Jack Lindamood [Fri, 15 Jul 2016 20:28:27 +0000 (13:28 -0700)]
context: add test for WithDeadline in the past
Adds a test case for calling context.WithDeadline() where the deadline
exists in the past. This change increases the code coverage of the
context package.
net/http: make Transport.RoundTrip return raw Conn.Read error on peek failure
From at least Go 1.4 to Go 1.6, Transport.RoundTrip would return the
error value from net.Conn.Read directly when the initial Read (1 byte
Peek) failed while reading the HTTP response, if a request was
outstanding. While never a documented or tested promise, Go 1.7 changed the
behavior (starting at https://golang.org/cl/23160).
This restores the old behavior and adds a test (but no documentation
promises yet) while keeping the fix for spammy logging reported in #15446.
This looks larger than it is: it just changes errServerClosedConn from
a variable to a type, where the type preserves the underlying
net.Conn.Read error, for unwrapping later in Transport.RoundTrip.
Fixes #16465
Change-Id: I6fa018991221e93c0cfe3e4129cb168fbd98bd27
Reviewed-on: https://go-review.googlesource.com/25153 Reviewed-by: Andrew Gerrand <adg@golang.org> Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
It was removed in upstream Chrome https://codereview.chromium.org/2016863004
Rather than update to the latest version, make the minimal change for Go 1.7 and
change the "showToUser" boolean from true to false.
Tested by hand that it goes away after this change.
Updates #16247
Change-Id: I051f49da878e554b1a34a88e9abc70ab50e18780
Reviewed-on: https://go-review.googlesource.com/25117 Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
David Chase [Wed, 20 Jul 2016 14:44:49 +0000 (10:44 -0400)]
cmd/compile: change phi location to be optimistic at backedges
This is:
(1) a simple trick that cuts the number of phi-nodes
(temporarily) inserted into the ssa representation by a factor
of 10, and can cut the user time to compile tricky inputs like
gogo/protobuf tests from 13 user minutes to 9.5, and memory
allocation from 3.4GB to 2.4GB.
(2) a fix to sparse lookup, that does not rely on
an assumption proven false by at least one pathological
input "etldlen".
These two changes fix unrelated compiler performance bugs,
both necessary to obtain good performance compiling etldlen.
Without them it takes 20 minutes or longer, with them it
completes in 2 minutes, without a gigantic memory footprint.
Updates #16407
Change-Id: Iaa8aaa8c706858b3d49de1c4865a7fd79e6f4ff7
Reviewed-on: https://go-review.googlesource.com/23136 Reviewed-by: Keith Randall <khr@golang.org>
Run-TryBot: David Chase <drchase@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Keith Randall [Tue, 19 Jul 2016 06:06:04 +0000 (23:06 -0700)]
cmd/compile: move phi args which are constants closer to the phi
entry:
x = MOVQconst [7]
...
b1:
goto b2
b2:
v = Phi(x, y, z)
Transform that program to:
entry:
...
b1:
x = MOVQconst [7]
goto b2
b2:
v = Phi(x, y, z)
This CL moves constant-generating instructions used by a phi to the
appropriate immediate predecessor of the phi's block.
We used to put all constants in the entry block. Unfortunately, in
large functions we have lots of constants at the start of the
function, all of which are used by lots of phis throughout the
function. This leads to the constants being live through most of the
function (especially if there is an outer loop). That's an O(n^2)
problem.
Note that most of the non-phi uses of constants have already been
folded into instructions (ADDQconst, MOVQstoreconst, etc.).
This CL may be generally useful for other instances of compiler
slowness, I'll have to check. It may cause some programs to run
slower, but probably not by much, as rematerializeable values like
these constants are allocated late (not at their originally scheduled
location) anyway.
This CL is definitely a minimal change that can be considered for 1.7.
We probably want to do a better job in the tighten pass generally, not
just for phi args. Leaving that for 1.8.
Ian Lance Taylor [Wed, 20 Jul 2016 22:40:10 +0000 (15:40 -0700)]
runtime: add explicit `INT $3` at end of Darwin amd64 sigtramp
The omission of this instruction could confuse the traceback code if a
SIGPROF occurred during a signal handler. The traceback code would
trace up to sigtramp, but would then get confused because it would see a
PC address that did not appear to be in the function.
runtime: support smaller physical pages than PhysPageSize
Most operations need an upper bound on the physical page size, which
is what sys.PhysPageSize is for (this is checked at runtime init on
Linux). However, a few operations need a *lower* bound on the physical
page size. Introduce a "minPhysPageSize" constant to act as this lower
bound and use it where it makes sense:
1) In addrspace_free, we have to query each page in the given range.
