Until now, the goroutine state has been scattered during the
execution of newstack and oldstack. It's all there, and those routines
know how to get back to a working goroutine, but other pieces of
the system, like stack traces, do not. If something does interrupt
the newstack or oldstack execution, the rest of the system can't
understand the goroutine. For example, if newstack decides there
is an overflow and calls throw, the stack tracer wouldn't dump the
goroutine correctly.
For newstack to save a useful state snapshot, it needs to be able
to rewind the PC in the function that triggered the split back to
the beginning of the function. (The PC is a few instructions in, just
after the call to morestack.) To make that possible, we change the
prologues to insert a jmp back to the beginning of the function
after the call to morestack. That is, the prologue used to be roughly:
TEXT myfunc
check for split
jmpcond nosplit
call morestack
nosplit:
sub $xxx, sp
Now an extra instruction is inserted after the call:
TEXT myfunc
start:
check for split
jmpcond nosplit
call morestack
jmp start
nosplit:
sub $xxx, sp
The jmp is not executed directly. It is decoded and simulated by
runtime.rewindmorestack to discover the beginning of the function,
and then the call to morestack returns directly to the start label
instead of to the jump instruction. So logically the jmp is still
executed, just not by the cpu.
The prologue thus repeats in the case of a function that needs a
stack split, but against the cost of the split itself, the extra few
instructions are noise. The repeated prologue has the nice effect of
making a stack split double-check that the new stack is big enough:
if morestack happens to return on a too-small stack, we'll now notice
before corruption happens.
The ability for newstack to rewind to the beginning of the function
should help preemption too. If newstack decides that it was called
for preemption instead of a stack split, it now has the goroutine state
correctly paused if rescheduling is needed, and when the goroutine
can run again, it can return to the start label on its original stack
and re-execute the split check.
Here is an example of a split stack overflow showing the full
trace, without any special cases in the stack printer.
(This one was triggered by making the split check incorrect.)
runtime: newstack framesize=0x0 argsize=0x18 sp=0x6aebd0 stack=[0x6b0000, 0x6b0fa0]
morebuf={pc:0x69f5b sp:0x6aebd8 lr:0x0}
sched={pc:0x68880 sp:0x6aebd0 lr:0x0 ctxt:0x34e700}
runtime: split stack overflow: 0x6aebd0 < 0x6b0000
fatal error: runtime: split stack overflow
goroutine 1 [stack split]:
runtime.mallocgc(0x290, 0x100000000, 0x1)
/Users/rsc/g/go/src/pkg/runtime/zmalloc_darwin_amd64.c:21 fp=0x6aebd8
runtime.new()
/Users/rsc/g/go/src/pkg/runtime/zmalloc_darwin_amd64.c:682 +0x5b fp=0x6aec08
go/build.(*Context).Import(0x5ae340, 0xc210030c71, 0xa, 0xc2100b4380, 0x1b, ...)
/Users/rsc/g/go/src/pkg/go/build/build.go:424 +0x3a fp=0x6b00a0
main.loadImport(0xc210030c71, 0xa, 0xc2100b4380, 0x1b, 0xc2100b42c0, ...)
/Users/rsc/g/go/src/cmd/go/pkg.go:249 +0x371 fp=0x6b01a8
main.(*Package).load(0xc21017c800, 0xc2100b42c0, 0xc2101828c0, 0x0, 0x0, ...)
/Users/rsc/g/go/src/cmd/go/pkg.go:431 +0x2801 fp=0x6b0c98
main.loadPackage(0x369040, 0x7, 0xc2100b42c0, 0x0)
/Users/rsc/g/go/src/cmd/go/pkg.go:709 +0x857 fp=0x6b0f80
----- stack segment boundary -----
main.(*builder).action(0xc2100902a0, 0x0, 0x0, 0xc2100e6c00, 0xc2100e5750, ...)
/Users/rsc/g/go/src/cmd/go/build.go:539 +0x437 fp=0x6b14a0
main.(*builder).action(0xc2100902a0, 0x0, 0x0, 0xc21015b400, 0x2, ...)
/Users/rsc/g/go/src/cmd/go/build.go:528 +0x1d2 fp=0x6b1658
main.(*builder).test(0xc2100902a0, 0xc210092000, 0x0, 0x0, 0xc21008ff60, ...)
/Users/rsc/g/go/src/cmd/go/test.go:622 +0x1b53 fp=0x6b1f68
----- stack segment boundary -----
main.runTest(0x5a6b20, 0xc21000a020, 0x2, 0x2)
/Users/rsc/g/go/src/cmd/go/test.go:366 +0xd09 fp=0x6a5cf0
main.main()
/Users/rsc/g/go/src/cmd/go/main.go:161 +0x4f9 fp=0x6a5f78
runtime.main()
/Users/rsc/g/go/src/pkg/runtime/proc.c:183 +0x92 fp=0x6a5fa0
runtime.goexit()
/Users/rsc/g/go/src/pkg/runtime/proc.c:1266 fp=0x6a5fa8
And here is a seg fault during oldstack:
SIGSEGV: segmentation violation
PC=0x1b2a6
runtime.oldstack()
/Users/rsc/g/go/src/pkg/runtime/stack.c:159 +0x76
runtime.lessstack()
/Users/rsc/g/go/src/pkg/runtime/asm_amd64.s:270 +0x22
goroutine 1 [stack unsplit]:
fmt.(*pp).printArg(0x2102e64e0, 0xe5c80, 0x2102c9220, 0x73, 0x0, ...)
