Alexander Morozov [Fri, 27 May 2016 22:02:31 +0000 (15:02 -0700)]
syscall: call setgroups for no groups on GNU/Linux
Skip setgroups only for one particular case: GidMappings != nil and
GidMappingsEnableSetgroup == false and list of supplementary groups is
empty.
This patch returns pre-1.5 behavior for simple exec and still allows to
use GidMappings with non-empty Credential.
Change-Id: Ia91c77e76ec5efab7a7f78134ffb529910108fc1
Reviewed-on: https://go-review.googlesource.com/23524 Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Steve Phillips [Thu, 2 Jun 2016 09:40:37 +0000 (02:40 -0700)]
doc/go1.7.html: typo fix; replace "," at end of sentence with "."
Signed-off-by: Steven Phillips <steve@tryingtobeawesome.com>
Change-Id: Ie7c3253a5e1cd43be8fa12bad340204cc6c5ca76
Reviewed-on: https://go-review.googlesource.com/23677 Reviewed-by: Ian Lance Taylor <iant@golang.org>
Alberto Donizetti [Thu, 2 Jun 2016 12:34:37 +0000 (14:34 +0200)]
doc/go1.7: fix typo in nsswitch.conf name
Fixes #15939
Change-Id: I120cbeac73a052fb3f328774e6d5e1534f11bf6b
Reviewed-on: https://go-review.googlesource.com/23682 Reviewed-by: Ian Lance Taylor <iant@golang.org>
Elias Naur [Thu, 2 Jun 2016 13:00:34 +0000 (15:00 +0200)]
misc/cgo/test: fix issue9400 test on android/386
The test for #9400 relies on an assembler function that manipulates
the stack pointer. Meanwile, it uses a global variable for
synchronization. However, position independent code on 386 use a
function call to fetch the base address for global variables.
That function call in turn overwrites the Go stack.
Fix that by fetching the global variable address once before the
stack register manipulation.
Fixes the android/386 builder.
Change-Id: Ib77bd80affaa12f09d582d09d8b84a73bd021b60
Reviewed-on: https://go-review.googlesource.com/23683
Run-TryBot: Elias Naur <elias.naur@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: David Crawshaw <crawshaw@golang.org>
Anmol Sethi [Thu, 2 Jun 2016 02:35:09 +0000 (22:35 -0400)]
net/http: http.Request.Context doc fix
The comment on http.Request.Context says that the context
is canceled when the client's connection closes even though
this has not been implemented. See #15927
Change-Id: I50b68638303dafd70f77f8f778e6caff102d3350
Reviewed-on: https://go-review.googlesource.com/23672 Reviewed-by: Andrew Gerrand <adg@golang.org>
Andrew Gerrand [Thu, 2 Jun 2016 04:31:16 +0000 (14:31 +1000)]
doc: mention net/http/httptrace package in release notes
Updates #15810
Change-Id: I689e18409a88c9e8941aa2e98f472c331efd455e
Reviewed-on: https://go-review.googlesource.com/23674 Reviewed-by: Ian Lance Taylor <iant@golang.org>
Ian Lance Taylor [Thu, 2 Jun 2016 04:09:58 +0000 (21:09 -0700)]
doc/go1.7.html: fix spelling of cancelation
We say "cancelation," not "cancellation."
Fixes #15928.
Change-Id: I66d545404665948a27281133cb9050eebf1debbb
Reviewed-on: https://go-review.googlesource.com/23673 Reviewed-by: Ian Lance Taylor <iant@golang.org>
Michael Hudson-Doyle [Fri, 27 May 2016 03:41:55 +0000 (15:41 +1200)]
cmd/compile: do not generate tail calls when dynamic linking on ppc64le
When a wrapper method calls the real implementation, it's not possible to use a
tail call when dynamic linking on ppc64le. The bad scenario is when a local
call is made to the wrapper: the wrapper will call the implementation, which
might be in a different module and so set the TOC to the appropriate value for
that module. But if it returns directly to the wrapper's caller, nothing will
reset it to the correct value for that function.
