This works well enough to run some code natively on arm64, but not well enough for more complicated code. I've been suggested to start a pull request anyway.
Bryan C. Mills [Thu, 28 Mar 2019 17:18:37 +0000 (13:18 -0400)]
cmd/go: only add a 'go' directive on 'go mod tidy' or when a conversion occurs
If the go.mod file exists and is empty, we initialize it from any of
various formats supported by legacy dependency-management tools.
We also initialize the 'go' directive at that point: we know that the
go.mod file is incomplete, because it does not reflect the information
in the legacy configuration file, and since we know that the go.mod
file is incomplete, we should complete it with as much information as
we have — including the version of the language currently in use.
However, if there is no legacy configuration file present, then we
cannot infer that the go.mod file is incomplete: it may correctly
specify a module without external dependencies. In that case, we
should not initialize the 'go' directive either: the user will not be
expecting unnecessary edits to the go.mod file, and we generally do
not make unnecessary-but-helpful edits unless 'go mod tidy' is invoked
explicitly.
Fixes #30790
Fixes #31100
Change-Id: I05a7872bce54a917c10d910cd9a616cab52e2730
Reviewed-on: https://go-review.googlesource.com/c/go/+/169877
Run-TryBot: Bryan C. Mills <bcmills@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Ian Lance Taylor <iant@golang.org>
Matthew Dempsky [Fri, 19 Apr 2019 19:20:56 +0000 (12:20 -0700)]
cmd/compile: fix ICE from go/defer call to variadic function
The special case logic for go/defer arguments in Escape.call was
scattered around a bit and was somewhat inconsistently handled across
different types of function calls and parameters. This CL pulls the
logic out into a separate callStmt method that's used uniformly for
all kinds of function calls and arguments.
David Chase [Tue, 16 Apr 2019 01:27:04 +0000 (21:27 -0400)]
cmd/compile: shortcut intrinsic inlining AFTER getcallerXX check
A check in inl.go to prevent inlining of functions calling
either getcallerpc or getcallersp does not work when these
functions are intrinsics. Swap checks to fix.
Includes test.
No bug, this was discovered in the course of a ridiculous
experiment with inlining.
Change-Id: Ie1392523bb89882d586678f2674e1a4eadc5e431
Reviewed-on: https://go-review.googlesource.com/c/go/+/172217
Run-TryBot: David Chase <drchase@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
Than McIntosh [Thu, 18 Apr 2019 13:45:40 +0000 (09:45 -0400)]
go/internal/gccgoimporter: improve alias handling for anonymous fields
The code in the parser that deals with anonymous structure fields
records the fact that a field is anonymous, then tries to install a proxy
name for the field based on the name of the type used to declare
the field. If that type was an alias, the current recipe for determining
the proxy name was not working properly; enhance the code to recover
and report the alias name used.
Fixes #31540.
Change-Id: I9b7369ed558a288b56d85170c6f1144daf5228eb
Reviewed-on: https://go-review.googlesource.com/c/go/+/172603 Reviewed-by: Ian Lance Taylor <iant@golang.org>
Than McIntosh [Fri, 19 Apr 2019 18:50:57 +0000 (14:50 -0400)]
cmd/link: adjust whitelist for -strictdups checking for plan9
Add a couple of additional entries to the white list used to screen
out errors for builtin functions; these correspond to cases
that appear to come up only on the plan9 builder.
Updates #31503.
Change-Id: I48ab942ab2894240efe651ec7b7eace7aa5cb45e
Reviewed-on: https://go-review.googlesource.com/c/go/+/172986 Reviewed-by: David du Colombier <0intro@gmail.com> Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Than McIntosh <thanm@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
os: disable the use of netpoll on directories as well on *BSDs
Follow up CL 156379.
Updates #19093
Change-Id: I5ea3177fc5911d3af71cbb32584249e419e9d4a3
Reviewed-on: https://go-review.googlesource.com/c/go/+/172937
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Ian Lance Taylor <iant@golang.org>
cmd/go/internal/modload: fix boundary conditions in matchPackages
This makes the boundary logic of matchPackages consistent with
modload.dirInModule.
