When rewriting a go.mod file, we currently sort all of the require
lines in a block. The way the parser works is that it considers
preceding blank lines to be empty comment lines, and preceding empty
comment lines are "owned" by their adjoining line. So when we go to sort
them, the empty lines follow around each sorted entry, which doesn't
make a whole lot of sense, since usually vertical space is inserted to
show sections, and if things get moved around by sorting, those sections
are no longer meaningful. This all results in one especially troublesome
edge case: blank lines between a block opening ("require (") and the
first block line ("golang.org/x/sys ...") are not treated the same way
and are rewritten out of existence.
Here's an example of the behavior this fixes.
Starting input file:
require (
golang.zx2c4.com/wireguard master
golang.org/x/crypto latest
golang.org/x/net latest
golang.org/x/sys latest
golang.org/x/text latest
github.com/lxn/walk latest
github.com/lxn/win latest
)
Now we run this through `GOPROXY=direct go get -d`:
require (
github.com/lxn/walk v0.0.0-20190619151032-86d8802c197a
github.com/lxn/win v0.0.0-20190716185335-d1d36f0e4f48
golang.org/x/crypto v0.0.0-20190820162420-60c769a6c586
golang.org/x/net v0.0.0-20190813141303-74dc4d7220e7
golang.org/x/sys v0.0.0-20190813064441-fde4db37ae7a
golang.org/x/text v0.3.2
golang.zx2c4.com/wireguard v0.0.20190806-0.20190822065259-3cedc22d7b49
)
Notice how the blank lines before lxn/walk and x/crypto were preserved.
Finally, we have this be rewritten yet again with a call to `go build`:
require (
github.com/lxn/walk v0.0.0-20190619151032-86d8802c197a
github.com/lxn/win v0.0.0-20190716185335-d1d36f0e4f48
golang.org/x/crypto v0.0.0-20190820162420-60c769a6c586
golang.org/x/net v0.0.0-20190813141303-74dc4d7220e7
golang.org/x/sys v0.0.0-20190813064441-fde4db37ae7a
golang.org/x/text v0.3.2
golang.zx2c4.com/wireguard v0.0.20190806-0.20190822065259-3cedc22d7b49
)
In this final resting point, the first blank line has been removed.
The discrepancy between those two last stages are especially bothersome,
because it makes for lots of dirty git commits and file contents
bouncing back and forth.
This commit fixes the problem as mentioned above, getting rid of those
preceding blank lines. The output in all cases looks as it should, like
this:
require (
github.com/lxn/walk v0.0.0-20190619151032-86d8802c197a
github.com/lxn/win v0.0.0-20190716185335-d1d36f0e4f48
golang.org/x/crypto v0.0.0-20190820162420-60c769a6c586
golang.org/x/net v0.0.0-20190813141303-74dc4d7220e7
golang.org/x/sys v0.0.0-20190813064441-fde4db37ae7a
golang.org/x/text v0.3.2
golang.zx2c4.com/wireguard v0.0.20190806-0.20190822065259-3cedc22d7b49
)
Fixes#33779
Change-Id: I11c894440bd35f343ee62db3e06a50fa871f2599
Reviewed-on: https://go-review.googlesource.com/c/go/+/199917
Run-TryBot: Jason A. Donenfeld <Jason@zx2c4.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Bryan C. Mills <bcmills@google.com>
This functionality already exists but was undocumented. Related to
comments in CL 198797.
Change-Id: Icce40bd7c362423e6ed9c20673ce3de1311e5fd5
Reviewed-on: https://go-review.googlesource.com/c/go/+/200040
Run-TryBot: Jay Conrod <jayconrod@google.com>
Reviewed-by: Bryan C. Mills <bcmills@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Test case with code that caused a gccgo error while emitting export
data for an inlinable function.
Updates #34577.
Change-Id: I28b598c4c893c77f4a76bb4f2d27e5b42f702992
Reviewed-on: https://go-review.googlesource.com/c/go/+/198057
Run-TryBot: Than McIntosh <thanm@google.com>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
For example, if a test calls os.Exit(0), that could trick a 'go test'
run into not running some of the other tests, and thinking that they all
succeeded. This can easily go unnoticed and cause developers headaches.
Add a simple sanity check as part of 'go test': if the test binary
succeeds and doesn't print anything, we should error, as something
clearly went very wrong.