Change-Id: Icebf24c9a2a0a9a7c2bce6bd6f1358657284fb10
Reviewed-on: https://go-review.googlesource.com/23468 Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Michael Hudson-Doyle <michael.hudson@canonical.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Mikio Hara [Wed, 1 Jun 2016 23:53:11 +0000 (08:53 +0900)]
vendor: update vendored route
Updates golang.org/x/net/route to rev fac978c for:
- route: fix typos in test
Change-Id: I35de1d3f8e887c6bb5fe50e7299f2fc12e4426de
Reviewed-on: https://go-review.googlesource.com/23660 Reviewed-by: Ian Lance Taylor <iant@golang.org>
Adam Langley [Wed, 1 Jun 2016 21:41:09 +0000 (14:41 -0700)]
crypto/tls: buffer handshake messages.
This change causes TLS handshake messages to be buffered and written in
a single Write to the underlying net.Conn.
There are two reasons to want to do this:
Firstly, it's slightly preferable to do this in order to save sending
several, small packets over the network where a single one will do.
Secondly, since 37c28759ca46cf381a466e32168a793165d9c9e9 errors from
Write have been returned from a handshake. This means that, if a peer
closes the connection during a handshake, a “broken pipe” error may
result from tls.Conn.Handshake(). This can mask any, more detailed,
fatal alerts that the peer may have sent because a read will never
happen.
Buffering handshake messages means that the peer will not receive, and
possibly reject, any of a flow while it's still being written.
Fixes #15709
Change-Id: I38dcff1abecc06e52b2de647ea98713ce0fb9a21
Reviewed-on: https://go-review.googlesource.com/23609 Reviewed-by: Andrew Gerrand <adg@golang.org>
Run-TryBot: Andrew Gerrand <adg@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Tom Bergan [Fri, 27 May 2016 23:53:13 +0000 (16:53 -0700)]
net/http: update bundled http2
Updates x/net/http2 to git rev 6bdd4be4 for CL 23526:
http2: GotFirstResponseByte hook should only fire once
Also updated the trace hooks test to verify that all trace hooks are called
exactly once except ConnectStart/End, which may be called multiple times (due
to happy-eyeballs).
Fixes #15777
Change-Id: Iea5c64eb322b58be27f9ff863b3a6f90e996fa9b
Reviewed-on: https://go-review.googlesource.com/23527 Reviewed-by: Andrew Gerrand <adg@golang.org>
David du Colombier [Wed, 1 Jun 2016 13:13:55 +0000 (15:13 +0200)]
cmd/compile: fix TestAssembly on Plan 9
Since CL 23620, TestAssembly is failing on Plan 9.
In CL 23620, the process environment is passed to 'go tool compile'
after setting GOARCH. On Plan 9, if GOARCH is already set in the
process environment, it would take precedence. On Unix, it works
as expected because the first GOARCH found takes precedence.
This change uses the mergeEnvLists function from cmd/go/main.go
to merge the two environment lists such that variables with the
same name in "in" replace those in "out".
Change-Id: Idee22058343932ee18666dda331c562c89c33507
Reviewed-on: https://go-review.googlesource.com/23593 Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: David du Colombier <0intro@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Dan Peterson [Wed, 1 Jun 2016 12:44:38 +0000 (09:44 -0300)]
doc: rename Unshare to Unshareflags in go1.7 release notes
Implementation changed in https://golang.org/cl/23612.
Updates #15810
Change-Id: I8fff9e3aa3e54162546bb9ec1cc2ebba2b6d9fed
Reviewed-on: https://go-review.googlesource.com/23614 Reviewed-by: Ian Lance Taylor <iant@golang.org>
Alexander Morozov [Wed, 1 Jun 2016 02:44:48 +0000 (19:44 -0700)]
syscall: rename SysProcAttr.Unshare to Unshareflags
For symmetry with Cloneflags and it looks slightly weird because there
is syscall.Unshare method.
Change-Id: I3d710177ca8f27c05b344407f212cbbe3435094b
Reviewed-on: https://go-review.googlesource.com/23612 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: Rob Pike <r@golang.org>
Kenny Grant [Tue, 31 May 2016 21:30:37 +0000 (22:30 +0100)]
context: fix typo in comments
Change-Id: I41310ec88c889fda79d80eaf4a742a1000284f60
Reviewed-on: https://go-review.googlesource.com/23591 Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Robert Griesemer [Sat, 28 May 2016 02:47:55 +0000 (19:47 -0700)]
crypto/tls: remove unused variable in benchmark code
This fixes `go test go/types`.
https://golang.org/cl/23487/ introduced this code which contains
two unused variables (declared and assigned to, but never read).
cmd/compile doesn't report the error due open issue #8560 (the
variables are assigned to in a closure), but go/types does. The
build bot only runs go/types tests in -short mode (which doesn't
typecheck the std lib), hence this doesn't show up on the dashboard
either.