Previously, matchPackages always stopped at go.mod file, even within
the vendor tree. However, we do not guarantee that the vendor tree is
free of such files in general.
matchPackages also issued needless stat operations for modules in the
module cach, which we already know to be free of nested modules. On
systems with slow filesystems (such as macOS), those extra calls could
potentially slow package matching considerably.
Change-Id: I71979ab752e1d3971b370b37085d30502690413b
Reviewed-on: https://go-review.googlesource.com/c/go/+/172985
Run-TryBot: Bryan C. Mills <bcmills@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Jay Conrod <jayconrod@google.com>
This resurrects CL 121198, except that this time we map read-only.
In case that we need to apply relocations to the symbol's
content that is backed by read-only memory, we do our own copy-
on-write. This can happen if we failed to mmap the output file,
or we build for Wasm.
Memory profile for building k8s.io/kubernetes/cmd/kube-apiserver
on Linux/AMD64:
Old (before this sequence of CLs):
inuse_space 1598.75MB total
669.87MB 41.90% 41.90% 669.87MB 41.90% cmd/link/internal/objfile.(*objReader).readSlices
New:
inuse_space 1280.45MB total
441.18MB 34.46% 34.46% 441.18MB 34.46% cmd/link/internal/objfile.(*objReader).readSlices
Move the phase of applying relocations later, after the sections
and segments are written to the mmap'd output region. Then apply
relocations directly in the output region, instead of the input.
So the input slices we read in don't need to be modified.
This is in preparation for mmap'ing input files read-only.
cmd/link: apply DWARF relocations while doing compression
We are preparing for applying relocations to the output buffer.
However, for DWARF compression, relocations need to be applied
before compression, but we don't have an output buffer at that
time. We also cannot delay DWARF compression to when we mmap the
output file, because we need the size of the DWARF sections to
compute the file size.
Instead of applying all the relocations together, we apply
relocations in DWARF sections one symbol at a time, right before
it is writing out for compression. As the symbol content may be
in read-only memory (in the future), we use a temporary buffer
for applying the relocations, and immediately write it out.
If compression is not used, relocations are still applied all
together.
This is in preparation for mmap'ing input files read-only.
Apply R_DWARFFILEREF relocations later, along with other
relocations, so that we don't modify symbols' contents before
they are written to the output buffer.
This is in preparation for mmap'ing input files read-only.
Use mmap for writing most of the output file content,
specifically, the sections and segments. After layout, we
already know the sizes and file offsets for the sections and
segments. So we can just write the bytes by copying to a mmap'd
backing store.
The writing of the output file is split into two parts. The first
part writes the sections and segments to the mmap'd region. The
second part writes some extra content, for which we don't know
the size, so we use direct file IO.
This is in preparation for mmap'ing input files read-only.
Pass "set print thread-events off" to gdb to suppress thread
event prints, like "[New Thread 0xe7b83b40 (LWP 18609)]". We
don't check them, and the extra output may confuse our other
checks, in particular, checkCleanBacktrace.
Ian Lance Taylor [Fri, 19 Apr 2019 16:48:36 +0000 (09:48 -0700)]
bootstrap.bash: make source writable before cleaning
Otherwise the "git clean" command fails with errors like
rm: cannot remove '/home/iant/go-linux-ppc64-bootstrap/pkg/mod/golang.org/x/text@v0.0.0-20170915032832-14c0d48ead0c/encoding/simplifiedchinese/all.go': Permission denied
Change-Id: Iecfb1fed6d59819d7fdceb9e391a2b3f81ea620c
Reviewed-on: https://go-review.googlesource.com/c/go/+/172998
Run-TryBot: Ian Lance Taylor <iant@golang.org> Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Ian Lance Taylor [Fri, 19 Apr 2019 16:50:01 +0000 (09:50 -0700)]
cmd/link: require cgo support for TestSectionsWithSameName
The test doesn't really require cgo, but it does require that we know
the right flags to use to run the C compiler, and that is not
necessarily correct if we don't support cgo.