Currently we increment by the upper bound on the physical page
size, which means we may skip over pages if the true size is
smaller. Worse, we currently pass a result buffer that only has
enough room for one page. If there are actually multiple pages in
the range passed to mincore, the kernel will overflow this buffer.
Fix these problems by incrementing by the lower-bound on the
physical page size and by passing "1" for the length, which the
kernel will round up to the true physical page size.
2) In the write barrier, the bad pointer check tests for pointers to
the first physical page, which are presumably small integers
masquerading as pointers. However, if physical pages are smaller
than we think, we may have legitimate pointers below
sys.PhysPageSize. Hence, use minPhysPageSize for this test since
pointers should never fall below that.
In particular, this applies to ARM64 and MIPS. The runtime is
configured to use 64kB pages on ARM64, but by default Linux uses 4kB
pages. Similarly, the runtime assumes 16kB pages on MIPS, but both 4kB
and 16kB kernel configurations are common. This also applies to ARM on
systems where the runtime is recompiled to deal with a larger page
size. It is also a step toward making the runtime use only a
dynamically-queried page size.
Change-Id: I1fdfd18f6e7cbca170cc100354b9faa22fde8a69
Reviewed-on: https://go-review.googlesource.com/25020 Reviewed-by: Ian Lance Taylor <iant@golang.org> Reviewed-by: Cherry Zhang <cherryyz@google.com>
Run-TryBot: Austin Clements <austin@google.com>
The leak was reported internally on a sever canary that runs for days.
After a day server consumes 5.6GB, after 6 days -- 12.2GB.
The leak is exposed by the added benchmark.
The leak is fixed upstream in :
http://llvm.org/viewvc/llvm-project/compiler-rt/trunk/lib/tsan/rtl/tsan_rtl_thread.cc?view=diff&r1=276102&r2=276103&pathrev=276103
Fixes #16441
Change-Id: I9d4f0adef48ca6cf2cd781b9a6990ad4661ba49b
Reviewed-on: https://go-review.googlesource.com/25091 Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Run-TryBot: Dmitry Vyukov <dvyukov@google.com>
Ian Lance Taylor [Tue, 19 Jul 2016 06:00:43 +0000 (23:00 -0700)]
runtime: add as many extra M's as needed
When a non-Go thread calls into Go, the runtime needs an M to run the Go
code. The runtime keeps a list of extra M's available. When the last
extra M is allocated, the needextram field is set to tell it to allocate
a new extra M as soon as it is running in Go. This ensures that an extra
M will always be available for the next thread.
However, if many threads need an extra M at the same time, this
serializes them all. One thread will get an extra M with the needextram
field set. All the other threads will see that there is no M available
and will go to sleep. The one thread that succeeded will create a new
extra M. One lucky thread will get it. All the other threads will see
that there is no M available and will go to sleep. The effect is
thundering herd, as all the threads looking for an extra M go through
the process one by one. This seems to have a particularly bad effect on
the FreeBSD scheduler for some reason.
With this change, we track the number of threads waiting for an M, and
create all of them as soon as one thread gets through. This still means
that all the threads will fight for the lock to pick up the next M. But
at least each thread that gets the lock will succeed, instead of going
to sleep only to fight again.
This smooths out the performance greatly on FreeBSD, reducing the
average wall time of `testprogcgo CgoCallbackGC` by 74%. On GNU/Linux
the average wall time goes down by 9%.
Fixes #13926
Fixes #16396
Change-Id: I6dc42a4156085a7ed4e5334c60b39db8f8ef8fea
Reviewed-on: https://go-review.googlesource.com/25047
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Dmitry Vyukov <dvyukov@google.com>
net/smtp: document that the smtp package is frozen
This copies the frozen wording from the log/syslog package.
Fixes #16436
Change-Id: If5d478023328925299399f228d8aaf7fb117c1b4
Reviewed-on: https://go-review.googlesource.com/25080 Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Andrew Gerrand <adg@golang.org>
net/http, net/http/cgi: fix for CGI + HTTP_PROXY security issue
Because,
* The CGI spec defines that incoming request header "Foo: Bar" maps to
environment variable HTTP_FOO == "Bar". (see RFC 3875 4.1.18)
* The HTTP_PROXY environment variable is conventionally used to configure
the HTTP proxy for HTTP clients (and is respected by default for
Go's net/http.Client and Transport)
That means Go programs running in a CGI environment (as a child
process under a CGI host) are vulnerable to an incoming request
containing "Proxy: attacker.com:1234", setting HTTP_PROXY, and
changing where Go by default proxies all outbound HTTP requests.