We cannot call b.Fatal and friends in the goroutine. Communicating
the error to the invoking function requires a channel or a mutex.
Unless the channel/sycnhronized variable is tested in each iteration
that follows, the iteration blocks if there's a failure. Testing in
each iteration may affect benchmark times.
One could use a time-out but that time depends on the underlying system.
Panicking seems good enough in this unlikely case; better than hanging
or affecting benchmark times.
Change-Id: Idce1172da8058e580fa3b3e398825b0eb4316325
Reviewed-on: https://go-review.googlesource.com/23528 Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Robert Griesemer [Tue, 31 May 2016 20:32:34 +0000 (13:32 -0700)]
spec: document that duplicate types are invalid in type switches
Both compilers and also go/types don't permit duplicate types in
type switches; i.e., this spec change is documenting a status quo
that has existed for some time.
Furthermore, duplicate nils are not accepted by gccgo or go/types;
and more recently started causing a compiler error in gc. Permitting
them is inconsistent with the existing status quo.
Rather than making it an implementation restriction (as we have for
expression switches), this is a hard requirement since it was enforced
from the beginning (except for duplicate nils); it is also a well
specified requirement that does not pose a significant burden for
an implementation.
Fixes #15896.
Change-Id: If12db5bafa87598b323ea84418cb05421e657dd8
Reviewed-on: https://go-review.googlesource.com/23584 Reviewed-by: Rob Pike <r@golang.org> Reviewed-by: Ian Lance Taylor <iant@golang.org>
Ian Lance Taylor [Tue, 31 May 2016 18:23:50 +0000 (11:23 -0700)]
flag: recognize "0s" as the zero value for a flag.Duration
Implemented by using a reflect-based approach to recognize the zero
value of any non-interface type that implements flag.Value. Interface
types will fall back to the old code.
Fixes #15904.
Change-Id: I594c3bfb30e9ab1aca3e008ef7f70be20aa41a0b
Reviewed-on: https://go-review.googlesource.com/23581
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Rob Pike <r@golang.org>
Ian Lance Taylor [Fri, 27 May 2016 17:05:52 +0000 (10:05 -0700)]
runtime: pass signal context to cgo traceback function
When doing a backtrace from a signal that occurs in C code compiled
without using -fasynchronous-unwind-tables, we have to rely on frame
pointers. In order to do that, the traceback function needs the signal
context to reliably pick up the frame pointer.
Change-Id: I7b45930fced01685c337d108e0f146057928f876
Reviewed-on: https://go-review.googlesource.com/23494
Run-TryBot: Ian Lance Taylor <iant@golang.org> Reviewed-by: Austin Clements <austin@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Joe Tsai [Tue, 31 May 2016 19:42:34 +0000 (12:42 -0700)]
doc/go1.7.html: make RFC an actual link
Change-Id: I5e8dad0c2534b5c3654cf0a0b51a38186d627a3c
Reviewed-on: https://go-review.googlesource.com/23582 Reviewed-by: Ian Lance Taylor <iant@golang.org>
Ian Lance Taylor [Fri, 27 May 2016 00:47:03 +0000 (17:47 -0700)]
runtime/cgo: add TSAN acquire/release calls
Add TSAN acquire/release calls to runtime/cgo to match the ones
generated by cgo. This avoids a false positive race around the malloc
memory used in runtime/cgo when other goroutines are simultaneously
calling malloc and free from cgo.
These new calls will only be used when building with CGO_CFLAGS and
CGO_LDFLAGS set to -fsanitize=thread, which becomes a requirement to
avoid all false positives when using TSAN. These are needed not just
for runtime/cgo, but also for any runtime package that uses cgo (such as
net and os/user).
Add an unused attribute to the _cgo_tsan_acquire and _cgo_tsan_release
functions, in case there are no actual cgo function calls.
Add a test that checks that setting CGO_CFLAGS/CGO_LDFLAGS avoids a
false positive report when using os/user.