Fixes #31565
Change-Id: I04dc8db26697caa470e91ad712376aa621cf765d
Reviewed-on: https://go-review.googlesource.com/c/go/+/172981
Run-TryBot: Ian Lance Taylor <iant@golang.org> Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
runtime, cmd/compile: re-order PCDATA and FUNCDATA indices
The pclntab encoding supports writing only some PCDATA and FUNCDATA values.
However, the encoding is dense: The max index in use determines the space used.
We should thus choose a numbering in which frequently used indices are smaller.
This change re-orders the PCDATA and FUNCDATA indices using that principle,
using a quick and dirty instrumentation to measure index frequency.
The phi tighten pass moves rematerializable phi args
to the immediate predecessor of the phis.
This reduces value lifetimes for regalloc.
However, the critical edge removal pass can introduce
new blocks, which can change what a block's
immediate precedessor is. This can result in tightened
phi args being spilled unnecessarily.
This change moves the phi tighten pass after the
critical edge pass, when the block structure is stable.
cmd/go/internal/modcmd: skip files with the "ignore" constraint in 'go mod vendor'
'go mod vendor' already drops test files and testdata directories, so
users should not expect the vendored module to include unnecessary
files in general.
Files tagged "ignore" are typically only used to refresh or regenerate
source files within the module to be vendored, so users of that module
do not need them.
Running `go mod init` outside of GOPATH with `GO111MODULE=off`
silently fails. This behavior was undocumented.
This CL makes go mod fail with the error:
go: modules disabled by GO111MODULE=off; see 'go help modules'
Comparing with already erroring GO111MODULE=<value> conditions:
* With GO111MODULE=auto, inside GOPATH:
go modules disabled inside GOPATH/src by GO111MODULE=auto; see 'go help modules'
* With GO111MODULE=auto outside of GOPATH:
go: cannot determine module path for source directory /path/to/dir (outside GOPATH, no import comments)
Ian Lance Taylor [Thu, 18 Apr 2019 05:41:51 +0000 (22:41 -0700)]
cmd/link: don't fail if multiple ELF sections have the same name
New versions of clang can generate multiple sections named ".text"
when using vague C++ linkage. This is valid ELF, but would cause the
Go linker to report an error when using internal linking:
symbol PACKAGEPATH(.text) listed multiple times
Avoid the problem by renaming section symbol names if there is a name
collision.
Change-Id: I41127e95003d5b4554aaf849177b3fe000382c02
Reviewed-on: https://go-review.googlesource.com/c/go/+/172697
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Cherry Zhang <cherryyz@google.com>
Rob Pike [Thu, 18 Apr 2019 03:28:55 +0000 (13:28 +1000)]
strconv: pre-allocate in appendQuotedWith
The byte-at-a-time allocation done quoting strings in appendQuotedWith
grows the output incrementally, which is poor behavior for very large
strings. An easy fix is to make sure the buffer has enough room at
least for an unquoted string.
Add a benchmark with a megabyte of non-ASCII data.
Before: 39 allocations.
After: 7 allocations.
We could do better by doing a lot more work but this seems like a big
result for little effort.
Fixes #31472.
Change-Id: I852139e0a2bd13722c4dd329ded8ae1759abad5b
Reviewed-on: https://go-review.googlesource.com/c/go/+/172677 Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
runtime: move libcall to stack for runtime syscalls on AIX
As the stackguard was increased on AIX by CL 157117, every syscalls can
now have libcall directly on the stack. This fixes some concurrency bugs
which seems to occur when semasleep is interrupted by a SIGPROF signal.
Change-Id: I905a9618d13ef227dad6f8328b0f958f2f917a5b
Reviewed-on: https://go-review.googlesource.com/c/go/+/172359
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Ian Lance Taylor <iant@golang.org>
variable setg_gcc in runtime/cgo/*.c should be static, otherwise it
will be mixed with the function of the same name in runtime/asm_*.s or
tls_*.s, which causes an error when building PIE with internal linking
mode.
