Russ Cox [Thu, 3 Nov 2016 01:52:27 +0000 (21:52 -0400)]
cmd/go: fix TestGoGetDashTIssue8181
The test case was importing golang.org/x/build/cmd/cl,
which is a package main and cannot be imported.
The test case (stored in a separate repo) has been changed
to import golang.org/x/build/gerrit. Update the test accordingly.
Fixes #17702.
Change-Id: I80e150092111b5a04bb00c992b32edb271edb086
Reviewed-on: https://go-review.googlesource.com/32616
Run-TryBot: Russ Cox <rsc@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Ian Lance Taylor <iant@golang.org>
Quentin Smith [Wed, 2 Nov 2016 20:18:22 +0000 (16:18 -0400)]
runtime/testdata/testprog: increase GCFairness2 timeout to 1s
OpenBSD's scheduler causes preemption to take 20+ms, so 30ms is not
enough time for 3 goroutines to run. This change continues to sleep for
30ms, but if it finds that the 3 goroutines have not run, it sleeps for
an additional 1s before declaring failure.
Updates #17712
Change-Id: I3e886e40d05192b7cb71b4f242af195836ef62a8
Reviewed-on: https://go-review.googlesource.com/32634 Reviewed-by: Rick Hudson <rlh@golang.org>
Run-TryBot: Quentin Smith <quentin@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Mohit Agarwal [Thu, 3 Nov 2016 14:23:27 +0000 (19:53 +0530)]
cmd/go: run mkalldocs.sh
Follow-up to CL 32114
Change-Id: I75247ed9c1c0a0e8a278eb75a60d4c5bee355409
Reviewed-on: https://go-review.googlesource.com/32690 Reviewed-by: Ian Lance Taylor <iant@golang.org>
David Crawshaw [Fri, 23 Sep 2016 22:39:31 +0000 (08:39 +1000)]
cmd/compile: write type symbols referenced in ptabs
The exported symbol for a plugin can be the only reference to a
type in a program. In particular, "var F func()" will have
the type *func(), which is uncommon.
David du Colombier [Thu, 3 Nov 2016 10:50:45 +0000 (11:50 +0100)]
net: disallow dialing and listening on port > 65535 on Plan 9
Since Dial(":80") was implemented on Plan 9 (CL 32593),
TestProtocolDialError is failing because dialing a port
superior to 65535 is supported on Plan 9.
This change disallows dialing and listening on ports
superior to 65535.
TestAssembly takes 20s on my machine,
which is too slow for normal operation.
Marking as -short has its dangers (#17472),
but hopefully we'll soon have a builder for that.
All the SSA tests are hermetic and not time sensitive
and can thus be run in parallel.
Reduces the cmd/compile/internal/gc test time during
all.bash on my laptop from 42s to 7s.
Keith Randall [Tue, 1 Nov 2016 22:28:10 +0000 (15:28 -0700)]
cmd/compile: compute faulting args before writing args to stack
when compiling f(a, b, c), we do something like:
*(SP+0) = eval(a)
*(SP+8) = eval(b)
*(SP+16) = eval(c)
call f
If one of those evaluations is later determined to unconditionally panic
(say eval(b) in this example), then the call is deadcode eliminated. But
any previous argument write (*(SP+0)=... here) is still around. Becuase
we only compute the size of the outarg area for calls which are still
around at the end of optimization, the space needed for *(SP+0)=v is not
accounted for and thus the outarg area may be too small.
The fix is to make sure that we evaluate any potentially panicing
operation before we write any of the args to the stack. It turns out
that fix is pretty easy, as we already have such a mechanism available
for function args. We just need to extend it to possibly panicing args
as well.
The resulting code (if b and c can panic, but a can't) is:
tmpb = eval(b)
*(SP+16) = eval(c)
*(SP+0) = eval(a)
*(SP+8) = tmpb
call f
This change tickled a bug in how we find the arguments for intrinsic
calls, so that latent bug is fixed up as well.
Keith Randall [Fri, 28 Oct 2016 18:37:45 +0000 (11:37 -0700)]
cmd/compile: do more type conversion inline
The code to do the conversion is smaller than the
call to the runtime.