Change-Id: I0905c644ff7f003b6718aac782393fa219514c48
Reviewed-on: https://go-review.googlesource.com/23492
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Dmitry Vyukov <dvyukov@google.com>
Joe Tsai [Sun, 29 May 2016 07:47:45 +0000 (00:47 -0700)]
compress/flate: use seperate const block for exported constants
As rendered on https://tip.golang.org/pkg/compress/flate/, there is an
extra new-line because of the unexported constants in the same block.
<<<
const (
NoCompression = 0
BestSpeed = 1
BestCompression = 9
DefaultCompression = -1
HuffmanOnly = -2 // Disables match search and only does Huffman entropy reduction.
)
>>>
Instead, seperate the exported compression level constants into its own
const block. This is both more readable and also fixes the issue.
Change-Id: I60b7966c83fb53356c02e4640d05f55a3bee35b7
Reviewed-on: https://go-review.googlesource.com/23557 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, 27 May 2016 23:03:44 +0000 (16:03 -0700)]
runtime/pprof, cmd/pprof: fix profiling for PIE
In order to support pprof for position independent executables, pprof
needs to adjust the PC addresses stored in the profile by the address at
which the program is loaded. The legacy profiling support which we use
already supports recording the GNU/Linux /proc/self/maps data
immediately after the CPU samples, so do that. Also change the pprof
symbolizer to use the information, if available, when looking up
addresses in the Go pcline data.
Fixes #15714.
Change-Id: I4bf679210ef7c51d85cf873c968ce82db8898e3e
Reviewed-on: https://go-review.googlesource.com/23525 Reviewed-by: Michael Hudson-Doyle <michael.hudson@canonical.com>
Andrew Gerrand [Tue, 31 May 2016 03:21:35 +0000 (13:21 +1000)]
crypto/tls: reduce size of buffer in throughput benchmarks
The Windows builders run the throughput benchmarks really slowly with a
64kb buffer. Lowering it to 16kb brings the performance back into line
with the other builders.
This is a work-around to get the build green until we can figure out why
the Windows builders are slow with the larger buffer size.
Update #15899
Change-Id: I215ebf115e8295295c87f3b3e22a4ef1f9e77f81
Reviewed-on: https://go-review.googlesource.com/23574 Reviewed-by: Ian Lance Taylor <iant@golang.org>
Richard Miller [Sat, 28 May 2016 09:06:37 +0000 (10:06 +0100)]
syscall: plan9 - mark gbit16 as go:nosplit
This is a correction to CL 22610. The gbit16 function is called in
StartProcess between fork and exec, and therefore must not split the
stack. Normally it's inlined so this is not an issue, but on one
occasion I've observed it to be compiled without inlining, and the
result was a panic. Mark it go:nosplit to be safe.
Change-Id: I0381754397b766431bf406d9767c73598d23b901
Reviewed-on: https://go-review.googlesource.com/23560 Reviewed-by: David du Colombier <0intro@gmail.com>
Run-TryBot: David du Colombier <0intro@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Minux Ma <minux@golang.org> Reviewed-by: Ian Lance Taylor <iant@golang.org>
Augusto Roman [Sat, 28 May 2016 23:59:28 +0000 (16:59 -0700)]
doc: correct release notes for non-string map keys in encoding/json
The original draft mentioned support for json.Marshaler, but that's
not the case. JSON supports only string keys (not arbitrary JSON)
so only encoding.TextMarshaller is supported.
Change-Id: I7788fc23ac357da88e92aa0ca17b513260840cee
Reviewed-on: https://go-review.googlesource.com/23529 Reviewed-by: Andrew Gerrand <adg@golang.org>
Joe Tsai [Sat, 28 May 2016 09:14:25 +0000 (02:14 -0700)]
io: use SeekStart, SeekCurrent, and SeekEnd in io.Seeker documentation
The documentation previously used C style enumerations: 0, 1, 2.
While this is pretty much universally correct, it does not help a user
become aware of the existence of the SeekStart, SeekCurrent, and SeekEnd
constants. Thus, we should use them in the documentation to direct people's
attention to them.
Updates #6885
Change-Id: I44b5e78d41601c68a0a1c96428c853df53981d52
Reviewed-on: https://go-review.googlesource.com/23551 Reviewed-by: Andrew Gerrand <adg@golang.org>
Document the following:
* That the algorithmic changes are still compliant with RFC 1951. I remember
people having questions regarding this issue, and it would be good to re-assure
them that it is still standards compliant.