This is done by inspecting each of the stdout writes from the spawned
process, since we don't want to read the entirety of the output into a
buffer. We need to introduce a "buffered" bool var, as there's now an
io.Writer layer between cmd.Stdout and &buf.
A few TestMain funcs in the standard library needed fixing, as they
returned without printing anything as a means to skip testing the entire
package. For that purpose add testenv.MainMust, which prints a warning
and prints SKIP, similar to when -run matches no tests.
Finally, add tests for both os.Exit(0) and os.Exit(1), both as part of
TestMain and as part of a single test, and test that the various stdout
modes still do the right thing.
Fixes#29062.
Change-Id: Ic6f8ef3387dfc64e4cd3e8f903d7ca5f5f38d397
Reviewed-on: https://go-review.googlesource.com/c/go/+/184457
Run-TryBot: Daniel Martí <mvdan@mvdan.cc>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Bryan C. Mills <bcmills@google.com>
The branch-on-count instructions on s390x decrement the input
register and then compare its value to 0. If not equal the branch
is taken.
These instructions are useful for implementing loops with a set
number of iterations (which might be in a register).
For example, this for loop:
for i := 0; i < n; i++ {
... // i is not used or modified in the loop
}
Could be implemented using this assembly:
MOVD Rn, Ri
loop:
...
BRCTG Ri, loop
Note that i will count down from n in the assembly whereas in the
original for loop it counted up to n which is why we can't use i
in the loop.
These instructions will only be used in hand-written codegen and
assembly for now since SSA blocks cannot currently modify values.
We could look into this in the future though.
Change-Id: Iaab93b8aa2699513b825439b8ea20d8fe2ea1ee6
Reviewed-on: https://go-review.googlesource.com/c/go/+/199977
Run-TryBot: Michael Munday <mike.munday@ibm.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
The element size for VSUMQF and VSUMQG was off by one. Fix this
and add tests for VSUM* instruction encodings.
Change-Id: I6de2dabb383e5bc6f85eef1e0f106ba949c9030b
Reviewed-on: https://go-review.googlesource.com/c/go/+/199978
Run-TryBot: Michael Munday <mike.munday@ibm.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
In CL 197059, I suppressed errors if the target package was already found.
However, that does not cover the case of passing a '/v2' module path to
'go get' when the module does not contain a package at its root.
This CL is a minimal fix for that case, intended to be backportable to 1.13.
(Longer term, I intend to rework the version-validation check to treat
all mismatched paths as ErrNotExist.)
Fixes#34746
Updates #34383
Change-Id: Ia963c2ea00fae424812b8f46a4d6c2c668252147
Reviewed-on: https://go-review.googlesource.com/c/go/+/199839
Run-TryBot: Bryan C. Mills <bcmills@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Jay Conrod <jayconrod@google.com>
You were a useful port and you've served your purpose.
Thanks for all the play.
A subsequent CL will remove amd64p32 (including assembly files and
toolchain bits) and remaining bits. The amd64p32 removal will be
separated into its own CL in case we want to support the Linux x32 ABI
in the future and want our old amd64p32 support as a starting point.
Updates #30439
Change-Id: Ia3a0c7d49804adc87bf52a4dea7e3d3007f2b1cd
Reviewed-on: https://go-review.googlesource.com/c/go/+/199499
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
n.Noescape() was overloaded for two uses: (1) to indicate a function
was annotated with //go:noescape, and (2) to indicate that certain
temporary allocations don't outlive the current statement.
The first use case is redundant with n.Func.Pragma&Noescape!=0, which
is the convention we use for checking other function-level pragmas.
The second use case is better served by renaming "Noescape" to
"Transient".
Passes toolstash-check.
Change-Id: I0f09d2d5767513894b7bf49da9cdabd04aa4a05e
Reviewed-on: https://go-review.googlesource.com/c/go/+/199822
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
The newly introduced "late-stage" cycle detection for types
(https://golang.org/cl/196338/) "skips" named types on the
RHS of a type declaration when reporting a cycle. For instance,
for:
type (
A B
B [10]C
C A
)
the reported cycle is:
illegal cycle in declaration of C
C refers to
C
because the underlying type of C resolves to [10]C (note that
cmd/compile does the same but simply says invalid recursive
type C).
This CL introduces the Named.orig field which always refers
to the RHS type in a type definition (and is never changed).