The 1-result asserts need to call panic if they fail, but that
code is out of line.
The only conversions left in the runtime are those which
might allocate and those which might need to generate an itab.
Given the following types:
type E interface{}
type I interface { foo() }
type I2 iterface { foo(); bar() }
type Big [10]int
func (b Big) foo() { ... }
This CL inlines the following conversions:
was assertE2T
var e E = ...
b := i.(Big)
was assertE2T2
var e E = ...
b, ok := i.(Big)
was assertI2T
var i I = ...
b := i.(Big)
was assertI2T2
var i I = ...
b, ok := i.(Big)
was assertI2E
var i I = ...
e := i.(E)
was assertI2E2
var i I = ...
e, ok := i.(E)
These are the remaining runtime calls:
convT2E:
var b Big = ...
var e E = b
convT2I:
var b Big = ...
var i I = b
convI2I:
var i2 I2 = ...
var i I = i2
assertE2I:
var e E = ...
i := e.(I)
assertE2I2:
var e E = ...
i, ok := e.(I)
assertI2I:
var i I = ...
i2 := i.(I2)
assertI2I2:
var i I = ...
i2, ok := i.(I2)
Fixes #17405
Fixes #8422
Change-Id: Ida2367bf8ce3cd2c6bb599a1814f1d275afabe21
Reviewed-on: https://go-review.googlesource.com/32313
Run-TryBot: Keith Randall <khr@golang.org> Reviewed-by: David Chase <drchase@google.com>
Keith Randall [Tue, 1 Nov 2016 04:18:00 +0000 (21:18 -0700)]
cmd/compile: On a runtime.KeepAlive call, keep whole variable alive
We generate an OpKeepAlive for the idata portion of the interface
for a runtime.KeepAlive call. But given such an op, we need to keep
the entire containing variable alive, not just the range that was
passed to the OpKeepAlive operation.
Fixes #17710
Change-Id: I90de66ec8065e22fb09bcf9722999ddda289ae6e
Reviewed-on: https://go-review.googlesource.com/32477
Run-TryBot: Keith Randall <khr@golang.org> Reviewed-by: Joe Tsai <thebrokentoaster@gmail.com>
Joe Tsai [Thu, 27 Oct 2016 23:23:53 +0000 (16:23 -0700)]
archive/tar: disable prefix field in Writer
The proper fix for the Writer is too involved to be done in time
for Go 1.8. Instead, we do a localized fix that simply disables the
prefix encoding logic. While this will prevent some legitimate uses
of prefix, it will ensure that we don't keep outputting invalid
GNU format files that have the prefix field populated.
For headers with long filenames that could have used the prefix field,
they will be promoted to use the PAX format, which ensures that we
will still be able to encode all headers that we were able to do before.
Matthew Dempsky [Wed, 2 Nov 2016 16:47:43 +0000 (09:47 -0700)]
cmd/compile: avoid double export of aliased objects
Instead of writing out the original object for each alias, ensure we
export the original object before any aliases. This allows the aliases
to simply refer back to the original object by qualified name.
Fixes #17636.
Change-Id: If80fa8c66b8fee8344a00b55d25a8aef22abd859
Reviewed-on: https://go-review.googlesource.com/32575
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Robert Griesemer <gri@golang.org> Reviewed-by: Alan Donovan <adonovan@google.com>
Russ Cox [Thu, 27 Oct 2016 01:07:52 +0000 (21:07 -0400)]
net: fix Dial(":80") on Windows
Windows sockets allow bind to 0.0.0.0:80 but not connect to it.
To make Listen(":80") / Dial(":80") work as documented on Windows,
connect to 127.0.0.1 or ::1 (depending on network) in place of 0.0.0.0.
Russ Cox [Mon, 31 Oct 2016 23:27:26 +0000 (19:27 -0400)]
testing: introduce testing/internal/testdeps for holding testmain dependencies
Currently, we don't have package testing to import package regexp directly,
because then regexp can't have internal tests (or at least they become more
difficult to write), for fear of an import cycle. The solution we've been using
is for the generated test main package (pseudo-import path "testmain", package main)
to import regexp and pass in a matchString function for use by testing when
implementing the -run flags. This lets testing use regexp but without depending
on regexp and creating unnecessary cycles.
