Aman Gupta [Tue, 7 Nov 2017 07:02:21 +0000 (23:02 -0800)]
net: implement (*syscall.RawConn).Read/Write on Windows
RawRead assumes the callback will perform either (a) a blocking read
and always return true, (b) a blocking read with a SO_RCVTIMEO set
returning false on WSAETIMEDOUT, or (c) a non-blocking read
returning false on WSAEWOULDBLOCK. In the latter two cases, it uses
a 0-byte overlapped read for notifications from the IOCP runtime
when the socket becomes readable before trying again.
RawWrite assumes the callback will perform blocking write and will
always return true, and makes no effort to tie into the runtime loop.
Alex Brainman [Sat, 14 Apr 2018 04:14:28 +0000 (14:14 +1000)]
cmd/internal/obj/arm64: do not clear environment in TestLarge and TestNoRet
Windows process cannot run properly, if it only has
GOOS and GOARCH environment variables set. It needs
other environment variables. So adjust TestLarge and
TestNoRet to add GOOS and GOARCH to the existing
variables set instead of clearing environment.
Alex Brainman [Sat, 14 Apr 2018 02:48:55 +0000 (12:48 +1000)]
syscall: remove WSAEMSGSIZE
CL 92475 added WSAEMSGSIZE const to syscall package. But there
is already copy of WSAEMSGSIZE in internal/syscall/windows.
So delete syscall.WSAEMSGSIZE
cmd/compile/internal/ssa: prefer non-indexed stores on amd64
We sometimes generate code like this:
v473 = MOVQconst <uintptr> // constant..
v580 = MOVBstoreidx1 <mem> v1056 v473 v819 v491 // ..only used as an index
Rewrite indexed stores to non-indexed version, where possible.
This allows to eliminate const->register move, reducing codesize and lowering register pressure.
Ilya Tocar [Fri, 23 Mar 2018 20:45:03 +0000 (15:45 -0500)]
runtime: avoid division in growslice
Add a special case for power-of-2 sized elements.
We can replace div/mul with left/right shift and avoid expensive operation.
growslice is hotter for short slices of small elements, such as int16, so
add an int16 version for GrowSlice benchmark.
Eric Daniels [Wed, 4 Apr 2018 01:35:46 +0000 (21:35 -0400)]
runtime/traceback: support tracking goroutine ancestor tracebacks with GODEBUG="tracebackancestors=N"
Currently, collecting a stack trace via runtime.Stack captures the stack for the
immediately running goroutines. This change extends those tracebacks to include
the tracebacks of their ancestors. This is done with a low memory cost and only
utilized when debug option tracebackancestors is set to a value greater than 0.
Cherry Zhang [Wed, 14 Feb 2018 21:26:39 +0000 (16:26 -0500)]
all: use new softfloat on GOARM=5
Use the new softfloat support in the compiler, originally added
for softfloat on MIPS. This support is portable, so we can just
use it for softfloat on ARM.
In the old softfloat support on ARM, the compiler generates
floating point instructions, then the assembler inserts calls
to _sfloat before FP instructions. _sfloat decodes the following
FP instructions and simulates them.
In the new scheme, the compiler generates runtime calls to do FP
operations at a higher level. It doesn't generate FP instructions,
and therefore the assembler won't insert _sfloat calls, i.e. the
old mechanism is automatically suppressed.
The old method may be still be triggered with assembly code
using FP instructions. In the standard library, the only
occurance is math/sqrt_arm.s, which is rewritten to call to the
Go implementation instead.
Some significant speedups for code using floating points:
Hana Kim [Thu, 12 Apr 2018 21:06:33 +0000 (17:06 -0400)]
cmd/trace: change span id computation for trace view use
golang.org/cl/102697 attempted to fix the span presentation by utilizing
the position of the span in the span slices of a task. But it is
not complete either.
First, id=0 is omitted in json encoding and the trace viewer silently
drops entries with the missing id field, so we must avoid zero-value id.
Second, it is possible that a goroutine handles multiple tasks. Then,
id collisions will happen.
This takes a simpler approach - have a counter that increments for every
emitSpan call, and use the value as the id value.