Fixes #31485
Change-Id: I79b311ffcaf450984328db65397840ae7d85e65d
Reviewed-on: https://go-review.googlesource.com/c/go/+/172498 Reviewed-by: Ian Lance Taylor <iant@golang.org>
cmd/link: increase the reserved space for ELF relocations
Currently the offset values of ELF relocations and Macho relocations
are 256 and 512 respectively, which means that the space reserved for
ELF relocations is only 256. But AARCH64 has more than 256 ELF relocation
types, in fact the maximum AARCH64 ELF relocation type recorded in file
src/debug/elf/elf.go is 1032 currently. So this CL increases the offset
of Macho relocations to 2048 to leave enough space for AARCH64 ELF
relocations.
Change-Id: I784ac38aeb3e102ac7825f6d621086849c8d3146
Reviewed-on: https://go-review.googlesource.com/c/go/+/172497
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Ian Lance Taylor <iant@golang.org>
Matthew Dempsky [Wed, 17 Apr 2019 22:40:46 +0000 (15:40 -0700)]
test: fix escape_runtime_atomic.go
Casp1 is implemented in Go on js/wasm, so escape analysis correctly
determines that the "old" parameter does not escape (which is good).
Unfortunately, test/run.go doesn't have a way to indicate that ERROR
messages are optional, and cmd/compile only emits diagnostics for "var
x int" when it's moved to the heap; not when it stays on the stack.
To accomodate that this test currently passes on some GOARCHes but not
others, rewrite the Casp1 test to use "x := new(int)" and allow both
"new(int) escapes to heap" or "new(int) does not escape".
We use a struct to allocate two structs simultaneously.
Because we embed structs rather than using named fields,
the compiler generates forwarding method stubs for the
anonymous type.
In theory, the compiler could detect that these stubs are unnecessary:
The value in question has a very limited scope, the methods are not
called, and there are operations where an interface would need
to be satisfied.
This compiler optimization is unlikely to happen, though;
the ROI is likely to be low.
Instead, just give the fields names. Cuts 64k off the cmd/compile binary.
Matthew Dempsky [Wed, 17 Apr 2019 18:23:53 +0000 (11:23 -0700)]
runtime/internal/atomic: remove bad go:noescape annotations on Loadp
The //go:noescape directive says that arguments don't leak at all,
which is too aggressive of a claim for functions that return pointers
derived from their parameters.
Remove the directive for now. Long term fix will require a new
directive that allows more fine-grained control over escape analysis
information supplied for functions implemented in assembly.
Also, update the BAD comments in the test cases for Loadp: we really
want that *ptr leaks to the result parameter, not that *ptr leaks to
the heap.
Matthew Dempsky [Wed, 17 Apr 2019 00:30:35 +0000 (17:30 -0700)]
test: add regress test cases for self-assignment
Cherry pointed out this case in review for CL 136496. That CL was
slightly too aggressive, and I likely would have made the same mistake
if I tried it myself.
Matthew Dempsky [Tue, 16 Apr 2019 23:32:26 +0000 (16:32 -0700)]
test: add escape regress tests for runtime and sync atomics
There weren't any tests to make sure these work correctly, and this
led to escape analysis regressions in both linux/s390x and js/wasm.
The underlying issue that cmd/compile is only getting some of these
correct because escape analysis doesn't understand //go:linkname is
still present, but at least this addresses the fragility aspect.
This test fails frequently in the longtest builder, and the failures
on the build dashboard have masked two other regressions so far.
Let's skip it until it can be fixed.
jfbus [Tue, 26 Mar 2019 18:21:53 +0000 (18:21 +0000)]
net: support single-request resolv.conf option in pure Go resolver
There is a DNS resolution issue in Kubernetes (UDP response packets get dropped due to a race in conntrack between the parallel A and AAAA queries, causing timeouts in DNS queries).
A workaround is to enable single-request / single-request-reopen in resolv.conf in order to use sequential A and AAAA queries instead of parallel queries.