We want to add a few dependencies to runtime/pprof, notably regexp
but also compress/gzip, without causing those packages to have to work
hard to write internal tests.
Restructure the (dare I say it) dependency injection of regexp.MatchString
to be more general, and use it for the runtime/pprof functionality in addition
to the regexp functionality. The new package testing/internal/testdeps is
the root for the testing dependencies handled this way.
Code using testing.MainStart will have to change from passing in a matchString
implementation to passing in testdeps.TestDeps{}. Users of 'go test' don't do this,
but other build systems that have recreated 'go test' (for example, Blaze/Bazel)
may need to be updated. The new testdeps setup should make future updates
unnecessary, but even so we keep the comment about MainStart not being
subject to Go 1 compatibility.
Change-Id: Iec821d2afde10c79f95f3b23de5e71b219f47b92
Reviewed-on: https://go-review.googlesource.com/32455 Reviewed-by: Ian Lance Taylor <iant@golang.org>
Russ Cox [Mon, 31 Oct 2016 16:09:03 +0000 (12:09 -0400)]
cmd/pprof: move cmd/internal/pprof back to cmd/pprof/internal
CL 21870 moved the entire cmd/pprof/internal directory to cmd/internal/pprof
for use by cmd/trace, but really cmd/trace only needed cmd/pprof/internal/profile,
which became cmd/internal/pprof/profile, and then internal/pprof/profile.
Move the rest back under cmd/pprof so that it is clear that no other code
is reaching into the guts of cmd/pprof. Just like functions should not be
exported unless necessary, internals should not be made visible to more
code than necessary.
Raúl Silvera noted after the commit of CL 21870 that only the profile package
should have moved, but there was no followup fix (until now).
Change-Id: I603f4dcb0616df1e5d5eb7372e6fccda57e05079
Reviewed-on: https://go-review.googlesource.com/32453
Run-TryBot: Russ Cox <rsc@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Ian Lance Taylor <iant@golang.org>
Michael Munday [Wed, 2 Nov 2016 15:41:40 +0000 (11:41 -0400)]
bytes, strings: update s390x code to match amd64 changes
Updates the s390x-specific files in these packages with the changes
to the amd64-specific files made during the review of CL 31690. I'd
like to keep these files in sync unless there is a reason to
diverge.
Ian Lance Taylor [Tue, 1 Nov 2016 19:06:22 +0000 (12:06 -0700)]
cmd/go: use cgo -srcdir when using SWIG
SWIG generates cgo input files in the work directory. When those files
are passed directly to cgo, cgo generates long file names that include
the object directory (with slashes replaced by underscores). Instead,
use cgo's new -srcdir option to give it short file names.
When using both SWIG and cgo, copy the cgo files into the object
directory first.
Use a shorter object file name when compiling the C file generated by
SWIG.
Fixes #17070.
Change-Id: Ic558603f1731636d9999f3130ad0224b24bd7dcb
Reviewed-on: https://go-review.googlesource.com/32485
Run-TryBot: Ian Lance Taylor <iant@golang.org> Reviewed-by: David Crawshaw <crawshaw@golang.org>
Ian Lance Taylor [Tue, 1 Nov 2016 21:13:34 +0000 (14:13 -0700)]
cmd/cgo: only record typedef name for pointer to struct
In a function argument, we handle a typedef for a pointer specially,
using the pointer type rather than the typedef, to permit the Go calls
to match the laxer type conversions permitted in C. We record the
typedef so that we use that type in the C code, in case it has a special
attribute. However, using the typedef is wrong when using a pointer to a
basic type, because the C code may sometimes use the typedef and
sometimes not, and using the typedef in all cases will cause incorrect
type errors on the Go side. Fortunately we only really need to use the
typedef when pointing to a struct/union/class, and in such a case
confusion is unlikely.
Fixes #17723.