By using Named.orig rather than Named.underlying for the type
validity check, the cycle as written in the source code is
reported:
illegal cycle in declaration of A
A refers to
B refers to
C refers to
A
Fixes#34771.
Change-Id: I41e260ceb3f9a15da87ffae6a3921bd8280e2ac4
Reviewed-on: https://go-review.googlesource.com/c/go/+/199937
Run-TryBot: Robert Griesemer <gri@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
Avoid confusion between (now gone) objSet and objset types.
Also: rename visited -> seen in initorder.go.
No functional changes.
Change-Id: Ib0aa25e006eee55a79a739194d0d26190354a9f2
Reviewed-on: https://go-review.googlesource.com/c/go/+/198044
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
- remove Checker.cycle in favor of using a "seen" map
- rename Checker.typeCycle -> Checker.cycle
- remove TODO in api.go since the API is frozen
Change-Id: I182a8215978dad54e9c6e79c21c5ec88ec802349
Reviewed-on: https://go-review.googlesource.com/c/go/+/198042
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
For Go 1.13, we rewrote the go/types cycle detection scheme. Unfortunately,
it was a bit too clever and introduced a bug (#34333). Here's an example:
type A struct {
f1 *B
f2 B
}
type B A
When type-checking this code, the first cycle A->*B->B->A (via field f1)
is ok because there's a pointer indirection. Though in the process B is
considered "type-checked" (and painted/marked from "grey" to black").
When type-checking f2, since B is already completely set up, go/types
doesn't complain about the invalid cycle A->B->A (via field f2) anymore.
On the other hand, with the fields f1, f2 swapped:
type A struct {
f2 B
f1 *B
}
go/types reports an error because the cycle A->B->A is type-checked first.
In general, we cannot know the "right" order in which types need to be
type-checked.
This CL fixes the issue as follows:
1) The global object path cycle detection does not take (pointer, function,
reference type) indirections into account anymore for cycle detection.
That mechanism was incorrect to start with and the primary cause for this
issue. As a consequence we don't need Checker.indirectType and indir anymore.
2) After processing type declarations, Checker.validType is called to
verify that a type doesn't expand indefinitively. This corresponds
essentially to cmd/compile's dowidth computation (without size computation).
3) Cycles involving only defined types (e.g.: type (A B; B C; C A))
require separate attention as those must now be detected when resolving
"forward chains" of type declarations. Checker.underlying was changed
to detect these cycles.
All three cycle detection mechanism use an object path ([]Object) to
report cycles. The cycle error reporting mechanism is now factored out
into Checker.cycleError and used by all three mechanisms. It also makes
an attempt to report the cycle starting with the "first" (earliest in the
source) object.
Fixes#34333.
Change-Id: I2c6446445e47344cc2cd034d3c74b1c345b8c1e6
Reviewed-on: https://go-review.googlesource.com/c/go/+/196338
Run-TryBot: Robert Griesemer <gri@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
This CL adds system register error checking test cases. There're two kinds of
error test cases:
1. illegal combination.
MRS should be used in this way: MRS <system register>, <general register>.
MSR should be used in this way: MSR <general register>, <system register>.
Error usage examples:
MRS R8, VTCR_EL2 // ERROR "illegal combination"
MSR VTCR_EL2, R8 // ERROR "illegal combination"
2. illegal read or write access.
Error usage examples:
MSR R7, MIDR_EL1 // ERROR "expected writable system register or pstate"
MRS OSLAR_EL1, R3 // ERROR "expected readable system register"
This CL reads system registers readable and writeable property to check whether
they're used with legal read or write access. This property is named AccessFlags
in sysRegEnc.go, and it is automatically generated by modifing the system register
generator.
Change-Id: Ic83d5f372de38d1ecd0df1ca56b354ee157f16b4
Reviewed-on: https://go-review.googlesource.com/c/go/+/194917
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reimplement syscall wrappers for linux/arm64 in terms of supported
syscalls (or in case of Ustat make it return ENOSYS) and remove the
manually added SYS_* consts for the deprecated syscalls. Adapted from
golang.org/x/sys/unix where this is already done since CL 119655.
Change-Id: I94ab48a4645924df3822497d0575f1a1573d509f
Reviewed-on: https://go-review.googlesource.com/c/go/+/199140
Run-TryBot: Tobias Klauser <tobias.klauser@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
When running wasm in the browser, the "process" global is not defined.