This is CVE-2016-5386, aka https://httpoxy.org/
Fixes #16405
Change-Id: I6f68ade85421b4807785799f6d98a8b077e871f0
Reviewed-on: https://go-review.googlesource.com/25010
Run-TryBot: Chris Broadfoot <cbro@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Chris Broadfoot <cbro@golang.org>
doc/go1.7.html: mention specific runtime improvements
Most of the runtime improvements are hard to quantify or summarize,
but it's worth mentioning some of the substantial improvements in STW
time, and that the scavenger now actually works on ARM64, PPC64, and
MIPS.
Joe Tsai [Sat, 16 Jul 2016 09:42:52 +0000 (02:42 -0700)]
fmt: properly handle early io.EOF Reads in readRune.readByte
Change https://golang.org/cl/19895 caused a regression
where the last character in a string would be dropped if it was
accompanied by an io.EOF.
This change fixes the logic so that the last byte is still returned
without a problem.
Fixes #16393
Change-Id: I7a4d0abf761c2c15454136a79e065fe002d736ea
Reviewed-on: https://go-review.googlesource.com/24981 Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
all: rename vendored golang.org/x/net packages to golang_org
Regression from Go 1.6 to Go 1.7rc1: we had broken the ability for
users to vendor "golang.org/x/net/http2" or "golang.org/x/net/route"
because we were vendoring them ourselves and cmd/go and cmd/compile do
not understand multiple vendor directories across multiple GOPATH
workspaces (e.g. user's $GOPATH and default $GOROOT).
As a short-term fix, since fixing cmd/go and cmd/compile is too
invasive at this point in the cycle, just rename "golang.org" to
"golang_org" for the standard library's vendored copy.
Fixes #16333
Change-Id: I9bfaed91e9f7d4ca6bab07befe80d71d437a21af
Reviewed-on: https://go-review.googlesource.com/24902
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Ian Lance Taylor <iant@golang.org>
doc/go1.7.html: document that http.Server now enforces request versions
Document that the http.Server is now stricter about rejecting
requests with invalid HTTP versions, and also that it rejects plaintext
HTTP/2 requests, except for `PRI * HTTP/2.0` upgrade requests.
The relevant CL is https://golang.org/cl/24505.
doc/effective_go: clarify advice on returning interfaces
New Gophers sometimes misconstrue the advice in the "Generality" section
as "export interfaces instead of implementations" and add needless
interfaces to their code as a result. Down the road, they end up
needing to add methods and either break existing callers or have to
resort to unpleasant hacks (e.g. using "magic method" type-switches).
Weaken the first paragraph of this section to only advise leaving types
unexported when they will never need additional methods.
Change-Id: I32a1ae44012b5896faf167c02e192398a4dfc0b8
Reviewed-on: https://go-review.googlesource.com/24892 Reviewed-by: Rob Pike <r@golang.org> Reviewed-by: Andrew Gerrand <adg@golang.org>
cmd/go: don't fail on invalid GOOS/GOARCH pair when using gccgo
Fixes #12272
Change-Id: I0306ce0ef4a87df2158df3b7d4d8d93a1cb6dabc
Reviewed-on: https://go-review.googlesource.com/24864 Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
Ian Lance Taylor [Mon, 11 Jul 2016 23:15:03 +0000 (16:15 -0700)]
runtime: add ctxt parameter to cgocallback called from Go
The cgocallback function picked up a ctxt parameter in CL 22508.
That CL updated the assembler implementation, but there are a few
mentions in Go code that were not updated. This CL fixes that.
Ian Lance Taylor [Tue, 12 Jul 2016 03:43:05 +0000 (03:43 +0000)]
Revert "regexp: add the Fanout benchmark
This is a copy of the "FANOUT" benchmark recently added to RE2 with the
following comment:
// This has quite a high degree of fanout.
// NFA execution will be particularly slow.
Most of the benchmarks on the regexp package have very little fanout and
are designed for comparing the regexp package's NFA with backtracking
engines found in other regular expression libraries. This benchmark
exercises the performance of the NFA on expressions with high fanout.Reviewed-on: https://go-review.googlesource.com/24846 Reviewed-by: Andrew Gerrand <adg@golang.org> Reviewed-by: Ian Lance Taylor <iant@golang.org>
"
Michael Matloob [Mon, 11 Jul 2016 20:59:03 +0000 (14:59 -0600)]
regexp: add the Fanout benchmark
This is a copy of the "FANOUT" benchmark recently added to RE2 with the
following comment:
// This has quite a high degree of fanout.
// NFA execution will be particularly slow.
Most of the benchmarks on the regexp package have very little fanout and
are designed for comparing the regexp package's NFA with backtracking
engines found in other regular expression libraries. This benchmark
exercises the performance of the NFA on expressions with high fanout.