Change-Id: Id2eaeb156faeaf2e8eb9cf0b8f95b44caf8cfbd2
Reviewed-on: https://go-review.googlesource.com/32536
Run-TryBot: Ian Lance Taylor <iant@golang.org> Reviewed-by: David Crawshaw <crawshaw@golang.org>
Robert Griesemer [Tue, 1 Nov 2016 21:46:59 +0000 (14:46 -0700)]
go/importer: handle multiple imports of the same object
Before aliases, and because we chose a simple export format for them,
a package may now export the same object more than once if there are
multiple exported aliases referring to that object. The go/importer
made the assumption this couldn't happen. Adjust it.
Fixes #17726.
Change-Id: Ibb9fc669a8748200b45ad78934d7453e5a5aad82
Reviewed-on: https://go-review.googlesource.com/32538 Reviewed-by: Matthew Dempsky <mdempsky@google.com>
Run-TryBot: Robert Griesemer <gri@golang.org>
Ian Lance Taylor [Tue, 1 Nov 2016 20:04:56 +0000 (13:04 -0700)]
cmd/go: remove _cgo_flags from gccgo archive built using SWIG
Since SWIG uses cgo, when using SWIG with gccgo, the go tool will add a
_cgo_flags file to the package archive, just as it does with cgo. We
need to remove that file from the archive passed to the linker, just as
we do with cgo.
Change-Id: I5ef4fea92eec419f870fbf6f678691d15901ee6c
Reviewed-on: https://go-review.googlesource.com/32535
Run-TryBot: Ian Lance Taylor <iant@golang.org> Reviewed-by: Michael Hudson-Doyle <michael.hudson@canonical.com>
David Crawshaw [Mon, 31 Oct 2016 03:53:09 +0000 (23:53 -0400)]
cmd/link: support plugins with no exported symbols
A plugin with no exported symbols is still potentially very useful.
Its init functions are called on load, and it so it can have visible
side effects.
Fixes #17681
Change-Id: Icdca31f48e5ab13c99020a2ef724f3de47dcd74b
Reviewed-on: https://go-review.googlesource.com/32437
Run-TryBot: David Crawshaw <crawshaw@golang.org> Reviewed-by: Ian Lance Taylor <iant@golang.org>
David Crawshaw [Tue, 1 Nov 2016 19:28:55 +0000 (15:28 -0400)]
runtime: resolve type offsets using source module
The runtime.typeEquals function is used during typelinksinit to
determine the canonical set of *_type values to use throughout the
runtime. As such, it is run against non-canonical *_type values, that
is, types from modules that are duplicates of a type from another
module that was loaded earlier in the program life.
These non-canonical *_type values sometimes contain pointers. These
pointers are pointing to position-independent data, and so they are set
by ld.so using dynamic relocations when the module is loaded. As such,
the pointer can point to the equivalent memory from a previous module.
This means if typesEqual follows a pointer inside a *_type, it can end
up at a piece of memory from another module. If it reads a typeOff or
nameOff from that memory and attempts to resolve it against the
non-canonical *_type from the later module, it will end up with a
reference to junk memory.
Instead, resolve against the pointer the offset was read from, so the
data is valid.
Fixes #17709.
Should no longer matter after #17724 is resolved in a later Go.
Change-Id: Ie88b151a3407d82ac030a97b5b6a19fc781901cb
Reviewed-on: https://go-review.googlesource.com/32513
Run-TryBot: David Crawshaw <crawshaw@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
When the err from ReadFile is non-nil, we call t.Fatal(err).
Switch t.Fatal to t.Error + return.
ensure that close(results) happens on that code path as well.
Ilya Tocar [Fri, 21 Oct 2016 20:23:48 +0000 (23:23 +0300)]
bytes,strings: use IndexByte more often in Index on AMD64
IndexByte+compare is faster than indexShortStr in good case, when
first byte is rare, but is more costly in bad cases.
Start with IndexByte and switch to indexShortStr if we encounter
false positives more often than once per 8 bytes.