Change-Id: Idaa9505634acf6d327c6f00af32d8260955b85e1
Reviewed-on: https://go-review.googlesource.com/106755 Reviewed-by: Heschi Kreinick <heschi@google.com>
Run-TryBot: Hyang-Ah Hana Kim <hyangah@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
fanzha02 [Wed, 21 Mar 2018 10:49:24 +0000 (10:49 +0000)]
cmd/internal/obj/arm64: fix the bug of incorrect handling negative offset of LDP/STP/LDPW/STPW
The current assembler will report error when the negative offset is in
the range of [-256, 0) and is not the multiples of 4/8.
The fix introduces C_NSAUTO_8, C_NSAUTO_4 and C_NAUTO4K. C_NPAUTO
includes C_NSAUTO_8 instead of C_NSAUTO, C_NAUTO4K includes C_NSAUTO_8,
C_NSAUTO_4 and C_NSAUTO. So that assembler will encode the negative offset
that is greater than -4095 and is not the multiples of 4/8 as two instructions.
cmd/compile: in escape analysis, propagate loop depth to field
The escape analysis models "loop depth". If the address of an
expression is assigned to something defined at a lower (outer)
loop depth, the escape analysis decides it escapes. However, it
uses the loop depth of the address operator instead of where
the RHS is defined. This causes an unnecessary escape if there is
an assignment inside a loop but the RHS is defined outside the
loop. This CL propagates the loop depth.
Change-Id: I7536aca1e90717283bd6a3bb4b1bab059b0cf720
Reviewed-on: https://go-review.googlesource.com/104677
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Ian Lance Taylor <iant@golang.org>
Tom Limoncelli [Sun, 8 Apr 2018 20:05:04 +0000 (16:05 -0400)]
io/ioutil: change TempFile prefix to a pattern
Users of TempFile need to be able to supply the suffix, especially
when using operating systems that give semantic meaning to the
filename extension such as Windows. Renaming the file to include
an extension after the fact is insufficient as it could lead to
race conditions.
If the string given to TempFile includes a "*", the random string
replaces the "*". For example "myname.*.bat" will result in a random
filename such as "myname.123456.bat". If no "*' is included the
old behavior is retained, and the random digits are appended to the
end.
If multiple "*" are included, the final one is replaced, thus
permitting a pathological programmer to create filenames such as
"foo*.123456.bat" but not "foo.123456.*.bat"
Ben Shi [Mon, 9 Apr 2018 11:12:15 +0000 (11:12 +0000)]
cmd/internal/obj/arm64: optimize constant pool
"MOVD $0xaaaaaaaa, R2"
"MOVD $-0x55555555, R3"
For the above instructions, 64-bit constants 0x00000000 aaaaaaaa
and 0xffffffff aaaaaaab are stored in the constant pool.
This CL optimizes them to
"MOVWU $0xaaaaaaaa, R2"
"MOVW $-0x05555555, R3"
and 32-bit constants 0xaaaaaaaa and 0xaaaaaaab are stored in the
constant pool.
There is a little size reduction (about total 5KB) in both the go
executable and the library files.
Elias Naur [Thu, 12 Apr 2018 18:33:47 +0000 (20:33 +0200)]
misc/ios: speed up the iOS exec wrapper
First, take the exclusive lock that ensures only one running binary
later: after assembling the gotest.app directory and signing it.
Second, don't pass -r to ios-deploy. The -r flag uninstalls the
app before installing it. It seems unnecessary, takes extra time
and if there was only the one developer app on the phone, it
will drop the developer permission on uninstall.
Change-Id: Ia222d3e5c2e1e2285f53074eb952941fd45fadd9
Reviewed-on: https://go-review.googlesource.com/106676 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
This commit includes efficiency improvements in two places in the
database/sql package where an "if err != nil" was redundant and
the error can be returned as-is (most of the code in the standard
library and even in the file I changed does it my suggested way).
Change-Id: Ib9dac69ed01ee846e570a776164cb87c2caee6ca
Reviewed-on: https://go-review.googlesource.com/106555 Reviewed-by: Daniel Theophanes <kardianos@gmail.com>
Run-TryBot: Daniel Theophanes <kardianos@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
The receiver was renamed 6 years ago in https://golang.org/cl/5674065
but the docs weren't updated to match.