With this PR, the pure Go resolver searches for "single-request" and "single-request-reopen" in resolv.conf and send A and AAAA queries sequentially when found.
David Benjamin [Sun, 28 Oct 2018 19:52:51 +0000 (14:52 -0500)]
crypto/tls: fix a minor MAC vs padding leak
The CBC mode ciphers in TLS are a disaster. By ordering authentication
and encryption wrong, they are very subtly dependent on details and
implementation of the padding check, admitting attacks such as POODLE
and Lucky13.
crypto/tls does not promise full countermeasures for Lucky13 and still
contains some timing variations. This change fixes one of the easy ones:
by checking the MAC, then the padding, rather than all at once, there is
a very small timing variation between bad MAC and (good MAC, bad
padding).
The consequences depend on the effective padding value used in the MAC
when the padding is bad. extractPadding simply uses the last byte's
value, leaving the padding bytes effectively unchecked. This is the
scenario in SSL 3.0 that led to POODLE. Specifically, the attacker can
take an input record which uses 16 bytes of padding (a full block) and
replace the final block with some interesting block. The MAC check will
succeed with 1/256 probability due to the final byte being 16. This
again means that after 256 queries, the attacker can decrypt one byte.
To fix this, bitwise AND the two values so they may be checked with one
branch. Additionally, zero the padding if the padding check failed, to
make things more robust.
Jay Conrod [Mon, 15 Apr 2019 19:34:49 +0000 (15:34 -0400)]
cmd/go: print package import chains for some build list errors
When we construct the build list by loading packages (e.g., in
"go build", "go list", or "go test"), we may load additional modules
not mentioned in the original build list. If we encounter an error
loading one of these modules, mvs.BuildList currently returns a
BuildListError with a chain of requirments. Unfortunately, this is not
helpful, since the graph is structured such that these missing modules
are direct requirements of the main module.
With this change, loader.load keeps track of the package that caused
each "missing" module to be added. If an error occurs in a missing
module, the chain of package imports is printed instead of the module
requirements.
Fixes #31475
Change-Id: Ie484814af42ceea3e85fedc38e705ba3a38cd495
Reviewed-on: https://go-review.googlesource.com/c/go/+/171859
Run-TryBot: Jay Conrod <jayconrod@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Bryan C. Mills <bcmills@google.com>
Jay Conrod [Mon, 8 Apr 2019 20:57:13 +0000 (16:57 -0400)]
cmd/go: describe dependencies in build list error messages
mvs.BuildList reports errors with a chain of modules to make it
clear why the module where the error occurred was part of the
build. This is a little confusing with "go get -u" since there are
edges in the module graph for requirements and for updates.
With this change, we now print "requires" or "updates to" between
each module version in the chain.
Updates #30661
Change-Id: Ie689500ea86857e715b250b9e0cae0bc6686dc32
Reviewed-on: https://go-review.googlesource.com/c/go/+/171150
Run-TryBot: Jay Conrod <jayconrod@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Bryan C. Mills <bcmills@google.com>
Jay Conrod [Mon, 8 Apr 2019 18:31:21 +0000 (14:31 -0400)]
cmd/go: handle wildcards for unknown modules in "go get"
For example, "go get golang.org/x/tools/cmd/..." will add a
requirement for "golang.org/x/tools" to go.mod and will install
executables from the "cmd" subdirectory.
Fixes #29363
Change-Id: Id53f051710708d7760ffe831d4274fd54533d2b7
Reviewed-on: https://go-review.googlesource.com/c/go/+/171138
Run-TryBot: Jay Conrod <jayconrod@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Bryan C. Mills <bcmills@google.com>
Constantin Konstantinidis [Sun, 14 Apr 2019 05:52:05 +0000 (07:52 +0200)]
cmd/go: remove auto-deriving module path for github.com
This fix removes the special case of auto-deriving the module path
only for VCS github.com. Error message now explicitly requests
the module path. Documentation and its FAQ do not need an update
as only the beginning of the message is mentioned and is not modified.