/Users/rsc/g/go/src/pkg/fmt/print.go:818 +0x3d3 fp=0x221031e6f8
fmt.(*pp).doPrintf(0x2102e64e0, 0x12fb20, 0x2, 0x221031eb98, 0x1, ...)
/Users/rsc/g/go/src/pkg/fmt/print.go:1183 +0x15cb fp=0x221031eaf0
fmt.Sprintf(0x12fb20, 0x2, 0x221031eb98, 0x1, 0x1, ...)
/Users/rsc/g/go/src/pkg/fmt/print.go:234 +0x67 fp=0x221031eb40
flag.(*stringValue).String(0x2102c9210, 0x1, 0x0)
/Users/rsc/g/go/src/pkg/flag/flag.go:180 +0xb3 fp=0x221031ebb0
flag.(*FlagSet).Var(0x2102f6000, 0x293d38, 0x2102c9210, 0x143490, 0xa, ...)
/Users/rsc/g/go/src/pkg/flag/flag.go:633 +0x40 fp=0x221031eca0
flag.(*FlagSet).StringVar(0x2102f6000, 0x2102c9210, 0x143490, 0xa, 0x12fa60, ...)
/Users/rsc/g/go/src/pkg/flag/flag.go:550 +0x91 fp=0x221031ece8
flag.(*FlagSet).String(0x2102f6000, 0x143490, 0xa, 0x12fa60, 0x0, ...)
/Users/rsc/g/go/src/pkg/flag/flag.go:563 +0x87 fp=0x221031ed38
flag.String(0x143490, 0xa, 0x12fa60, 0x0, 0x161950, ...)
/Users/rsc/g/go/src/pkg/flag/flag.go:570 +0x6b fp=0x221031ed80
testing.init()
/Users/rsc/g/go/src/pkg/testing/testing.go:-531 +0xbb fp=0x221031edc0
strings_test.init()
/Users/rsc/g/go/src/pkg/strings/strings_test.go:1115 +0x62 fp=0x221031ef70
main.init()
strings/_test/_testmain.go:90 +0x3d fp=0x221031ef78
runtime.main()
/Users/rsc/g/go/src/pkg/runtime/proc.c:180 +0x8a fp=0x221031efa0
runtime.goexit()
/Users/rsc/g/go/src/pkg/runtime/proc.c:1269 fp=0x221031efa8
goroutine 2 [runnable]:
runtime.MHeap_Scavenger()
/Users/rsc/g/go/src/pkg/runtime/mheap.c:438
runtime.goexit()
/Users/rsc/g/go/src/pkg/runtime/proc.c:1269
created by runtime.main
/Users/rsc/g/go/src/pkg/runtime/proc.c:166
rax 0x23ccc0
rbx 0x23ccc0
rcx 0x0
rdx 0x38
rdi 0x2102c0170
rsi 0x221032cfe0
rbp 0x221032cfa0
rsp 0x7fff5fbff5b0
r8 0x2102c0120
r9 0x221032cfa0
r10 0x221032c000
r11 0x104ce8
r12 0xe5c80
r13 0x1be82baac718
r14 0x13091135f7d69200
r15 0x0
rip 0x1b2a6
rflags 0x10246
cs 0x2b
fs 0x0
gs 0x0
Fixes#5723.
R=r, dvyukov, go.peter.90, dave, iant
CC=golang-dev
https://golang.org/cl/10360048
Setenv("AN_ENV_VAR", "") deletes AN_ENV_VAR instead of setting it
to "" at this moment. Also Getenv("AN_ENV_VAR") returns "not found",
if AN_ENV_VAR is "". Change it, so they behave like unix.
Fixes#5610
R=golang-dev, bradfitz
CC=golang-dev
https://golang.org/cl/10594043
Makes it easy to ask the simple question, what is the hash of this data?
Also fix the commentary and prints in Sum256.