Change-Id: I5e72cedc0e0f067382545d272f48a9c7dfb5a9b7
Reviewed-on: https://go-review.googlesource.com/104116 Reviewed-by: Ian Lance Taylor <iant@golang.org>
Michael Hudson-Doyle [Wed, 4 Apr 2018 22:07:41 +0000 (10:07 +1200)]
cmd/link: do not pass -no-pie to host linker when -linkshared is passed
As the comment above the code I'm changing says, when building with
-buildmode=exe, the default compiler flags produce code incompatible with PIE.
But when -linkshared is passed, the default compiler flags are not used so this
does not apply. And now I've found a system (linux/arm64 with glibc 2.27) where
this combination of flags causes a problem, albeit for reasons I don't really
understand, so stop passing -no-pie when -linkshared is passed.
Change-Id: I412ec7941dc0cb89e6d1b171fc29288aadcb9f20
Reviewed-on: https://go-review.googlesource.com/104815
Run-TryBot: Michael Hudson-Doyle <michael.hudson@canonical.com>
TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Ian Lance Taylor <iant@golang.org>
Matthew Dempsky [Wed, 11 Apr 2018 22:37:16 +0000 (15:37 -0700)]
cmd/compile, cmd/link: encode cgo directives using JSON
The standard library has plenty of polished encoder/decoder
implementations. No need for another ad-hoc one.
I considered using encoding/gob instead, but these strings go into the
package data part of the object file, so it's important they don't
contain "\n$$\n". Package json escapes newlines in strings, so it's
safe to use here.
os: document that Chown with -1 means to leave values unchanged, like POSIX
And fix the nacl implementation.
Fixes #24710
Change-Id: I31ffeea03a72dac5021ffb183fde31e9ffd060ad
Reviewed-on: https://go-review.googlesource.com/106464 Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Elias Naur [Tue, 10 Apr 2018 23:24:09 +0000 (01:24 +0200)]
misc/ios: don't wait for response to lldb run in the exec wrapper
CL 106096 changed the iOS exec wrapper to directly run the binary
without waiting for a SIGINT signal, but did so in a way that
expects a "(lldb)" response from lldb in 2 seconds. Lldb might
not out output anything until the program finishes, so change the
exec wrapper to just fire and forget the the run command and go
straight to waiting for exit, successfully or otherwise.
Robert Griesemer [Thu, 5 Apr 2018 23:04:56 +0000 (16:04 -0700)]
go/printer, gofmt: handle raw string literals containing newlines better
A raw string containing newlines breaks whatever columns structure
has been established so far. Recognize the situation and force a
new section of alignment with the first line break seen after the
the raw string.
Applied gofmt to src and misc.
Fixes #9064.
Change-Id: I961e94b529b1fd421908311f366b113e2ec9b7f0
Reviewed-on: https://go-review.googlesource.com/105040 Reviewed-by: Matthew Dempsky <mdempsky@google.com>
runtime: make Playground timestamps change when the stream fd changes
The process reading the output of the binary may read stderr and
stdout separately, and may interleave reads from the two streams
arbitrarily. Because we explicitly serialize writes on the writer
side, we can reuse a timestamp within a single stream without losing
information; however, if we use the same timestamp for write on both
streams, the reader can't tell how to interleave them.
This change ensures that every time we change between the two fds, we
also bump the timestamp. That way, writes within a stream continue to
show the same timestamp, but a sorted merge of the contents of the two
streams always interleaves them in the correct order.
This still requires a corresponding change to the Playground parser to
actually reconstruct the correct interleaving. It currently merges the
two streams without reordering them; it should instead buffer them
separately and perform a sorted merge. (See
https://golang.org/cl/105496.)
Updates golang/go#24615.
Updates golang/go#24659.