Brad Fitzpatrick [Tue, 1 Nov 2016 17:12:48 +0000 (17:12 +0000)]
net/http: add Transport.ProxyConnectHeader to control headers to proxies
Fixes #13290
Change-Id: I0f7e7683d86db501cbedb6a0b7349ceb0769701c
Reviewed-on: https://go-review.googlesource.com/32481 Reviewed-by: Martin Möhrmann <martisch@uos.de> Reviewed-by: Ian Lance Taylor <iant@golang.org>
Bryan C. Mills [Thu, 27 Oct 2016 21:36:39 +0000 (17:36 -0400)]
runtime: align stack pointer in sigfwd
sigfwd calls an arbitrary C signal handler function. The System V ABI
for x86_64 (and the most recent revision of the ABI for i386) requires
the stack to be 16-byte aligned.
Fixes: #17641
Change-Id: I77f53d4a8c29c1b0fe8cfbcc8d5381c4e6f75a6b
Reviewed-on: https://go-review.googlesource.com/32107
Run-TryBot: Bryan Mills <bcmills@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Ian Lance Taylor <iant@golang.org>
Brad Fitzpatrick [Tue, 1 Nov 2016 15:24:11 +0000 (15:24 +0000)]
net/http, net/http/httptest: make http2's TrailerPrefix work for http1
Go's http1 implementation originally had a mechanism to send HTTP
trailers based on pre-declaring the trailer keys whose values you'd
later let after the header was written.
http2 copied the same mechanism, but it was found to be unsufficient
for gRPC's wire protocol. A second trailer mechanism was added later
(but only to http2) for handlers that want to send a trailer without
knowing in advance they'd need to.
Copy the same mechanism back to http1 and document it.
Fixes #15754
Change-Id: I8c40d55e28b0e5b7087d3d1a904a392c56ee1f9b
Reviewed-on: https://go-review.googlesource.com/32479 Reviewed-by: Ian Lance Taylor <iant@golang.org>
Lynn Boger [Tue, 1 Nov 2016 15:43:34 +0000 (10:43 -0500)]
cmd/vendor/golang.org/x/arch/ppc64/ppc64asm: skip TestObjdumpPowerManual if not ppc64x
Skip TestObjdumpPowerManual if the host system is not ppc64 or ppc64le.
This test depends on using the host objdump and comparing output, which
does not work as expected if the test is run on another host.
Orignates from golang.org/x/arch/ppc64/ppc64asm commit 8e2d4898.
David Crawshaw [Mon, 31 Oct 2016 12:59:05 +0000 (08:59 -0400)]
plugin: do not leak cRelName on error path
Fixes #17683
Change-Id: I46f45c63796b58e8a8f14e37592231cbe7cd6934
Reviewed-on: https://go-review.googlesource.com/32438 Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
David Crawshaw [Mon, 31 Oct 2016 00:30:38 +0000 (20:30 -0400)]
runtime: access modules via a slice
The introduction of -buildmode=plugin means modules can be added to a
Go program while it is running. This means there exists some time
while the program is running with the module is on the moduledata
linked list, but it has not been initialized to the satisfaction of
other parts of the runtime. Notably, the GC.
This CL adds a new way of access modules, an activeModules function.
It returns a slice of modules that is built in the background and
atomically swapped in. The parts of the runtime that need to wait on
module initialization can use this slice instead of the linked list.
Fixes #17455
Change-Id: I04790fd07e40c7295beb47cea202eb439206d33d
Reviewed-on: https://go-review.googlesource.com/32357 Reviewed-by: Ian Lance Taylor <iant@golang.org>
Michael Munday [Fri, 7 Oct 2016 16:16:26 +0000 (12:16 -0400)]
cmd/compile: improve s390x rules for folding ADDconst into loads/stores
There is no benefit to folding ADDconsts unless the resultant immediate
will fit into a 20-bit signed integer, so limit these rules accordingly.
Also the signed load operations were missing, so I've added them, and
I've also removed some MOVDaddr rules that were dead code (MOVDaddrs
are rematerializable on s390x which means they can't take inputs other
than SP or SB).
