays and slices are sent

Fixes #13378

Change-Id: Ia78624ca1aa36ee906cef15416ea5554fa8229f2
Reviewed-on: https://go-review.googlesource.com/17201
Reviewed-by: Ian Lance Taylor <iant@golang.org>
9 years agosyscall: don't check result of close(fd) in forkAndExecInChild on Plan9
David du Colombier [Tue, 24 Nov 2015 17:54:58 +0000 (18:54 +0100)]
syscall: don't check result of close(fd) in forkAndExecInChild on Plan9

On multiprocessor machines, a file descriptor could be
closed twice in forkAndExecInChild. Consequently, the close
syscall returns the "fd out of range or not open" error
and forkAndExecInChild fails.

This changes forkAndExecInChild to ignore the error
returned by close(fd), as on other operating systems.

Fixes #12851.

Change-Id: I96a8463ce6599bfd1362353283e0329a00f738da
Reviewed-on: https://go-review.googlesource.com/17188
Reviewed-by: Rob Pike <r@golang.org>
9 years agoruntime: use a proper type, sigset, for m.sigmask
Elias Naur [Tue, 17 Nov 2015 10:41:06 +0000 (11:41 +0100)]
runtime: use a proper type, sigset, for m.sigmask

Replace the cross platform but unsafe [4]uintptr type with a OS
specific type, sigset. Most OSes already define sigset, and this
change defines a suitable sigset for the OSes that don't (darwin,
openbsd). The OSes that don't use m.sigmask (windows, plan9, nacl)
now defines sigset as the empty type, struct{}.

The gain is strongly typed access to m.sigmask, saving a dynamic
size sanity check and unsafe.Pointer casting. Also, some storage is
saved for each M, since [4]uinptr was conservative for most OSes.

The cost is that OSes that don't need m.sigmask has to define sigset.

completes ./all.bash with GOOS linux, on amd64
completes ./make.bash with GOOSes openbsd, android, plan9, windows,
darwin, solaris, netbsd, freebsd, dragonfly, all amd64.

With GOOS=nacl ./make.bash failed with a seemingly unrelated error.

[Replay of CL 16942 by Elias Naur.]

Change-Id: I98f144d626033ae5318576115ed635415ac71b2c
Reviewed-on: https://go-review.googlesource.com/17033
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Russ Cox <rsc@golang.org>

9 years agotest: add ability to run tests with dynamic linking
Michael Hudson-Doyle [Tue, 27 Oct 2015 01:54:19 +0000 (14:54 +1300)]
test: add ability to run tests with dynamic linking

This is a bit ugly but it's a useful test. Run go install -buildmode=shared std
and then go run run.go -linkshared (it passes on linux/amd64).

Change-Id: I5684c79cd03817fa1fc399788b7320f8535c08da
Reviewed-on: https://go-review.googlesource.com/16343
Reviewed-by: Russ Cox <rsc@golang.org>
9 years agounicode: add function level examples
Carlos Cirello [Wed, 26 Aug 2015 20:57:52 +0000 (22:57 +0200)]
unicode: add function level examples

Change-Id: I9e968c03c6615a539b18701c84f830898242441d
Reviewed-on: https://go-review.googlesource.com/13926
Reviewed-by: Russ Cox <rsc@golang.org>
9 years agounicode/utf8: don't imply that the empty string is incorrect UTF-8
Aaron Jacobs [Tue, 27 Oct 2015 21:58:26 +0000 (08:58 +1100)]
unicode/utf8: don't imply that the empty string is incorrect UTF-8

Change-Id: Idd9523949ee4f2f304b12be39f8940ba34a420be
Reviewed-on: https://go-review.googlesource.com/16361
Reviewed-by: Russ Cox <rsc@golang.org>
9 years agotesting: pause the test timer while waiting in T.Parallel
Caleb Spare [Tue, 17 Nov 2015 03:34:54 +0000 (19:34 -0800)]
testing: pause the test timer while waiting in T.Parallel

Before, we reset the timer at the end of T.Parallel, which is okay
assuming that T.Parallel is the first thing in the test.

Snapshot the elapsed time at the beginning of Parallel and include it in
the total duration so that any time spent in the test before calling
Parallel is reported in the test duration as well.

Updates #12243.

Change-Id: Ieca553e1f801e16b9b6416463fa8f7fa65425185
Reviewed-on: https://go-review.googlesource.com/16989
Reviewed-by: Russ Cox <rsc@golang.org>
9 years agoruntime/race: add tests for channels
dvyukov [Wed, 28 Oct 2015 09:49:17 +0000 (17:49 +0800)]
runtime/race: add tests for channels

These tests were failing on one of the versions of cl/9345
("runtime: simplify buffered channels").