Change-Id: Id789dfcc02eb4247906c9ddad38dac50cf829979
Reviewed-on: https://go-review.googlesource.com/105235
Run-TryBot: Bryan C. Mills <bcmills@google.com>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Yury Smolsky <yury@smolsky.by> Reviewed-by: Ian Lance Taylor <iant@golang.org>
cmd/link: set S_ATTR_PURE_INSTRUCTIONS bit for text section on darwin
Mac otool and llvm-objdump distinguishes a Mach-O section is
text or data by looking at S_ATTR_PURE_INSTRUCTIONS bit. Without
this bit it thinks our function symbols are data, not functions.
Set this bit for text section to make otool/objdump happy.
Fixes #24706.
Change-Id: I5236482cb9a72474c23fbea0f35d5b5cc8491ea4
Reviewed-on: https://go-review.googlesource.com/105256
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Ian Lance Taylor <iant@golang.org>
Nick Patavalis [Sun, 11 Mar 2018 17:11:33 +0000 (19:11 +0200)]
os: use poller when NewFile is called with a blocking descriptor.
If NewFile is called with a file descriptor that is already set to
non-blocking mode, it tries to return a pollable file (one for which
SetDeadline methods work) by adding the filedes to the poll/netpoll
mechanism. If called with a filedes in blocking mode, it returns a
non-pollable file, as it always did.
Fixes #22939
Updates #24331
Change-Id: Id54c8be1b83e6d35e14e76d7df0e57a9fd64e176
Reviewed-on: https://go-review.googlesource.com/100077
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Ian Lance Taylor <iant@golang.org>
David Chase [Thu, 5 Apr 2018 20:14:42 +0000 (16:14 -0400)]
cmd/compile: use Block.Likely to put likely branch first
When a neither of a conditional block's successors follows,
the block must end with a conditional branch followed by a
an unconditional branch. If the (conditional) branch is
"unlikely", invert it and swap successors to make it
likely instead.
This doesn't matter to most benchmarks on amd64, but in one
instance on amd64 it caused a 30% improvement, and it is
otherwise harmless. The problematic loop is
for i := 0; i < w; i++ {
if pw[i] != 0 {
return true
}
}
compiled under GOEXPERIMENT=preemptibleloops
This the very worst-case benchmark for that experiment.
Also in this CL is a commoning up of heavily-repeated
boilerplate, which made it much easier to see that the
changes were applied correctly. In the future this should
allow un-exporting of SSAGenState.Branches once the
boilerplate-replacement is done everywhere.
TestServerDuplicateBackgroundRead has been causing crashes on the
netbsd-arm-bsiegert builder, with the system becoming completely
unresponsive (probably deadlocked). Skip this test while the crash
is under investigation.
Upstream bug report is http://gnats.netbsd.org/53173.
Ben Shi [Tue, 20 Mar 2018 08:25:08 +0000 (08:25 +0000)]
cmd/internal/obj/arm64: support SWPD/SWPW/SWPH/SWPB
SWPD/SWPW/SWPH/SWPB were introduced in ARMv8.1. They swap content
of register and memory atomically. And their difference is
SWPD: 64-bit double word data
SWPW: 32-bit word data (zero extended to 64-bit)
SWPH: 16-bit half word data (zero extended to 64-bit)
SWPB: 8-bit byte data (zero extended to 64-bit)
In #17528 it was discussed (off-topic to the actual issue) to reserve
GOARCH names for the RISC-V architecture. With the first RISC-V
Linux-capable development boards released (e.g. HiFive Unleashed),
Linux distributions being ported to RISC-V (e.g. Debian, Fedora) and
RISC-V support being added to gccgo (CL 96377), it becomes more likely
that Go software (and maybe Go itself) will be ported as well.
Add riscv and riscv64 (which is already used by gccgo), so Go 1.11 will
already recognize "*_riscv{,64}.go" as reserved files.
Change-Id: I042aab19c68751d82ea513e40f7b1d7e1ad924ea
Reviewed-on: https://go-review.googlesource.com/106256
Run-TryBot: Tobias Klauser <tobias.klauser@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Ian Lance Taylor <iant@golang.org>
fanzha02 [Tue, 20 Mar 2018 03:53:21 +0000 (03:53 +0000)]
internal/bytealg: add optimized Compare for arm64
Use LDP instructions to load 16 bytes per loop when the source length is long. Specially
process the 8 bytes length, 4 bytes length and 2 bytes length to get a better performance.