This causes functions like os.Getpid() to panic, which is unusual.
For example on Windows os.Getpid() returns -1 and does not panic.
This change adds a dummy polyfill for "process" which returns -1 or an
error. It also extends the polyfill for "fs".
Fixes#34627
Replaces CL 199357
Change-Id: Ifeb12fe7e152c517848933a9ab5f6f749896dcef
Reviewed-on: https://go-review.googlesource.com/c/go/+/199698
Run-TryBot: Richard Musiol <neelance@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
The lines on nodes within the IF-tree generated for switch
statements looks like control flow so the lines get marked
as statement boundaries. Except for the first/root comparison,
explicitly disable the marks.
Change-Id: I64b966ed8e427cdc6b816ff6b6a2eb754346edc7
Reviewed-on: https://go-review.googlesource.com/c/go/+/198738
Run-TryBot: David Chase <drchase@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Jeremy Faller <jeremy@golang.org>
Previously, when emitting type switches without an explicit "case nil"
clause, we would emit:
if x == nil { goto Lnil }
...
Lnil: goto Ldefault
But we can instead just emit:
if x == nil { goto Ldefault }
Doesn't pass toolstash-check; seems like it causes some harmless
instruction scheduling changes.
Change-Id: Ie233dda26756911e93a08b3db40407ba38694c62
Reviewed-on: https://go-review.googlesource.com/c/go/+/199644
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
Reviewed-by: Keith Randall <khr@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Currently, escape analysis is able to record at most one dereference
when a parameter leaks to the heap; that is, at call sites, it can't
distinguish between any of these three functions:
func x1(p ****int) { sink = *p }
func x2(p ****int) { sink = **p }
func x3(p ****int) { sink = ***p }
Similarly, it's limited to recording parameter leaks to only the first
4 parameters, and only up to 6 dereferences.
All of these limitations are due to the awkward encoding scheme used
at the moment.
This CL replaces the encoding scheme with a simple [8]uint8 array,
which can handle up to the first 7 parameters, and up to 254
dereferences, which ought to be enough for anyone. And if not, it's
much more easily increased.
Shrinks export data size geometric mean for Kubernetes by 0.07%.
Fixes#33981.
Change-Id: I10a94b9accac9a0c91490e0d6d458316f5ca1e13
Reviewed-on: https://go-review.googlesource.com/c/go/+/197680
Reviewed-by: Cherry Zhang <cherryyz@google.com>
This CL better abstracts away the parameter leak info that was
directly encoded into the uint16 value. Followup CL will rewrite the
implementation.
Passes toolstash-check.
Updates #33981.
Change-Id: I27f81d26f5dd2d85f5b0e5250ca529819a1f11c2
Reviewed-on: https://go-review.googlesource.com/c/go/+/197679
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
On wasm there is a special goroutine that handles asynchronous events.
Blocking this goroutine often causes a deadlock. However, the stack
trace of this goroutine was omitted when printing the deadlock error.
This change adds an exception so the goroutine is not considered as
an internal system goroutine and the stack trace gets printed, which
helps with debugging the deadlock.
Updates #32764
Change-Id: Icc8f5ba3ca5a485d557b7bdd76bf2f1ffb92eb3e
Reviewed-on: https://go-review.googlesource.com/c/go/+/199537
Run-TryBot: Richard Musiol <neelance@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
CL 170950 had a regression that makes the compiler produce
an invalid wasm binary if the data section is too large.
Loading such a binary gives the following error:
"LinkError: WebAssembly.instantiate(): data segment is out of bounds"
This change fixes the issue by ensuring that the minimum size of the
linear memory is larger than the end of the data section.
Fixes#34395.
Change-Id: I0c8629de7ffd0d85895ad31bf8c9d45fef197a57
Reviewed-on: https://go-review.googlesource.com/c/go/+/199358
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
We're allowed to remove a write barrier when both the old
value in memory and the new value we're writing are not heap pointers.
Improve both those checks a little bit.
A pointer is known to not be a heap pointer if it is read from
read-only memory. This sometimes happens for loads of pointers
from string constants in read-only memory.
Do a better job of tracking which parts of memory are known to be
zero. Before we just kept track of a range of offsets in the most
recently allocated object. For code that initializes the new object's
fields in a nonstandard order, that tracking is imprecise. Instead,
keep a bit map of the first 64 words of that object, so we can track
precisely what we know to be zeroed.