Change-Id: I920ffcd28de428bcb7c2d5a300068644260e1017
Reviewed-on: https://go-review.googlesource.com/16416
Reviewed-by: Keith Randall <khr@golang.org>
Run-TryBot: Dmitry Vyukov <dvyukov@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agoruntime: speed up memclr with avx2 on amd64
Ilya Tocar [Fri, 30 Oct 2015 14:54:39 +0000 (17:54 +0300)]
runtime: speed up memclr with avx2 on amd64

Results are a bit noisy, but show good improvement (haswell)

name            old time/op    new time/op     delta
Memclr5-48        6.06ns ± 8%     5.65ns ± 8%    -6.81%  (p=0.000 n=20+20)
Memclr16-48       5.75ns ± 6%     5.71ns ± 6%      ~     (p=0.545 n=20+19)
Memclr64-48       6.54ns ± 5%     6.14ns ± 9%    -6.12%  (p=0.000 n=18+20)
Memclr256-48      10.1ns ±12%      9.9ns ±14%      ~     (p=0.285 n=20+19)
Memclr4096-48      104ns ± 8%       57ns ±15%   -44.98%  (p=0.000 n=20+20)
Memclr65536-48    2.45µs ± 5%     2.43µs ± 8%      ~     (p=0.665 n=16+20)
Memclr1M-48       58.7µs ±13%     56.4µs ±11%    -3.92%  (p=0.033 n=20+19)
Memclr4M-48        233µs ± 9%      234µs ± 9%      ~     (p=0.728 n=20+19)
Memclr8M-48        469µs ±11%      472µs ±16%      ~     (p=0.947 n=20+20)
Memclr16M-48       947µs ±10%      916µs ±10%      ~     (p=0.050 n=20+19)
Memclr64M-48      10.9ms ±10%      4.5ms ± 9%   -58.43%  (p=0.000 n=20+20)
GoMemclr5-48      3.80ns ±13%     3.38ns ± 6%   -11.02%  (p=0.000 n=20+20)
GoMemclr16-48     3.34ns ±15%     3.40ns ± 9%      ~     (p=0.351 n=20+20)
GoMemclr64-48     4.10ns ±15%     4.04ns ±10%      ~     (p=1.000 n=20+19)
GoMemclr256-48    7.75ns ±20%     7.88ns ± 9%      ~     (p=0.227 n=20+19)

name            old speed      new speed       delta
Memclr5-48       826MB/s ± 7%    886MB/s ± 8%    +7.32%  (p=0.000 n=20+20)
Memclr16-48     2.78GB/s ± 5%   2.81GB/s ± 6%      ~     (p=0.550 n=20+19)
Memclr64-48     9.79GB/s ± 5%  10.44GB/s ±10%    +6.64%  (p=0.000 n=18+20)
Memclr256-48    25.4GB/s ±14%   25.6GB/s ±12%      ~     (p=0.647 n=20+19)
Memclr4096-48   39.4GB/s ± 8%   72.0GB/s ±13%   +82.81%  (p=0.000 n=20+20)
Memclr65536-48  26.6GB/s ± 6%   27.0GB/s ± 9%      ~     (p=0.517 n=17+20)
Memclr1M-48     17.9GB/s ±12%   18.5GB/s ±11%      ~     (p=0.068 n=20+20)
Memclr4M-48     18.0GB/s ± 9%   17.8GB/s ±14%      ~     (p=0.547 n=20+20)
Memclr8M-48     17.9GB/s ±10%   17.8GB/s ±14%      ~     (p=0.947 n=20+20)
Memclr16M-48    17.8GB/s ± 9%   18.4GB/s ± 9%      ~     (p=0.050 n=20+19)
Memclr64M-48    6.19GB/s ±10%  14.87GB/s ± 9%  +140.11%  (p=0.000 n=20+20)
GoMemclr5-48    1.31GB/s ±10%   1.48GB/s ± 6%   +13.06%  (p=0.000 n=19+20)
GoMemclr16-48   4.81GB/s ±14%   4.71GB/s ± 8%      ~     (p=0.341 n=20+20)
GoMemclr64-48   15.7GB/s ±13%   15.8GB/s ±11%      ~     (p=0.967 n=20+19)

Change-Id: I393f3f20e2f31538d1b1dd70d6e5c201c106a095
Reviewed-on: https://go-review.googlesource.com/16773
Run-TryBot: Ilya Tocar <ilya.tocar@intel.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Klaus Post <klauspost@gmail.com>
Reviewed-by: Keith Randall <khr@golang.org>
9 years agoruntime: skip CallbackGC test in short mode on linux/arm
Dave Cheney [Fri, 11 Sep 2015 00:15:17 +0000 (10:15 +1000)]
runtime: skip CallbackGC test in short mode on linux/arm

Fixes #11959
Fixes #12035

Skip the CallbackGC test on linux/arm. This test takes between 30 and 60
seconds to run by itself, and is run 4 times over the course of ./run.bash
(once during the runtime test, three times more later in the build).