Change-Id: Ie9c8e3bbeffeb1fe9fb90474ddd19e53f2f57a52
Reviewed-on: https://go-review.googlesource.com/24846 Reviewed-by: Andrew Gerrand <adg@golang.org> Reviewed-by: Ian Lance Taylor <iant@golang.org>
Ian Lance Taylor [Sun, 10 Jul 2016 02:38:04 +0000 (19:38 -0700)]
math/rand: fix raciness in Rand.Read
There are no synchronization points protecting the readVal and readPos
variables. This leads to a race when Read is called concurrently.
Fix this by adding methods to lockedSource, which is the case where
a race matters.
Fixes #16308.
Change-Id: Ic028909955700906b2d71e5c37c02da21b0f4ad9
Reviewed-on: https://go-review.googlesource.com/24852 Reviewed-by: Joe Tsai <thebrokentoaster@gmail.com>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
Ian Lance Taylor [Thu, 7 Jul 2016 23:41:29 +0000 (16:41 -0700)]
runtime: fix nanotime for macOS Sierra
In the beta version of the macOS Sierra (10.12) release, the
gettimeofday system call changed on x86. Previously it always returned
the time in the AX/DX registers. Now, if AX is returned as 0, it means
that the system call has stored the values into the memory pointed to by
the first argument, just as the libc gettimeofday function does. The
libc function handles both cases, and we need to do so as well.
Ian Lance Taylor [Fri, 8 Jul 2016 00:43:08 +0000 (17:43 -0700)]
runtime: handle selects with duplicate channels in shrinkstack
The shrinkstack code locks all the channels a goroutine is waiting for,
but didn't handle the case of the same channel appearing in the list
multiple times. This led to a deadlock. The channels are sorted so it's
easy to avoid locking the same channel twice.
Fixes #16286.
Change-Id: Ie514805d0532f61c942e85af5b7b8ac405e2ff65
Reviewed-on: https://go-review.googlesource.com/24815
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Austin Clements <austin@google.com>
The test was checking for 1 of 2 possible error values. But based on
goroutine scheduling and the randomness of select statement receive
cases, it was possible for a 3rd type of error to be returned.
This modifies the code (not the test) to make that third type of error
actually the second type of error, which is a nicer error message.
The test is no longer flaky. The flake was very reproducible with a
5ms sleep before the select at the end of Transport.getConn.
Ian Lance Taylor [Tue, 28 Jun 2016 21:19:27 +0000 (14:19 -0700)]
cmd/compile, syscall: add //go:uintptrescapes comment, and use it
This new comment can be used to declare that the uintptr arguments to a
function may be converted from pointers, and that those pointers should
be considered to escape. This is used for the Call methods in
dll_windows.go that take uintptr arguments, because they call Syscall.
We can't treat these functions as we do syscall.Syscall, because unlike
Syscall they may cause the stack to grow. For Syscall we can assume that
stack arguments can remain on the stack, but for these functions we need
them to escape.
Fixes #16035.
Change-Id: Ia0e5b4068c04f8d303d95ab9ea394939f1f57454
Reviewed-on: https://go-review.googlesource.com/24551 Reviewed-by: David Chase <drchase@google.com>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Sam Whited [Wed, 6 Jul 2016 01:06:00 +0000 (20:06 -0500)]
encoding/xml: update docs to follow convention
Fixes #8833
Change-Id: I4523a1de112ed02371504e27882659bce8028a45
Reviewed-on: https://go-review.googlesource.com/24745 Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Docs for Encode and EncodeValue do not mention that
nil pointers are not permitted hence we panic,
because Gobs encode values yet nil pointers have no value
to encode. It moves a comment that was internal to EncodeValue
to the top level to make it clearer to users what to expect
when they pass in nil pointers.
Supplements test TestTopLevelNilPointer.
Change-Id: Ie54f609fde4b791605960e088456047eb9aa8738
Reviewed-on: https://go-review.googlesource.com/24740 Reviewed-by: Andrew Gerrand <adg@golang.org>
Run-TryBot: Andrew Gerrand <adg@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Ian Lance Taylor [Fri, 1 Jul 2016 20:52:26 +0000 (13:52 -0700)]
cmd/vet: remove copylock warning about result types and calls
Don't issue a copylock warning about a result type; the function may
return a composite literal with a zero value, which is OK.
Don't issue a copylock warning about a function call on the RHS, or an
indirection of a function call; the function may return a composite
literal with a zero value, which is OK.
Updates #16227.