The new scheme is only precise up to the first 512 bytes of the object.
After that, we'll use write barriers unnecessarily. Hopefully most
initializers of large objects will use typedmemmove, which does only one
write barrier check for the whole initialization.
Fixes#34723
Update #21561
Change-Id: Idf6e1b7d525042fb67961302d4fc6f941393cac8
Reviewed-on: https://go-review.googlesource.com/c/go/+/199558
Run-TryBot: Keith Randall <khr@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
For commuting ops, check whether the second argument is dead before
checking if the first argument is rematerializeable. Reusing the register
holding a dead value is always best.
Fixes#33580
Change-Id: I7372cfc03d514e6774d2d9cc727a3e6bf6ce2657
Reviewed-on: https://go-review.googlesource.com/c/go/+/199559
Run-TryBot: Keith Randall <khr@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: David Chase <drchase@google.com>
The current bytes test suit didn't come with endian based test
which causing #34549 can passed the try-bot.
This test will failed when little endian architecture simply using
load and compare uint.
Update #34549
Change-Id: I0973c2cd505ce21c2bed1deeb7d526f1e872118d
Reviewed-on: https://go-review.googlesource.com/c/go/+/198358
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Add test to check that struct type in reflectlite is mirror of reflect.
Note that the test does not check the field types, only check for number
of fields and field name are the same.
Updates #34486
Change-Id: Id5f9b26d35faec97863dd1fe7e5eab37d4913181
Reviewed-on: https://go-review.googlesource.com/c/go/+/199280
Run-TryBot: Cuong Manh Le <cuong.manhle.vn@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
CL 197938 actually fixes those regression on Darwin as syscalls
are no longer labeled as always blocking and consume a thread.
Fixes#33953Fixes#32326
Change-Id: I82c98516c23cd36f762bc5433d7b71ea8939a0ac
Reviewed-on: https://go-review.googlesource.com/c/go/+/199477
Run-TryBot: Minux Ma <minux@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Tobias Klauser <tobias.klauser@gmail.com>
CL 191198 updated runtime rtype and mapType without adopting the changes
to reflectlite, causing mismatch between them.
This CL updates those changes to reflectlite.
Fixes#34486
Change-Id: I2bb043673d997f97bb0b12c4ad471474803b2160
Reviewed-on: https://go-review.googlesource.com/c/go/+/197559
Run-TryBot: Cuong Manh Le <cuong.manhle.vn@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Bryan C. Mills <bcmills@google.com>
Reviewed-by: Keith Randall <khr@golang.org>
TestScripts/mod_get_svn would stop with the following prompt if the real
user didn't have vcs-test.golang.org in their known_hosts file:
The authenticity of host 'vcs-test.golang.org (35.184.38.56)' can't be established.
ECDSA key fingerprint is SHA256:[...]
Are you sure you want to continue connecting (yes/no/[fingerprint])?
This was bad because it relied on the user's real ssh known_hosts file.
Worse even, if the user didn't expert or notice the prompt, it could
hang a 'go test' run for quite a while.
Work around that by forcing svn to not use ssh at all. Other potentially
better approaches were tried, but none worked on svn 1.12.2 with openssh
8.0p1.
Fixes#33883.
Change-Id: I2f925fa892f2fa53c77d86b0034141162517ee69
Reviewed-on: https://go-review.googlesource.com/c/go/+/199142
Run-TryBot: Daniel Martí <mvdan@mvdan.cc>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Updates #34388
Change-Id: I327a1c1557c47fa6c113c7a1a507a8e7355f9d1a
Reviewed-on: https://go-review.googlesource.com/c/go/+/199277
Run-TryBot: Alex Brainman <alex.brainman@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Reviewed-by: Jason A. Donenfeld <Jason@zx2c4.com>
When translating C types, cache the in-progress type under its parent
names, so that anonymous structs can also be translated for multiple
typedefs, without clashing.
Standalone types are not affected by this change.
Also updated the test for issue 9026 because the C struct name
generation algorithm has changed.
Fixes#31891
Change-Id: I00cc64852a2617ce33da13f74caec886af05b9f2
Reviewed-on: https://go-review.googlesource.com/c/go/+/181857
Run-TryBot: Tobias Klauser <tobias.klauser@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