Change-Id: I4e7d3046031cd8c08f39634bdd91da6e00054caf
Reviewed-on: https://go-review.googlesource.com/14485
Reviewed-by: Russ Cox <rsc@golang.org>
9 years agoruntime: fix handling VirtualAlloc failure in sysUsed
Alex Brainman [Tue, 24 Nov 2015 03:29:06 +0000 (14:29 +1100)]
runtime: fix handling VirtualAlloc failure in sysUsed

Original code is mistakenly panics on VirtualAlloc failure - we want
it to go looking for smaller memory region that VirtualAlloc will
succeed to allocate. Also return immediately if VirtualAlloc succeeds.
See rsc comment on issue #12587 for details.

I still don't have a test for this. So I can only hope that this

Fixes #12587

Change-Id: I052068ec627fdcb466c94ae997ad112016f734b7
Reviewed-on: https://go-review.googlesource.com/17169
Reviewed-by: Russ Cox <rsc@golang.org>
9 years agotest: add test case for issue 13268
Robert Griesemer [Mon, 23 Nov 2015 23:21:13 +0000 (15:21 -0800)]
test: add test case for issue 13268

See https://go-review.googlesource.com/#/c/17047/ for the bug fix.

Change-Id: Id5b0a37439d0f8f1e668f7c2ac84335b37f3a300
Reviewed-on: https://go-review.googlesource.com/17177
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Reviewed-by: Russ Cox <rsc@golang.org>
Run-TryBot: Robert Griesemer <gri@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agodoc: add Code of Conduct
Andrew Gerrand [Tue, 24 Nov 2015 03:39:03 +0000 (14:39 +1100)]
doc: add Code of Conduct

Fixes #13073

Change-Id: I4fd9c6c61f1b9d49f66816839ca35209b4147ae3
Reviewed-on: https://go-review.googlesource.com/17167
Reviewed-by: Russ Cox <rsc@golang.org>
9 years agocmd/compile: remove dead code
Robert Griesemer [Tue, 24 Nov 2015 00:48:31 +0000 (16:48 -0800)]
cmd/compile: remove dead code

Get rid of residue after removing old parser.

Change-Id: I0dace1037d50959071a082c276f9f374eef6edb2
Reviewed-on: https://go-review.googlesource.com/17179
Run-TryBot: Robert Griesemer <gri@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agotest: remove dead branch from run.go
Caleb Spare [Tue, 17 Nov 2015 04:14:50 +0000 (20:14 -0800)]
test: remove dead branch from run.go

It relates to an obsolete notion of 'skipped-but-okay' tests.

Fixes #12394

Change-Id: Ib7e6444dc8565e1390bd42d65d30fac136593f78
Reviewed-on: https://go-review.googlesource.com/17000
Run-TryBot: Caleb Spare <cespare@gmail.com>
Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Dmitri Shuralyov <shurcool@gmail.com>
9 years agoruntime: mark {g,m,p}uintptr methods as nosplit
Michael Hudson-Doyle [Mon, 23 Nov 2015 20:15:36 +0000 (09:15 +1300)]
runtime: mark {g,m,p}uintptr methods as nosplit

These are methods that are "obviously" going to get inlined -- until you build
with -l, when they can trigger a stack split at a bad time.

Fixes #11482

Change-Id: Ia065c385978a2e7fe9f587811991d088c4d68325
Reviewed-on: https://go-review.googlesource.com/17165
Run-TryBot: Michael Hudson-Doyle <michael.hudson@canonical.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Russ Cox <rsc@golang.org>
9 years agoruntime: take stack barrier lock during copystack
Austin Clements [Mon, 23 Nov 2015 20:03:38 +0000 (15:03 -0500)]
runtime: take stack barrier lock during copystack

Commit bbd1a1c prevented SIGPROF from scanning stacks that were being
copied, but it didn't prevent a stack copy (specifically a stack
shrink) from happening while SIGPROF is scanning the stack. As a
result, a stack copy may adjust stack barriers while SIGPROF is in the
middle of scanning a stack, causing SIGPROF to panic when it detects
an inconsistent stack barrier.

Fix this by taking the stack barrier lock while adjusting the stack.
In addition to preventing SIGPROF from scanning this stack, this will
block until any in-progress SIGPROF is done scanning the stack.

For 1.5.2.

Fixes #13362.
Updates #12932.

Change-Id: I422219c363054410dfa56381f7b917e04690e5dd
Reviewed-on: https://go-review.googlesource.com/17191
Run-TryBot: Austin Clements <austin@google.com>
Reviewed-by: Russ Cox <rsc@golang.org>
Run-TryBot: Russ Cox <rsc@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>