* io.EOF can now be returned early (c27efce66bce7534dbb357ac1779bbc08395b267)
* Use the term "decompress" when referred to as an action. The term "uncompressed"
or "decompressed" are both valid as ways to represent the current state of the data.
Change-Id: Ie29ebce709357359e7c36d3e7f3d53b260eaadfa
Reviewed-on: https://go-review.googlesource.com/23552 Reviewed-by: Andrew Gerrand <adg@golang.org>
Keith Randall [Fri, 27 May 2016 21:07:37 +0000 (14:07 -0700)]
cmd/compile: SSA, don't let write barrier clobber return values
When we do *p = f(), we might need to copy the return value from
f to p with a write barrier. The write barrier itself is a call,
so we need to copy the return value of f to a temporary location
before we call the write barrier function. Otherwise, the call
itself (specifically, marshalling the args to typedmemmove) will
clobber the value we're trying to write.
Fixes #15854
Change-Id: I5703da87634d91a9884e3ec098d7b3af713462e7
Reviewed-on: https://go-review.googlesource.com/23522 Reviewed-by: David Chase <drchase@google.com>
Run-TryBot: Keith Randall <khr@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Mikio Hara [Fri, 27 May 2016 08:35:45 +0000 (17:35 +0900)]
net/http/httptrace: fix nit in test
Change-Id: I6dc3666398b4cd7a7195bb9c0e321fa8b733fa15
Reviewed-on: https://go-review.googlesource.com/23502 Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Mikio Hara [Fri, 27 May 2016 08:34:22 +0000 (17:34 +0900)]
runtime: skip TestGdbBacktrace on netbsd
Also adds missing copyright notice.
Updates #15603.
Change-Id: Icf4bb45ba5edec891491fe5f0039a8a25125d168
Reviewed-on: https://go-review.googlesource.com/23501 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 [Fri, 27 May 2016 16:21:14 +0000 (12:21 -0400)]
runtime: always call stackfree on the system stack
Currently when the garbage collector frees stacks of dead goroutines
in markrootFreeGStacks, it calls stackfree on a regular user stack.
This is a problem, since stackfree manipulates the stack cache in the
per-P mcache, so if it grows the stack or gets preempted in the middle
of manipulating the stack cache (which are both possible since it's on
a user stack), it can easily corrupt the stack cache.
Fix this by calling markrootFreeGStacks on the system stack, so that
all calls to stackfree happen on the system stack. To prevent this bug
in the future, mark stack functions that manipulate the mcache as
go:systemstack.
Fixes #15853.
Change-Id: Ic0d1c181efb342f134285a152560c3a074f14a3d
Reviewed-on: https://go-review.googlesource.com/23511
Run-TryBot: Austin Clements <austin@google.com> Reviewed-by: Keith Randall <khr@golang.org> Reviewed-by: Rick Hudson <rlh@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Russ Cox [Fri, 27 May 2016 13:50:06 +0000 (09:50 -0400)]
crypto/tls: adjust dynamic record sizes to grow arithmetically
The current code, introduced after Go 1.6 to improve latency on
low-bandwidth connections, sends 1 kB packets until 1 MB has been sent,
and then sends 16 kB packets (the maximum record size).
Unfortunately this decreases throughput for 1-16 MB responses by 20% or so.
Following discussion on #15713, change cutoff to 128 kB sent
and also grow the size allowed for successive packets:
1 kB, 2 kB, 3 kB, ..., 15 kB, 16 kB.
This fixes the throughput problems: the overhead is now closer to 2%.
I hope this still helps with latency but I don't have a great way to test it.
At the least, it's not worse than Go 1.6.
The small MB runs are bimodal in both cases, probably GC pauses.
But there's clearly no general slowdown anymore.
Fixes #15713.
Change-Id: I5fc44680ba71812d24baac142bceee0e23f2e382
Reviewed-on: https://go-review.googlesource.com/23487 Reviewed-by: Ian Lance Taylor <iant@golang.org>
Russ Cox [Fri, 27 May 2016 16:15:04 +0000 (12:15 -0400)]
doc/go1.7.html: fix broken sentence
Change-Id: Ia540c890767dcb001d3b3b55d98d9517b13b21da
Reviewed-on: https://go-review.googlesource.com/23510 Reviewed-by: Ian Lance Taylor <iant@golang.org>
Russ Cox [Fri, 27 May 2016 15:05:14 +0000 (11:05 -0400)]
net/http: change Transport.Dialer to Transport.DialContext
New in Go 1.7 so still possible to change.