Change-Id: Iebeba78da37d3d71d32d4b7f49fe4ea9095d40ec
Reviewed-on: https://go-review.googlesource.com/30616
Run-TryBot: Michael Munday <munday@ca.ibm.com>
TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org> Reviewed-by: David Chase <drchase@google.com>
Martin Möhrmann [Sat, 29 Oct 2016 23:54:19 +0000 (01:54 +0200)]
runtime: improve atoi implementation
- Adds overflow checks
- Adds parsing of negative integers
- Adds boolean return value to signal parsing errors
- Adds atoi32 for parsing of integers that fit in an int32
- Adds tests
Handling of errors to provide error messages
at the call sites is left to future CLs.
Updates #17718
Change-Id: I3cacd0ab1230b9efc5404c68edae7304d39bcbc0
Reviewed-on: https://go-review.googlesource.com/32390 Reviewed-by: Ian Lance Taylor <iant@golang.org> Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Daniel Martí [Thu, 27 Oct 2016 18:47:47 +0000 (19:47 +0100)]
testing: don't warn if -bench was passed
In a previous change, cmd/go was taught to show a "no tests ran" warning
if test did nothing. But it missed a case - if no tests nor examples ran
but any benchmarks were meant to be run, it would still produce the
warning. This meant that running only benchmarks, which is common, would
be confusing:
$ go test -run='^$' -bench=.
testing: warning: no tests to run
BenchmarkFoo-4 300000 5056 ns/op
[...]
I believe this was because of a copy-paste error in the tests. This was
being tested, but on the wrong file which does contain a test that was
being run. Fix the path and fix the now failing test by never showing
the warning if -bench was given a non-empty string.
The rationale is that if -bench was given but there was no output, it's
obvious that nothing happened as benchmarks always produce output even
without -v. So showing a warning in those cases is redundant.
To make future typos less likely, make sure that no tests are being run
in the cases where we only want to run benchmarks.
Fixes #17603.
Change-Id: I4c626caf39f72260c6a9761c06446663f465f947
Reviewed-on: https://go-review.googlesource.com/32157 Reviewed-by: Marcel van Lohuizen <mpvl@golang.org>
Run-TryBot: Marcel van Lohuizen <mpvl@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Emmanuel Odeke [Fri, 28 Oct 2016 15:11:39 +0000 (08:11 -0700)]
net/http/httputil: add ModifyResponse to reverseProxy
Adds ModifyResponse, an optional func to ReverseProxy
that modifies a response in the backend, right before
the headers of the response are written to the internal
response writer.
If ModifyResponse returns an error, the proxy returns
a StatusBadGateway error.
Joe Tsai [Tue, 1 Nov 2016 00:34:42 +0000 (17:34 -0700)]
encoding/json: marshal with null when RawMessage is nil
This CL expands upon a change made in (http://golang.org/cl/21811)
to ensure that a nil RawMessage gets serialized as "null" instead of
being a nil slice.
The added check only triggers when the RawMessage is nil. We do not
handle the case when the RawMessage is non-nil, but empty.
Austin Clements [Mon, 31 Oct 2016 21:37:05 +0000 (17:37 -0400)]
runtime: perform write barriers on direct channel receive
Currently we have write barriers for direct channel sends, where the
receiver is blocked and the sender is writing directly to the
receiver's stack; but not for direct channel receives, where the
sender is blocked and the receiver is reading directly from the
sender's stack.
This was okay with the old write barrier because either 1) the
receiver would write the received pointer into the heap (causing it to
be shaded), 2) the pointer would still be on the receiver's stack at
mark termination and we would rescan it, or 3) the receiver dropped
the pointer so it wasn't necessarily reachable anyway.
This is not okay with the write barrier because it lets a grey stack
send a white pointer to a black stack and then remove it from its own
stack. If the grey stack was the sole grey-protector of this pointer,
this hides the object from the garbage collector.
Fix this by making direct receives perform a stack-to-stack write
barrier just like direct sends do.
Fixes #17694.
Change-Id: I1a4cb904e4138d2ac22f96a3e986635534a5ae41
Reviewed-on: https://go-review.googlesource.com/32450 Reviewed-by: Rick Hudson <rlh@golang.org>
Dhananjay Nakrani [Sat, 29 Oct 2016 19:10:21 +0000 (12:10 -0700)]
cmd/compile: avoid nil-ing out a node's Type in typecheckcomplit() on error
typecheckcomplit nils out node's type, upon finding new errors.