Change-Id: I94f0e066bbfbca5d4f8ba96106210083e36694a2
Reviewed-on: https://go-review.googlesource.com/24711 Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
Run-TryBot: Ian Lance Taylor <iant@golang.org> Reviewed-by: Rob Pike <r@golang.org>
Monty Taylor [Fri, 1 Jul 2016 13:47:41 +0000 (08:47 -0500)]
cmd/go: remove noVCSSuffix check for OpenStack
The original intent of the code was to allow both with and without .git
suffix for now to allow a transition period. The noVCSSuffix check was a
copy pasta error.
Fixes #15979.
Change-Id: I3d39aba8d026b40fc445244d6d01d8bc1979d1e4
Reviewed-on: https://go-review.googlesource.com/24645 Reviewed-by: Ian Lance Taylor <iant@golang.org>
Cherry Zhang [Thu, 30 Jun 2016 10:36:31 +0000 (06:36 -0400)]
cmd/compile: don't set line number to 0 when building SSA
The frontend may emit node with line number missing. In this case,
use the parent line number. Instead of changing every call site of
pushLine, do it in pushLine itself.
Alan Donovan [Thu, 30 Jun 2016 18:32:03 +0000 (14:32 -0400)]
cmd/vet: lostcancel: treat naked return as a use of named results
+ test.
Fixes #16230
Change-Id: Idac995437146a9df9e73f094d2a31abc25b1fa62
Reviewed-on: https://go-review.googlesource.com/24681 Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Austin Clements [Wed, 29 Jun 2016 21:41:50 +0000 (17:41 -0400)]
runtime/internal/sys: implement Ctz and Bswap in assembly for 386
Ctz is a hot-spot in the Go 1.7 memory manager. In SSA it's
implemented as an intrinsic that compiles to a few instructions, but
on the old backend (all architectures other than amd64), it's
implemented as a fairly complex Go function. As a result, switching to
bitmap-based allocation was a significant hit to allocation-heavy
workloads like BinaryTree17 on non-SSA platforms.
For unknown reasons, this hit 386 particularly hard. We can regain a
lot of the lost performance by implementing Ctz in assembly on the
386. This isn't as good as an intrinsic, since it still generates a
function call and prevents useful inlining, but it's much better than
the pure Go implementation:
Brad Fitzpatrick [Wed, 29 Jun 2016 23:56:43 +0000 (16:56 -0700)]
net/http: be consistent about spelling of HTTP/1.x
There was only one use of "HTTP/1.n" compared to "HTTP/1.x":
h2_bundle.go:// "Just as in HTTP/1.x, header field names are strings of ASCII
httputil/dump.go:// DumpRequest returns the given request in its HTTP/1.x wire
httputil/dump.go:// intact. HTTP/2 requests are dumped in HTTP/1.x form, not in their
response.go:// Write writes r to w in the HTTP/1.x server response format,
server.go: // Request.Body. For HTTP/1.x requests, handlers should read any
server.go:// The default HTTP/1.x and HTTP/2 ResponseWriter implementations
server.go:// The default ResponseWriter for HTTP/1.x connections supports
server.go:// http1ServerSupportsRequest reports whether Go's HTTP/1.x server
server.go: // about HTTP/1.x Handlers concurrently reading and writing, like
server.go: // HTTP/1.x from here on.
transport.go: return fmt.Errorf("net/http: HTTP/1.x transport connection broken: %v", err)
Be consistent.
Change-Id: I93c4c873e500f51af2b4762055e22f5487a625ac
Reviewed-on: https://go-review.googlesource.com/24610 Reviewed-by: Andrew Gerrand <adg@golang.org>
Tom Bergan [Wed, 29 Jun 2016 14:45:23 +0000 (07:45 -0700)]
crypto/tls: Use the same buffer size in the client and server in the TLS throughput benchmark
I believe it's necessary to use a buffer size smaller than 64KB because
(at least some versions of) Window using a TCP receive window less than
64KB. Currently the client and server use buffer sizes of 16KB and 32KB,
respectively (the server uses io.Copy, which defaults to 32KB internally).
Since the server has been using 32KB, it should be safe for the client to
do so as well.
Dmitry Vyukov [Wed, 29 Jun 2016 09:09:36 +0000 (11:09 +0200)]
runtime: fix race atomic operations on external memory
The assembly is broken: it does `MOVQ g(R12), R14` expecting that
R12 contains tls address, but it does not do get_tls(R12) before.
This magically works on linux: `MOVQ g(R12), R14` is compiled to
`mov %fs:0xfffffffffffffff8,%r14` which does not use R12.
But it crashes on windows.
Add explicit `get_tls(R12)`.
Fixes #16206
Change-Id: Ic1f21a6fef2473bcf9147de6646929781c9c1e98
Reviewed-on: https://go-review.googlesource.com/24590 Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Ian Lance Taylor [Wed, 29 Jun 2016 00:23:01 +0000 (17:23 -0700)]
runtime: correct printing of blocked field in scheduler trace
When the blocked field was first introduced back in
https://golang.org/cl/61250043 the scheduler trace code incorrectly used
m->blocked instead of mp->blocked. That has carried through the
conversion to Go. This CL fixes it.