Fixes #27951
Change-Id: Icaf87a38b5c58451edba9beaa12ae9a68e288ca1
Reviewed-on: https://go-review.googlesource.com/c/go/+/172019 Reviewed-by: Daniel Lublin <daniel@lublin.se> Reviewed-by: Bryan C. Mills <bcmills@google.com>
Run-TryBot: Jay Conrod <jayconrod@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Matthew Dempsky [Tue, 2 Apr 2019 21:43:27 +0000 (14:43 -0700)]
cmd/compile: enable -newescape by default
RELNOTE=yes
The new escape analysis pass is more precise, which for most Go code
should be an improvement. However, it may also break code that
happened to work before (e.g., code that violated the unsafe.Pointer
safety rules).
The old escape analysis pass can be re-enabled with "go build
-gcflags=all=-newescape=false". N.B., it's NOT recommended to mix the
old and new escape analysis passes such as by omitting "all=". While
the old and new escape analysis passes use similar and mostly
compatible metadata, there are cases (e.g., closure handling) where
they semantically differ and could lead to memory corruption errors in
compiled programs.
Fixes #23109.
Change-Id: I0b1b6a6de5e240cb30c87a165f47bb8795491158
Reviewed-on: https://go-review.googlesource.com/c/go/+/170448
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: David Chase <drchase@google.com>
Gergely Brautigam [Sat, 9 Mar 2019 12:28:50 +0000 (13:28 +0100)]
cmd/go: report non-Go files as package error
This change modifies cmd/go/list to format the error correctly in case
-e flag is set. It also fixes a bug where the package loader was only
ever checking the first pattern if it had the go extension. This caused
and error when a file without .go extension was not the first argument.
Fixes #29899
Change-Id: I029bf4465ad4ad054434b8337c1d2a59369783da
Reviewed-on: https://go-review.googlesource.com/c/go/+/166398
Run-TryBot: Bryan C. Mills <bcmills@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Bryan C. Mills <bcmills@google.com>
On some configurations of Windows, directories containing executable
files may be locked for a while after the executable exits (perhaps
due to antivirus scans?). It's probably worth a little extra latency
on exit to avoid filling up the user's temporary directory with leaked
files.
expvar: improve Map.addKey for large number of keys
The existing implementation has poor performance for inserting large
number of keys because it chooses to append the new key to the sorted
keys array and then sort the new array again (time complexity is
O(nlogn)).
The improvement tries to utilize the sorted keys array by searching the
index and doing an insertion if any (time complexity is O(logn)).
Benchmarked on 4-core machine with `go test -cpu 1,2,4,8 -count 5
-benchmem -bench=.`(the equal results are omitted):
net/http: rename DialerAndTLSConfigSupportsHTTP2 to ForceAttemptHTTP2
Transport.DialerAndTLSConfigSupportsHTTP2 was added just earlier
in CL 130256 but we thought of a better name moments after submitting.
ForceAttemptHTTP2 is shorter, more direct, and doesn't constrain what
we can use it with in the future.
Matthew Dempsky [Tue, 2 Apr 2019 21:44:13 +0000 (14:44 -0700)]
cmd/compile: update escape analysis tests for newescape
The new escape analysis implementation tries to emit debugging
diagnostics that are compatible with the existing implementation, but
there's a handful of cases that are easier to handle by updating the
test expectations instead.
For regress tests that need updating, the original file is copied to
oldescapeXXX.go.go with -newescape=false added to the //errorcheck
line, while the file is updated in place with -newescape=true and new
test requirements.
Notable test changes:
1) escape_because.go looks for a lot of detailed internal debugging
messages that are fairly particular to how esc.go works and that I
haven't attempted to port over to escape.go yet.
2) There are a lot of "leaking param: x to result ~r1 level=-1"
messages for code like
func(p *int) *T { return &T{p} }
that were simply wrong. Here &T must be heap allocated unconditionally
(because it's being returned); and since p is stored into it, p
escapes unconditionally too. esc.go incorrectly reports that p escapes
conditionally only if the returned pointer escaped.