This hides new errors in children's node as well as the type info
of current node. This change fixes that.
Fixes #17645.
Change-Id: Ib473291f31c7e8fa0307cb1d494e0c112ddd3583
Reviewed-on: https://go-review.googlesource.com/32324 Reviewed-by: Matthew Dempsky <mdempsky@google.com>
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Austin Clements [Sun, 30 Oct 2016 21:59:06 +0000 (17:59 -0400)]
runtime: make assists perform root jobs
Currently, assists can only perform heap marking jobs. However, at the
beginning of GC, there are only root jobs and no heap marking jobs. As
a result, there's often a period at the beginning of a GC cycle where
no goroutine has accumulated assist credit, but at the same time it
can't get any credit because there are no heap marking jobs for it to
do yet. As a result, many goroutines often block on the assist queue
at the very beginning of the GC cycle.
This commit fixes this by allowing assists to perform root marking
jobs. The tricky part of this (and the reason we haven't done this
before) is that stack scanning jobs can lead to deadlocks if the
goroutines performing the stack scanning are themselves
non-preemptible, since two non-preemptible goroutines may try to scan
each other. To address this, we use the same insight d6625ca used to
simplify the mark worker stack scanning: as long as we're careful with
the stacks and only drain jobs while on the system stack, we can put
the goroutine into a preemptible state while we drain jobs. This means
an assist's user stack can be scanned while it continues to do work.
This reduces the rate of assist blocking in the x/benchmarks HTTP
benchmark by a factor of 3 and all remaining blocking happens towards
the *end* of the GC cycle, when there may genuinely not be enough work
to go around.
Ideally, assists would get credit for working on root jobs. Currently
they do not; however, this change prioritizes heap work over root jobs
in assists, so they're likely to mostly perform heap work. In contrast
with mark workers, for assists, the root jobs act only as a backstop
to create heap work when there isn't enough heap work.
Fixes #15361.
Change-Id: If6e169863e4ad75710b0c8dc00f6125b41e9a595
Reviewed-on: https://go-review.googlesource.com/32432 Reviewed-by: Rick Hudson <rlh@golang.org>
Austin Clements [Sun, 30 Oct 2016 21:46:49 +0000 (17:46 -0400)]
runtime: lift systemstack part of gcAssistAlloc
This lifts the part of gcAssistAlloc that runs on the system stack to
its own function in preparation for letting assists perform root jobs
(notably stack scanning). This makes it easy to see that there are no
references to the user stack once we've entered gcAssistAlloc1, which
means it's safe to shrink the stack while in gcAssistAlloc1.
This does not yet make assists perform root jobs, so it's not actually
possible for the stack to shrink yet. That will happen in the next
commit.
The code in gcAssistAlloc1 is identical to the code that's currently
passed in a closure to systemstack with one exception. Currently, we
set the "completed" variable in the enclosing scope to indicate that
the assist completed the mark phase. This is exactly the sort of
cross-stack reference lifting this function is meant to prevent. We
replace this variable with setting gp.param to nil or non-nil to
indicate the completion status.
Dhananjay Nakrani [Sun, 30 Oct 2016 19:47:53 +0000 (12:47 -0700)]
cmd/compile: initialize Decldepth in all cases
Previously, on encountering Func.Nname.Type == nil, typecheckfunc()
returned without initializing Decldepth for that func. This causes
typecheckclosure() to fatal. This change ensures that we initialize
Decldepth in all cases.
Keith Randall [Mon, 31 Oct 2016 04:10:03 +0000 (21:10 -0700)]
cmd/compile: make [0]T and [1]T SSAable types
We used to have to keep on-stack copies of these types.
Now they can be registerized.
[0]T is kind of trivial but might as well handle it.
This change enables another change I'm working on to improve how x.(T)
expressions are handled (#17405). This CL helps because now all
types that are direct interface types are registerizeable (e.g. [1]*byte).
No higher-degree arrays for now because non-constant indexes are hard.