Alberto Donizetti [Wed, 11 Apr 2018 08:50:09 +0000 (10:50 +0200)]
cmd/go: document GOTOOLDIR environment variable
Also, make the variables list sorted.
Fixes #24794
Change-Id: I55f77004b00391875d26df8e55e54d79cef168dc
Reviewed-on: https://go-review.googlesource.com/106255 Reviewed-by: Ian Lance Taylor <iant@golang.org>
Rob Pike [Tue, 10 Apr 2018 23:49:32 +0000 (09:49 +1000)]
doc/install.html: address comments from review of previous edit
Point out that one can just run the commands now; it's not necessary
to log out first.
Change-Id: I48d0cc0273d97ba54ce59b3a3bbcae0b5af9aaef
Reviewed-on: https://go-review.googlesource.com/106195 Reviewed-by: Ian Lance Taylor <iant@golang.org>
We'll always generate method expression wrappers for declared
interface types in their own package, so no need to generate them in
downstream packages.
Noticed by gri@ while looking into #21282.
Change-Id: I4fb7051b4e15297933da05fdd2b111d6b8f4178e
Reviewed-on: https://go-review.googlesource.com/106175
Run-TryBot: Matthew Dempsky <mdempsky@google.com> Reviewed-by: Robert Griesemer <gri@golang.org>
go/doc: tune association of a function with a type
Previously, we used to associate a function with its first returned type
assuming that it is a factory function for that type.
However, a function may return multiple types in which case it is usually
doing something else. Check for multiple return types, and treat it as
a normal function in that case. Maintain same behavior if the function
returns just one type.
Fixes #12839
Change-Id: Ic4ac11d322996f216f593b71f4e61ad4270d5213
Reviewed-on: https://go-review.googlesource.com/105575 Reviewed-by: Robert Griesemer <gri@golang.org>
Run-TryBot: Robert Griesemer <gri@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Hana Kim [Wed, 4 Apr 2018 18:24:02 +0000 (14:24 -0400)]
cmd/trace: pprof computation for span types
/spanio, /spanblock, /spansched, /spansyscall provide
the pprof-style summary of span execution's
io, block, scheduling, syscall latency distributions
respectively.
The computation logic for /io, /block, /sched, /syscall
analysis was refactored and extended for reuse in these
new types of analysis. Upon the analysis query, we create
a map of goroutine id to time intervals based on the query
parameter, that represents the interesting time intervals
of matching goroutines. Only the events from the matching
goroutines that fall into the intervals are considered
in the pprof computation.
The new endpoints are not yet hooked into other span
analysis page (e.g. /userspan) yet.
Vlad Krasnov [Fri, 9 Mar 2018 00:07:42 +0000 (16:07 -0800)]
crypto/elliptic: improve P256 implementation on amd64 a bit
Minor modifications to the optimized amd64 implememntation.
* Reduce window size: reduces size of the lookup tables by 40%
* Revised scalar inversion formula, with less operations
* Field square function now uses intental loop, saving call overhead
This change will serve as a basis for an arm64 implementation.
Mike Samuel [Tue, 23 Jan 2018 19:27:47 +0000 (14:27 -0500)]
net/http: don't sniff Content-type in Server when X-Content-Type-Options:nosniff
The docs for ResponseWriter.Write say
// If the Header
// does not contain a Content-Type line, Write adds a Content-Type set
// to the result of passing the initial 512 bytes of written data to
// DetectContentType.
The header X-Content-Type-Options:nosniff is an explicit directive that
content-type should not be sniffed.
This changes the behavior of Response.WriteHeader so that, when
there is an X-Content-Type-Options:nosniff header, but there is
no Content-type header, the following happens:
1. A Content-type:application/octet-stream is added
2. A warning is logged via the server's logging mechanism.
Previously, a content-type would have been silently added based on
heuristic analysis of the first 512B which might allow a hosted
GIF like http://www.thinkfu.com/blog/gifjavascript-polyglots to be
categorized as JavaScript which might allow a CSP bypass, loading
as a script despite `Content-Security-Policy: script-src 'self' `.