3) esc.go used to print each "leaking param" analysis result as it
discovered them, which could lead to redundant messages (e.g., that a
param leaks at level=0 and level=1). escape.go instead prints
everything at the end, once it knows the shortest path to each sink.
4) esc.go didn't precisely model direct-interface types, resulting in
some values unnecessarily escaping to the heap when stored into
non-escaping interface values.
5) For functions written in assembly, esc.go only printed "does not
escape" messages, whereas escape.go prints "does not escape" or
"leaking param" as appropriate, consistent with the behavior for
functions written in Go.
6) 12 tests included "BAD" annotations identifying cases where esc.go
was unnecessarily heap allocating something. These are all fixed by
escape.go.
Updates #23109.
Change-Id: Iabc9eb14c94c9cadde3b183478d1fd54f013502f
Reviewed-on: https://go-review.googlesource.com/c/go/+/170447
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: David Chase <drchase@google.com>
Kunpei Sakai [Tue, 21 Aug 2018 10:30:30 +0000 (19:30 +0900)]
net/http: introduce DialerAndTLSConfigSupportsHTTP2 in Transport
Even when a custom TLS config or custom dialer is specified,
enables HTTP/2 if DialerAndTLSConfigSupportsHTTP2 is true.
By this change, avoid automatically enabling HTTP/2 if DialContext is set.
This change also ensures that DefaultTransport still automatically
enable HTTP/2 as discussed in #14391.
Elias Naur [Tue, 16 Apr 2019 06:52:42 +0000 (08:52 +0200)]
runtime: avoid getg in preinit on Android
sigaction is called as part of library mode initializers
(_rt0_*_lib). Sigaction in turn calls getg, but on Android the TLS
offset for g has not been initialized and getg might return garbage.
Add a check for initialization before calling getg.
Fixes the golang.org/x/mobile/bind/java tests on amd64 and 386.
Fixes #31476
Change-Id: Id2c41fdc983239eca039b49a54b8853c5669d127
Reviewed-on: https://go-review.googlesource.com/c/go/+/172158 Reviewed-by: Ian Lance Taylor <iant@golang.org>
Richard Musiol [Sun, 7 Apr 2019 22:36:55 +0000 (00:36 +0200)]
runtime, cmd/link: optimize memory allocation on wasm
WebAssembly's memory is contiguous. Allocating memory at a high address
also allocates all memory up to that address. This change reduces
the initial memory allocated on wasm from 1GB to 16MB by using multiple
heap arenas and reducing the size of a heap arena.
Daniel Martí [Mon, 15 Apr 2019 14:10:50 +0000 (23:10 +0900)]
all: clean up code with token.IsExported
A handful of packages were reimplementing IsExported, so use
token.IsExported instead. This caused the deps test to fail for net/rpc.
However, net/rpc deals with Go types, and go/token is light and fairly
low-level in terms of Go tooling packages, so that's okay.
While at it, replace all uses of ast.IsExported with token.IsExported.
This is more consistent, and also means that the import graphs are
leaner. A couple of files no longer need to import go/ast, for example.
We can't get rid of cmd/compile/internal/types.IsExported, as the
compiler can only depend on go/token as of Go 1.4. However, gc used
different implementations in a couple of places, so consolidate the use
of types.IsExported there.
Finally, we can't get rid of the copied IsExported implementation in
encoding/gob, as go/token depends on it as part of a test. That test
can't be an external test either, so there's no easy way to break the
import cycle.
Overall, this removes about forty lines of unnecessary code.
Michael Munday [Thu, 24 Jan 2019 16:16:41 +0000 (16:16 +0000)]
cmd/asm: add s390x 'rotate then ... selected bits' instructions
This CL adds the following instructions, useful for shifting/rotating
and masking operations:
* RNSBG - rotate then and selected bits
* ROSBG - rotate then or selected bits
* RXSBG - rotate then exclusive or selected bits
* RISBG - rotate then insert selected bits
It also adds the 'T' (test), 'Z' (zero), 'H' (high), 'L' (low) and
'N' (no test) variants of these instructions as appropriate.