This allows implementations not tied to *net.Dialer.
Fixes #15748.
Change-Id: I5fabbf13c7f1951c06587a4ccd120def488267ce
Reviewed-on: https://go-review.googlesource.com/23489 Reviewed-by: Ian Lance Taylor <iant@golang.org>
Russ Cox [Wed, 25 May 2016 14:01:58 +0000 (10:01 -0400)]
cmd/compile: additional paranoia and checking in plive.go
The main check here is that liveness now crashes if it finds an instruction
using a variable that should be tracked but is not.
Comments and adjustments in nodarg to explain what's going on and
to remove the "-1" argument added a few months ago, plus a sketch
of a future simplification.
The need for n.Orig in the earlier CL seems to have been an intermediate
problem rather than fundamental: the new explanations in nodarg make
clear that nodarg is not causing the problem I thought, and in fact now
using n instead of n.Orig works fine in plive.go.
Dmitri Shuralyov [Thu, 26 May 2016 06:22:11 +0000 (23:22 -0700)]
cmd/go: fixup for parsing SCP-like addresses
This is a fixup change for commit 5cd294480364eb166751838a3df8f58649c214e1
that added parsing of SCP-like addresses. To get the expected output
from (*url.URL).String(), Path needs to be set, not RawPath.
Add a test for this, since it has already regressed multiple times.
Updates #11457.
Change-Id: I806f5abbd3cf65e5bdcef01aab872caa8a5b8891
Reviewed-on: https://go-review.googlesource.com/23447
Run-TryBot: Andrew Gerrand <adg@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Andrew Gerrand <adg@golang.org>
Russ Cox [Wed, 25 May 2016 14:29:50 +0000 (10:29 -0400)]
cmd/compile: eliminate PPARAMREF
As in the elimination of PHEAP|PPARAM in CL 23393,
this is something the front end can trivially take care of
and then not bother the back ends with.
It also eliminates some suspect (and only lightly exercised)
code paths in the back ends.
I don't have a smoking gun for this one but it seems
more clearly correct.
Russ Cox [Wed, 25 May 2016 05:33:24 +0000 (01:33 -0400)]
cmd/compile: fix liveness computation for heap-escaped parameters
The liveness computation of parameters generally was never
correct, but forcing all parameters to be live throughout the
function covered up that problem. The new SSA back end is
too clever: even though it currently keeps the parameter values live
throughout the function, it may find optimizations that mean
the current values are not written back to the original parameter
stack slots immediately or ever (for example if a parameter is set
to nil, SSA constant propagation may replace all later uses of the
parameter with a constant nil, eliminating the need to write the nil
value back to the stack slot), so the liveness code must now
track the actual operations on the stack slots, exposing these
problems.
One small problem in the handling of arguments is that nodarg
can return ONAME PPARAM nodes with adjusted offsets, so that
there are actually multiple *Node pointers for the same parameter
in the instruction stream. This might be possible to correct, but
not in this CL. For now, we fix this by using n.Orig instead of n
when considering PPARAM and PPARAMOUT nodes.
The major problem in the handling of arguments is general
confusion in the liveness code about the meaning of PPARAM|PHEAP
and PPARAMOUT|PHEAP nodes, especially as contrasted with PAUTO|PHEAP.
The difference between these two is that when a local variable "moves"
to the heap, it's really just allocated there to start with; in contrast,
when an argument moves to the heap, the actual data has to be copied
there from the stack at the beginning of the function, and when a
result "moves" to the heap the value in the heap has to be copied
back to the stack when the function returns
This general confusion is also present in the SSA back end.
The PHEAP bit worked decently when I first introduced it 7 years ago (!)
in 391425ae. The back end did nothing sophisticated, and in particular
there was no analysis at all: no escape analysis, no liveness analysis,
and certainly no SSA back end. But the complications caused in the
various downstream consumers suggest that this should be a detail
kept mainly in the front end.
This CL therefore eliminates both the PHEAP bit and even the idea of
"heap variables" from the back ends.
First, it replaces the PPARAM|PHEAP, PPARAMOUT|PHEAP, and PAUTO|PHEAP
variable classes with the single PAUTOHEAP, a pseudo-class indicating
a variable maintained on the heap and available by indirecting a
local variable kept on the stack (a plain PAUTO).