----
https://fetch.spec.whatwg.org/#x-content-type-options-header
defines the X-Content-Type-Options header.
["Polyglots: Crossing Origins by Crossing Formats"](http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.905.2946&rep=rep1&type=pdf)
explains Polyglot attacks in more detail.
Elias Naur [Sat, 7 Apr 2018 15:22:43 +0000 (17:22 +0200)]
misc/ios: make detect.go more robust
To enable the exec wrapper go_darwin_arm_exec.go to run binaries
on iOS devices, the GOIOS_DEV_ID variable needs to be set to a code
signing identity. The program detect.go attempts to detect suitable
values for GOIOS_DEV_ID (along with GOIOS_APP_ID and GOIOS_TEAM_ID).
Before this change, detect.go would use "security find-identity
-p codesigning -v" to list all available identities for code signing
and pick the first one with "iPhone Developer" in its name. However,
that pick might be invalid since if it was replaced by an identity
issued later.
Instead of attempting to find an identity from the "security
find-identity" list, use the identity from the CommonName in the
embedded certificate in the provisioning file. The CommonName only
lists the identity name (iPhone Developer: xxx@xxx (2754T98W8E)),
not the fingerprint (FC6D96F24A3223C98BF7A2C2C5194D82E04CD23E), but
fortunately the codesign tool accepts both.
Identity names may not be unique, as demonstrated by the example,
but that will result in an ambiguity error at codesigning instead of
a more obscure error about an invalid identity when
go_darwin_arm_exec.go runs a binary.
The fix is then to delete the invalid identity from the system
keychain.
While here, find all connected devices instead of the first connected
and only consider provision files that covers them all. This matters
for the mobile builder where two devices are connected.
Change-Id: I6beb59ace3fc5e071ba76222a20a607765943989
Reviewed-on: https://go-review.googlesource.com/105436
Run-TryBot: Elias Naur <elias.naur@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
After CL 104636 the feature flags in internal/cpu are initialized before
alginit and can now be used for aeshash feature detection. Also remove
now unused runtime variables:
x86:
support_ssse3
support_sse42
support_aes
arm64:
support_aes
Elias Naur [Tue, 10 Apr 2018 11:26:05 +0000 (13:26 +0200)]
misc/ios,runtime/cgo: remove SIGINT handshake for the iOS exec wrapper
Once upon a time, the iOS exec wrapper needed to change the current
working directory for the binary being tested. To allow that, the
runtime raised a SIGINT signal that the wrapper caught, changed the
working directory and resumed the process.
These days, the current working directory is passed from the wrapper
to the runtime through a special entry in the app metadata and the
SIGINT handshake is not necessary anymore.
Remove the signaling from the runtime and the exec harness.
Change-Id: Ia53bcc9e4724d2ca00207e22b91ce80a05271b55
Reviewed-on: https://go-review.googlesource.com/106096
Run-TryBot: Elias Naur <elias.naur@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
internal/cpu,runtime: call cpu.initialize before alginit
runtime.alginit needs runtime/support_{aes,ssse3,sse41} feature flag
to init aeshash function but internal/cpu.init not be called yet.
This CL will call internal/cpu.initialize before runtime.alginit, so
that we can move all cpu features related code to internal/cpu.
Keith Randall [Mon, 9 Apr 2018 23:00:54 +0000 (16:00 -0700)]
runtime: use fixed TLS offsets on darwin/amd64 and darwin/386
Fixes #23617
Note that this CL does not affect darwin/arm and darwin/arm64,
still TBD what, if anything, needs to be done for those.
Change-Id: Ie1ee02a9f4d4d1fb9cd5dc432d900f926cc157db
Reviewed-on: https://go-review.googlesource.com/105975 Reviewed-by: Ian Lance Taylor <iant@golang.org>
Matthew Dempsky [Thu, 5 Apr 2018 19:48:28 +0000 (12:48 -0700)]
cmd/compile: sort method sets using package height
Also, when statically building itabs, compare *types.Sym instead of
name alone so that method sets with duplicate non-exported methods are
handled correctly.
Fixes #24693.