Konstantin Shaposhnikov [Mon, 27 Jun 2016 09:13:15 +0000 (17:13 +0800)]
cmd/vet: make checking example names in _test packages more robust
Prior to this change package "foo" had to be installed in order to check
example names in "foo_test" package.
However by the time "foo_test" package is checked a parsed "foo" package
has been already constructed. Use it to check example names.
Also change TestDivergentPackagesExamples test to pass directory of the
package to the vet tool as it is the most common way to invoke it. This
requires changes to errchk to add support for grabbing source files from
a directory.
Fixes #16189
Change-Id: Ief103d07b024822282b86c24250835cc591793e8
Reviewed-on: https://go-review.googlesource.com/24488 Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Brad Fitzpatrick [Tue, 28 Jun 2016 19:06:08 +0000 (12:06 -0700)]
database/sql: deflake TestPendingConnsAfterErr and fix races, panics
TestPendingConnsAfterErr only cared that things didn't deadlock, so 5
seconds is a sufficient timer. We don't need 100 milliseconds.
I was able to reproduce with a tiny (5 nanosecond) timeout value,
instead of 100 milliseconds. In the process of testing with -race and
a high -count= value, I noticed several data races and panics
(sendings on a closed channel) which are also fixed in this change.
Fixes #15684
Change-Id: Ib4605fcc0f296e658cb948352ed642b801cb578c
Reviewed-on: https://go-review.googlesource.com/24550 Reviewed-by: Marko Tiikkaja <marko@joh.to>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Andrew Gerrand <adg@golang.org>
Brad Fitzpatrick [Mon, 27 Jun 2016 23:39:40 +0000 (16:39 -0700)]
net/http: conditionally configure HTTP/2 in Server.Serve(Listener)
Don't configure HTTP/2 in http.Server.Serve(net.Listener) if the
Server's TLSConfig is set and doesn't include the "h2" NextProto
value. This avoids mutating a *tls.Config already in use if
previously passed to tls.NewListener.
Also document this. (it's come up a few times now)
Fixes #15908
Change-Id: I283eed82fdb29a791f80d801aadd9f75db244de0
Reviewed-on: https://go-review.googlesource.com/24508 Reviewed-by: Andrew Gerrand <adg@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Ian Lance Taylor <iant@golang.org>
Marcel van Lohuizen [Tue, 28 Jun 2016 10:31:02 +0000 (12:31 +0200)]
unicode: upgrade to version 9.0.0
Changes beyond generated tables:
- Now supports aliases to handle deprecated
property classes.
- Some Mongolian letters are now modifiers.
Other changes:
- strconv: newly generated table to be in sync
- regexp/syntax: updated maxFold
Fixes #16191
Change-Id: I56bdf21ee2f775f2a82d0465b3772faf5c24cb61
Reviewed-on: https://go-review.googlesource.com/24496
Run-TryBot: Marcel van Lohuizen <mpvl@golang.org> Reviewed-by: Ian Lance Taylor <iant@golang.org>
David Crawshaw [Tue, 28 Jun 2016 01:37:19 +0000 (21:37 -0400)]
reflect, runtime: optimize Name method
Several minor changes that remove a good chunk of the overhead added
to the reflect Name method over the 1.7 cycle, as seen from the
non-SSA architectures.
In particular, there are ~20 fewer instructions in reflect.name.name
on 386, and the method now qualifies for inlining.
The simple JSON decoding benchmark on darwin/386:
name old time/op new time/op delta
CodeDecoder-8 49.2ms ± 0% 48.9ms ± 1% -0.77% (p=0.000 n=10+9)
name old speed new speed delta
CodeDecoder-8 39.4MB/s ± 0% 39.7MB/s ± 1% +0.77% (p=0.000 n=10+9)
On darwin/amd64 the effect is less pronounced:
name old time/op new time/op delta
CodeDecoder-8 38.9ms ± 0% 38.7ms ± 1% -0.38% (p=0.005 n=10+10)
name old speed new speed delta
CodeDecoder-8 49.9MB/s ± 0% 50.1MB/s ± 1% +0.38% (p=0.006 n=10+10)
Counterintuitively, I get much more useful benchmark data out of my
MacBook Pro than a linux workstation with more expensive Intel chips.
While the laptop has fewer cores and an active GUI, the single-threaded
performance is significantly better (nearly 1.5x decoding throughput)
so the differences are more pronounced.
For #16117.