Michael Munday [Fri, 12 Apr 2019 10:04:11 +0000 (11:04 +0100)]
cmd/internal/obj/s390x: handle RestArgs in s390x assembler
Allow up to 3 RestArgs arguments to be specified. This is needed to
for us to add the 'rotate and ... bits' instructions, which require
5 arguments, cleanly.
Change-Id: Ie0171f48aaf48d8399ef578f95352445741d83a9
Reviewed-on: https://go-review.googlesource.com/c/go/+/171773 Reviewed-by: Ian Lance Taylor <iant@golang.org>
Matthew Dempsky [Tue, 2 Apr 2019 17:40:12 +0000 (10:40 -0700)]
cmd/compile: add new escape analysis implementation
This CL adds a new escape analysis implementation, which can be
enabled through the -newescape compiler flag.
This implementation focuses on simplicity, but in the process ends up
using less memory, speeding up some compile-times, fixing memory
corruption issues, and overall significantly improving escape analysis
results.
Updates #23109.
Change-Id: I6176d9a7ae9d80adb0208d4112b8a1e1f4c9143a
Reviewed-on: https://go-review.googlesource.com/c/go/+/170322
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: David Chase <drchase@google.com>
Michael Munday [Thu, 28 Mar 2019 12:05:43 +0000 (08:05 -0400)]
runtime: fix GDB tests on s390x running Ubuntu 18.04
On Ubuntu 18.04 I am seeing GDB fail to restore the stack pointer
during this test because stack unwinding can't find the PC. This CL
is essentially a partial revert of CL 23940 and fixes the issue on
s390x.
Michael Munday [Thu, 28 Mar 2019 16:51:30 +0000 (12:51 -0400)]
cmd/link, runtime: mark goexit as the top of the call stack
This CL adds a new attribute, TOPFRAME, which can be used to mark
functions that should be treated as being at the top of the call
stack. The function `runtime.goexit` has been marked this way on
architectures that use a link register.
This will stop programs that use DWARF to unwind the call stack
from unwinding past `runtime.goexit` on architectures that use a
link register. For example, it eliminates "corrupt stack?"
warnings when generating a backtrace that hits `runtime.goexit`
in GDB on s390x.
Similar code should be added for non-link-register architectures
(i.e. amd64, 386). They mark the top of the call stack slightly
differently to link register architectures so I haven't added
that code (they need to mark "rip" as undefined).
Daniel Martí [Sat, 23 Mar 2019 16:20:35 +0000 (16:20 +0000)]
go/token: add IsIdentifier, IsKeyword, and IsExported
Telling whether a string is a valid Go identifier can seem like an easy
task, but it's easy to forget about the edge cases. For example, some
implementations out there forget that an empty string or keywords like
"func" aren't valid identifiers.
Add a simple implementation with proper Unicode support, and start using
it in cmd/cover and cmd/doc. Other pieces of the standard library
reimplement part of this logic, but don't use a "func(string) bool"
signature, so we're leaving them untouched for now.
Add some tests too, to ensure that we actually got these edge cases
correctly.
Since telling whether a string is a valid identifier requires knowing
that it's not a valid keyword, add IsKeyword too. The internal map was
already accessible via Lookup, but "Lookup(str) != IDENT" isn't as easy
to understand as IsKeyword(str). And, as per Josh's suggestion, we could
have IsKeyword (and probably Lookup too) use a perfect hash function
instead of a global map.
Finally, for consistency with these new functions, add IsExported. That
makes go/ast.IsExported a bit redundant, so perhaps it can be deprecated
in favor of go/token.IsExported in the future. Clarify that
token.IsExported doesn't imply token.IsIdentifier, to avoid ambiguity.
Fixes #30064.
Change-Id: I0e0e49215fd7e47b603ebc2b5a44086c51ba57f7
Reviewed-on: https://go-review.googlesource.com/c/go/+/169018
Run-TryBot: Daniel Martí <mvdan@mvdan.cc>
TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Robert Griesemer <gri@golang.org> Reviewed-by: Alan Donovan <adonovan@google.com>