Change-Id: I2db8a3d6e80991a71fef5586a15134b6de116269
Reviewed-on: https://go-review.googlesource.com/105039
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Robert Griesemer <gri@golang.org>
Previously, constant pointer-typed expressions could use either Mpint
or NilVal as their Val depending on their construction, but const.go
expects each type to have a single corresponding Val kind.
This CL changes pointer-typed expressions to exclusively use Mpint.
Fixes #21221.
Change-Id: I6ba36c9b11eb19a68306f0b296acb11a8c254c41
Reviewed-on: https://go-review.googlesource.com/105315
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Robert Griesemer <gri@golang.org>
Matthew Dempsky [Mon, 9 Apr 2018 20:57:56 +0000 (13:57 -0700)]
cmd/compile: refactor symbol sorting logic
This used to be duplicated in methcmp and siglt, because Sig used its
own representation for Syms. Instead, just use Syms, and add a
(*Sym).Less method that both methcmp and siglt can use.
Also, prune some impossible cases purportedly related to blank
methods: the Go spec disallows blank methods in interface method sets,
and addmethod drops blank methods without actually recording them in
the type's method set.
Daniel Martí [Sat, 7 Apr 2018 15:24:52 +0000 (16:24 +0100)]
api: remove unnecessary lines from except.txt
When I added the text/template/parse lines, I thought that both removed
and added APIs should be listed here (i.e. both -pkg and +pkg lines).
However that was wrong, as one can see by reading cmd/api/goapi.go, or
seeing how removing the +pkg lines does not break the API test.
Michael Munday [Mon, 26 Mar 2018 20:18:27 +0000 (21:18 +0100)]
cmd/compile: optimize comparisons using load merging where available
Multi-byte comparison operations were used on amd64, arm64, i386
and s390x for comparisons with constant arrays, but only amd64 and
i386 for comparisons with string constants. This CL combines the
check for platform capability, since they have the same requirements,
and also enables both on ppc64le which also supports load merging.
Note that these optimizations currently use little endian byte order
which results in byte reversal instructions on s390x. This should
be fixed at some point.
Change-Id: Ie612d13359b50c77f4d7c6e73fea4a59fa11f322
Reviewed-on: https://go-review.googlesource.com/102558
Run-TryBot: Michael Munday <mike.munday@ibm.com>
TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
Daniel Martí [Sun, 8 Apr 2018 16:52:16 +0000 (17:52 +0100)]
doc: add a note about loading profile files
If one is somewhat new to the command line or shell, it might be
surprising that changes applied to a file like $HOME/.profile will
seemingly not take effect, even if new shells are started.
Add a note about how shells usually only load these when the user logs
into a machine, to minimize the amount of people stuck and confused by
this.
Fixes #24756.
Change-Id: Ic68d8c97933f3f080b151a107633ecad76a163a4
Reviewed-on: https://go-review.googlesource.com/105557 Reviewed-by: Agniva De Sarker <agniva.quicksilver@gmail.com> Reviewed-by: Bryan C. Mills <bcmills@google.com>
Matthew Dempsky [Tue, 27 Mar 2018 20:50:08 +0000 (13:50 -0700)]
cmd/compile: insert instrumentation during SSA building
Insert appropriate race/msan calls before each memory operation during
SSA construction.
This is conceptually simple, but subtle because we need to be careful
that inserted instrumentation calls don't clobber arguments that are
currently being prepared for a user function call.
reorder1 already handles introducing temporary variables for arguments
in some cases. This CL changes it to use them for all arguments when
instrumenting.
Also, we can't SSA struct types with more than one field while
instrumenting. Otherwise, concurrent uses of disjoint fields within an
SSA-able struct can introduce false races.
This is both somewhat better and somewhat worse than the old racewalk
instrumentation pass. We're now able to easily recognize cases like
constructing non-escaping closures on the stack or accessing closure
variables don't need instrumentation calls. On the other hand,
spilling escaping parameters to the heap now results in an
instrumentation call.
Overall, this CL results in a small net reduction in the number of
instrumentation calls, but a small net increase in binary size for
instrumented executables. cmd/go ends up with 5.6% fewer calls, but a
2.4% larger binary.