Second, walkexpr replaces all references to PAUTOHEAP variables
with indirections of the corresponding PAUTO variable.
The back ends and the liveness code now just see plain indirected
variables. This may actually produce better code, but the real goal
here is to eliminate these little-used and somewhat suspect code
paths in the back end analyses.
The OPARAM node type goes away too.
A followup CL will do the same to PPARAMREF. I'm not sure that
the back ends (SSA in particular) are handling those right either,
and with the framework established in this CL that change is trivial
and the result clearly more correct.
Ian Lance Taylor [Thu, 19 May 2016 01:42:25 +0000 (18:42 -0700)]
cmd/cgo: remove -O options when generating compiler errors
The cgo tool generates compiler errors to find out what kind of name it
is using. Turning on optimization can confuse that process by producing
new unexpected messages.
Fixes #14669.
Change-Id: Idc8e35fd259711ecc9638566b691c11d17140325
Reviewed-on: https://go-review.googlesource.com/23231
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Russ Cox <rsc@golang.org>
Keith Randall [Tue, 24 May 2016 21:09:02 +0000 (14:09 -0700)]
cmd/compile: testing harness for checking generated assembly
Add a test which compiles a function and checks the
generated assembly to make sure certain patterns are present.
This test allows us to do white box tests of the compiler
to make sure optimizations don't regress.
Added a few simple tests for now. More to come.
Change-Id: I4ab5ce5d95b9e04e7d0d9328ffae47b8d1f95e74
Reviewed-on: https://go-review.googlesource.com/23403 Reviewed-by: David Chase <drchase@google.com>
Run-TryBot: Keith Randall <khr@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Quentin Smith [Thu, 26 May 2016 21:53:21 +0000 (17:53 -0400)]
encoding/json: improve Decode example
Decoding a JSON message does not touch unspecified or null fields;
always use a new underlying struct to prevent old field values from
sticking around.
Fixes: #14640
Change-Id: Ica78c208ce104e2cdee1d4e92bf58596ea5587c8
Reviewed-on: https://go-review.googlesource.com/23483 Reviewed-by: Andrew Gerrand <adg@golang.org>
domorder has some non-obvious useful properties
that we’re relying on in cse.
Document them and provide an argument that they hold.
While we’re here, do some minor renaming.
The argument is a re-working of a private email
exchange with Todd Neal and David Chase.
Russ Cox [Wed, 25 May 2016 18:37:43 +0000 (14:37 -0400)]
build: enable framepointer mode by default
This has a minor performance cost, but far less than is being gained by SSA.
As an experiment, enable it during the Go 1.7 beta.
Having frame pointers on by default makes Linux's perf, Intel VTune,
and other profilers much more useful, because it lets them gather a
stack trace efficiently on profiling events.
(It doesn't help us that much, since when we walk the stack we usually
need to look up PC-specific information as well.)
Austin Clements [Thu, 26 May 2016 00:56:56 +0000 (20:56 -0400)]
runtime: unwind BP in jmpdefer to match SP unwind
The irregular calling convention for defers currently incorrectly
manages the BP if frame pointers are enabled. Specifically, jmpdefer
manipulates the SP as if its own caller, deferreturn, had returned.
However, it does not manipulate the BP to match. As a result, when a
BP-based traceback happens during a deferred function call, it unwinds
to the function that performed the defer and then thinks that function
called itself in an infinite regress.
Fix this by making jmpdefer manipulate the BP as if deferreturn had
actually returned.
Austin Clements [Thu, 26 May 2016 02:59:19 +0000 (22:59 -0400)]
cmd/link/internal/ld: fix DWARF offsets with GOEXPERIMENT=framepointer
The offsets computed by the DWARF expressions for local variables
currently don't account for the extra stack slot used by the frame
pointer when GOEXPERIMENT=framepointer is enabled.
Fix this by adding the extra stack slot to the offset.
This fixes TestGdbPython with GOEXPERIMENT=framepointer.
Russ Cox [Thu, 26 May 2016 00:01:25 +0000 (20:01 -0400)]
runtime: make framepointer mode safe for Windows
A few other architectures have already defined a NOFRAME flag.
Use it to disable frame pointer code on a few very low-level functions
that must behave like Windows code.
Makes the failing os/signal test pass on a Windows gomote.