Change-Id: I4e0cc1cc2d271d47d5127b1ee1ca926faf34cabf
Reviewed-on: https://go-review.googlesource.com/24510 Reviewed-by: Ian Lance Taylor <iant@golang.org>
Lynn Boger [Thu, 16 Jun 2016 19:57:45 +0000 (14:57 -0500)]
runtime/internal/atomic: Use power5 compatible instructions for ppc64
This modifies a recent performance improvement to the
And8 and Or8 atomic functions which required both ppc64le
and ppc64 to use power8 instructions. Since then it was
decided that ppc64 (BE) should work for power5 and later.
This change uses instructions compatible with power5 for
ppc64 and uses power8 for ppc64le.
Fixes #16004
Change-Id: I623c75e8e6fd1fa063a53d250d86cdc9d0890dc7
Reviewed-on: https://go-review.googlesource.com/24181 Reviewed-by: Keith Randall <khr@golang.org>
Run-TryBot: Andrew Gerrand <adg@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Konstantin Shaposhnikov [Sat, 25 Jun 2016 12:32:40 +0000 (20:32 +0800)]
net/http/httptest: show usage of httptest.NewRequest in example
Change ExampleResponseRecorder to use httptest.NewRequest instead of
http.NewRequest. This makes the example shorter and shows how to use
one more function from the httptest package.
Ian Lance Taylor [Mon, 27 Jun 2016 15:21:55 +0000 (08:21 -0700)]
cmd/compile: avoid function literal name collision with "glob"
The compiler was treating all global function literals as occurring in a
function named "glob", which caused a symbol name collision when there
was an actual function named "glob". Fixed by adding a period.
Fixes #16193.
Change-Id: I67792901a8ca04635ba41d172bfaee99944f594d
Reviewed-on: https://go-review.googlesource.com/24500 Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Robert Griesemer <gri@golang.org>
Raul Silvera [Mon, 27 Jun 2016 16:43:14 +0000 (09:43 -0700)]
runtime/pprof: update comments to point to new pprof
In the comments for this file there is a reference to gperftools
for more info on pprof. pprof now live on its own repo on github,
and the version in gperftools is deprecated.
Change-Id: I8a188f129534f73edd132ef4e5a2d566e69df7e9
Reviewed-on: https://go-review.googlesource.com/24502 Reviewed-by: Ian Lance Taylor <iant@golang.org>
Keith Randall [Sun, 19 Jun 2016 02:40:57 +0000 (19:40 -0700)]
cmd/compile: keep heap pointer for escaping output parameters live
Make sure the pointer to the heap copy of an output parameter is kept
live throughout the function. The function could panic at any point,
and then a defer could recover. Thus, we need the pointer to the heap
copy always available so the post-deferreturn code can copy the return
value back to the stack.
Before this CL, the pointer to the heap copy could be considered dead in
certain situations, like code which is reverse dominated by a panic call.
Fixes #16095.
Change-Id: Ic3800423e563670e5b567b473bf4c84cddb49a4c
Reviewed-on: https://go-review.googlesource.com/24213
Run-TryBot: Keith Randall <khr@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: David Chase <drchase@google.com>
David Crawshaw [Sat, 25 Jun 2016 14:23:30 +0000 (10:23 -0400)]
encoding/json: copy-on-write cacheTypeFields
Swtich from a sync.RWMutex to atomic.Value for cacheTypeFields.
On GOARCH=386, this recovers most of the remaining performance
difference from the 1.6 release. Compared with tip on linux/386:
name old time/op new time/op delta
CodeDecoder-40 92.8ms ± 1% 87.7ms ± 1% -5.50% (p=0.000 n=10+10)
name old speed new speed delta
CodeDecoder-40 20.9MB/s ± 1% 22.1MB/s ± 1% +5.83% (p=0.000 n=10+10)
With more time and care, I believe more of the JSON decoder's work
could be shifted so it is done before decoding, and independent of
the number of bytes processed. Maybe someone could explore that for
Go 1.8.
Konstantin Shaposhnikov [Thu, 26 May 2016 01:45:57 +0000 (09:45 +0800)]
math/rand: fix comment about bits of seed used by the default Source
Fixes #15788
Change-Id: I5a1fd1e5992f1c16cf8d8437d742bf02e1653b9c
Reviewed-on: https://go-review.googlesource.com/23461 Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Ian Lance Taylor [Fri, 24 Jun 2016 23:59:39 +0000 (16:59 -0700)]
cmd/pprof: don't use local symbolization for remote source
If we are using a remote source (a URL), and the user did not specify
the executable file to use, then don't try to use a local source.
This was misbehaving because the local symbolizer will not fail
if there is any memory map available, but the presence of a memory map
does not ensure that the files and symbols are actually available.