Update #17405
Change-Id: I2399940965d17b3969ae66f6fe447a8cefdd6edd
Reviewed-on: https://go-review.googlesource.com/32416
Run-TryBot: Keith Randall <khr@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: David Chase <drchase@google.com>
David Chase [Fri, 28 Oct 2016 17:33:57 +0000 (13:33 -0400)]
cmd/compile: mark temps with new AutoTemp flag, and use it.
This is an extension of
https://go-review.googlesource.com/c/31662/
to mark all the temporaries, not just the ssa-generated ones.
Before-and-after ls -l `go tool -n compile` shows a 3%
reduction in size (or rather, a prior 3% inflation for
failing to filter temps out properly.)
Replaced name-dependent "is it a temp?" tests with calls to
*Node.IsAutoTmp(), which depends on AutoTemp. Also replace
calls to istemp(n) with n.IsAutoTmp(), to reduce duplication
and clean up function name space. Generated temporaries
now come with a "." prefix to avoid (apparently harmless)
clashes with legal Go variable names.
Fixes #17644.
Fixes #17240.
Change-Id: If1417f29c79a7275d7303ddf859b51472890fd43
Reviewed-on: https://go-review.googlesource.com/32255
Run-TryBot: David Chase <drchase@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Matthew Dempsky <mdempsky@google.com>
Matthew Dempsky [Mon, 31 Oct 2016 19:01:32 +0000 (12:01 -0700)]
cmd/compile: remove legacy debug flags
-M, -P, and -R were for debugging backend passes that no longer
exists.
-g is used for debugging instructions generated with Gins, but the SSA
backend mostly generates instructions directly. The handful of
instructions still generated with Gins are pretty useless for
debugging.
-x was used to debug the old lexer, but now it only causes us to print
file names as they're parsed, and only if we manually hack the
compiler to enable tracing.
Mike Strosaker [Fri, 28 Oct 2016 23:50:16 +0000 (19:50 -0400)]
crypto/sha256: improve performance for sha256.block on ppc64le
Adds an assembly implementation of sha256.block for ppc64le to improve its
performance. This implementation is largely based on the original amd64
implementation, which unrolls the 64 iterations of the inner loop.
Fixes #17652
benchmark old ns/op new ns/op delta
BenchmarkHash8Bytes 1263 767 -39.27%
BenchmarkHash1K 14048 7766 -44.72%
BenchmarkHash8K 102245 55626 -45.60%
benchmark old MB/s new MB/s speedup
BenchmarkHash8Bytes 6.33 10.43 1.65x
BenchmarkHash1K 72.89 131.85 1.81x
BenchmarkHash8K 80.12 147.27 1.84x
Change-Id: Ib4adf429423b20495580400be10bd7e171bcc70b
Reviewed-on: https://go-review.googlesource.com/32318 Reviewed-by: Carlos Eduardo Seo <cseo@linux.vnet.ibm.com> Reviewed-by: David Chase <drchase@google.com>
Run-TryBot: David Chase <drchase@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Mike Strosaker [Sat, 29 Oct 2016 06:36:41 +0000 (02:36 -0400)]
crypto/sha512: improve performance for sha512.block on ppc64le
Adds an assembly implementation of sha512.block for ppc64le to improve its
performance. This implementation is largely based on the original amd64
implementation, unrolling the 80 iterations of the inner loop.
Fixes #17660
benchmark old ns/op new ns/op delta
BenchmarkHash8Bytes 1715 1133 -33.94%
BenchmarkHash1K 10098 5513 -45.41%
BenchmarkHash8K 68004 35278 -48.12%
benchmark old MB/s new MB/s speedup
BenchmarkHash8Bytes 4.66 7.06 1.52x
BenchmarkHash1K 101.40 185.72 1.83x
BenchmarkHash8K 120.46 232.21 1.93x
Change-Id: Ifd55a49a24cb159b3a09a8e928c3f37727aca103
Reviewed-on: https://go-review.googlesource.com/32320 Reviewed-by: Carlos Eduardo Seo <cseo@linux.vnet.ibm.com> Reviewed-by: David Chase <drchase@google.com>
Run-TryBot: David Chase <drchase@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>