We still need a pprof testsuite.
Fixes #16159.
Change-Id: I0250082a4d5181c7babc7eeec6bc95b2f3bcaec9
Reviewed-on: https://go-review.googlesource.com/24464
Run-TryBot: Ian Lance Taylor <iant@golang.org> Reviewed-by: Joe Tsai <thebrokentoaster@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Ian Lance Taylor [Fri, 24 Jun 2016 20:19:46 +0000 (13:19 -0700)]
cmd/pprof: ignore symbols with address 0 and size 0
Handling a symbol with address 0 and size 0, such as an ELF STT_FILE
symbols, was causing us to disassemble the entire program. We started
adding STT_FILE symbols to help fix issue #13247.
Fixes #16154.
Change-Id: I174b9614e66ddc3d65801f7c1af7650f291ac2af
Reviewed-on: https://go-review.googlesource.com/24460
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Joe Tsai <thebrokentoaster@gmail.com>
Nathan VanBenschoten [Thu, 23 Jun 2016 23:46:13 +0000 (19:46 -0400)]
math/big: special-case a 0 mantissa during Rat parsing
Previously, a 0 mantissa was special-cased during big.Float
parsing, but not during big.Rat parsing. This meant that a value
like 0e9999999999 would parse successfully in big.Float.SetString,
but would hang in big.Rat.SetString. This discrepancy became an
issue in https://golang.org/src/go/constant/value.go?#L250,
where the big.Float would report an exponent of 0, so
big.Rat.SetString would be used and would subsequently hang.
A Go Playground example of this is https://play.golang.org/p/3fy28eUJuF
The solution is to special-case a zero mantissa during big.Rat
parsing as well, so that neither big.Rat nor big.Float will hang when
parsing a value with 0 mantissa but a large exponent.
This was discovered using go-fuzz on CockroachDB:
https://github.com/cockroachdb/go-fuzz/blob/master/examples/parser/main.go
Fixes #16176
Change-Id: I775558a8682adbeba1cc9d20ba10f8ed26259c56
Reviewed-on: https://go-review.googlesource.com/24430 Reviewed-by: Robert Griesemer <gri@golang.org>
Run-TryBot: Robert Griesemer <gri@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
David Crawshaw [Fri, 24 Jun 2016 19:28:58 +0000 (15:28 -0400)]
cmd/compile, etc: use tflag to optimize Name()==""
Improves JSON decoding benchmark:
name old time/op new time/op delta
CodeDecoder-8 41.3ms ± 6% 39.8ms ± 1% -3.61% (p=0.000 n=10+10)
name old speed new speed delta
CodeDecoder-8 47.0MB/s ± 6% 48.7MB/s ± 1% +3.66% (p=0.000 n=10+10)
Change-Id: I524ee05c432fad5252e79b29222ec635c1dee4b4
Reviewed-on: https://go-review.googlesource.com/24452 Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Rob Pike [Thu, 23 Jun 2016 22:28:36 +0000 (15:28 -0700)]
time: update documentation for Duration.String regarding the zero value
It was out of date; in 1.7 the format changes to 0s.
Change-Id: I2013a1b0951afc5607828f313641b51c74433257
Reviewed-on: https://go-review.googlesource.com/24421 Reviewed-by: Joe Tsai <thebrokentoaster@gmail.com> Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Sameer Ajmani [Fri, 24 Jun 2016 14:48:06 +0000 (10:48 -0400)]
context: update documentation on cancelation and go vet check.
Also replace double spaces after periods with single spaces.
Change-Id: Iedaea47595c5ce64e7e8aa3a368f36d49061c555
Reviewed-on: https://go-review.googlesource.com/24431 Reviewed-by: Alan Donovan <adonovan@google.com> Reviewed-by: Ian Lance Taylor <iant@golang.org>
David Crawshaw [Fri, 24 Jun 2016 15:09:48 +0000 (11:09 -0400)]
reflect: avoid lock for some NumMethod()==0 cases
The encoding/json package uses NumMethod()==0 as a fast check for
interface satisfaction. In the case when a type has no methods at
all, we don't need to grab the RWMutex.
Improves JSON decoding benchmark on linux/amd64:
name old time/op new time/op delta
CodeDecoder-8 44.2ms ± 2% 40.6ms ± 1% -8.11% (p=0.000 n=10+10)
name old speed new speed delta
CodeDecoder-8 43.9MB/s ± 2% 47.8MB/s ± 1% +8.82% (p=0.000 n=10+10)
For #16117
Change-Id: Id717e7fcd2f41b7d51d50c26ac167af45bae3747
Reviewed-on: https://go-review.googlesource.com/24433 Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: David Crawshaw <crawshaw@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>