R=golang-dev, rsc
CC=golang-dev
https://golang.org/cl/10630043
With this CL, go test -short -cover std successfully builds and
runs all the standard package tests. The tests that look a file
line numbers (log and runtime/debug) fail, because cover is
not inserting //line directives. Everything else passes.
ok cmd/api 0.038s coverage: 66.6% of statements
? cmd/cgo [no test files]
ok cmd/fix 0.043s coverage: 27.2% of statements
ok cmd/go 0.063s coverage: 2.4% of statements
? cmd/godoc [no test files]
ok cmd/gofmt 0.085s coverage: 61.3% of statements
? cmd/yacc [no test files]
ok archive/tar 0.023s coverage: 74.2% of statements
ok archive/zip 0.075s coverage: 71.8% of statements
ok bufio 0.149s coverage: 88.2% of statements
ok bytes 0.135s coverage: 90.4% of statements
ok compress/bzip2 0.087s coverage: 85.1% of statements
ok compress/flate 0.632s coverage: 79.3% of statements
ok compress/gzip 0.027s coverage: 76.7% of statements
ok compress/lzw 0.141s coverage: 71.2% of statements
ok compress/zlib 1.123s coverage: 77.2% of statements
ok container/heap 0.020s coverage: 85.8% of statements
ok container/list 0.021s coverage: 92.5% of statements
ok container/ring 0.030s coverage: 86.5% of statements
? crypto [no test files]
ok crypto/aes 0.054s coverage: 54.3% of statements
ok crypto/cipher 0.027s coverage: 68.8% of statements
ok crypto/des 0.041s coverage: 83.8% of statements
ok crypto/dsa 0.027s coverage: 33.1% of statements
ok crypto/ecdsa 0.048s coverage: 48.7% of statements
ok crypto/elliptic 0.030s coverage: 91.6% of statements
ok crypto/hmac 0.019s coverage: 83.3% of statements
ok crypto/md5 0.020s coverage: 78.7% of statements
ok crypto/rand 0.057s coverage: 20.8% of statements
ok crypto/rc4 0.092s coverage: 70.8% of statements
ok crypto/rsa 0.261s coverage: 80.8% of statements
ok crypto/sha1 0.019s coverage: 83.9% of statements
ok crypto/sha256 0.021s coverage: 89.0% of statements
ok crypto/sha512 0.023s coverage: 88.7% of statements
ok crypto/subtle 0.027s coverage: 83.9% of statements
ok crypto/tls 0.833s coverage: 79.7% of statements
ok crypto/x509 0.961s coverage: 74.9% of statements
? crypto/x509/pkix [no test files]
ok database/sql 0.033s coverage: 75.0% of statements
ok database/sql/driver 0.020s coverage: 46.2% of statements
ok debug/dwarf 0.023s coverage: 71.5% of statements
ok debug/elf 0.035s coverage: 58.2% of statements
ok debug/gosym 0.022s coverage: 1.8% of statements
ok debug/macho 0.023s coverage: 63.7% of statements
ok debug/pe 0.024s coverage: 50.5% of statements
ok encoding/ascii85 0.021s coverage: 89.7% of statements
ok encoding/asn1 0.022s coverage: 77.9% of statements
ok encoding/base32 0.022s coverage: 91.4% of statements
ok encoding/base64 0.020s coverage: 90.7% of statements
ok encoding/binary 0.022s coverage: 66.2% of statements
ok encoding/csv 0.022s coverage: 88.5% of statements
ok encoding/gob 0.064s coverage: 82.2% of statements
ok encoding/hex 0.019s coverage: 86.3% of statements
ok encoding/json 0.047s coverage: 77.3% of statements
ok encoding/pem 0.026s coverage: 80.5% of statements
ok encoding/xml 0.039s coverage: 85.0% of statements
ok errors 0.022s coverage: 100.0% of statements
ok expvar 0.048s coverage: 72.0% of statements
ok flag 0.019s coverage: 86.9% of statements
ok fmt 0.062s coverage: 91.2% of statements
ok go/ast 0.028s coverage: 46.3% of statements
ok go/build 0.190s coverage: 75.4% of statements
ok go/doc 0.095s coverage: 76.7% of statements
ok go/format 0.036s coverage: 79.8% of statements
ok go/parser 0.075s coverage: 82.0% of statements
ok go/printer 0.733s coverage: 88.6% of statements
ok go/scanner 0.031s coverage: 86.5% of statements
ok go/token 0.062s coverage: 79.7% of statements
? hash [no test files]
ok hash/adler32 0.029s coverage: 49.0% of statements
ok hash/crc32 0.020s coverage: 64.2% of statements
ok hash/crc64 0.021s coverage: 53.5% of statements
ok hash/fnv 0.018s coverage: 75.5% of statements
ok html 0.022s coverage: 4.5% of statements
ok html/template 0.087s coverage: 83.9% of statements
ok image 0.108s coverage: 67.1% of statements
ok image/color 0.026s coverage: 20.1% of statements
ok image/draw 0.049s coverage: 69.6% of statements
ok image/gif 0.019s coverage: 65.2% of statements
ok image/jpeg 0.197s coverage: 78.6% of statements
ok image/png 0.055s coverage: 56.5% of statements
ok index/suffixarray 0.027s coverage: 82.4% of statements
ok io 0.037s coverage: 83.4% of statements
ok io/ioutil 0.022s coverage: 70.1% of statements
FAIL log 0.020s
ok log/syslog 2.063s coverage: 71.1% of statements
ok math 0.023s coverage: 76.5% of statements
ok math/big 0.235s coverage: 79.2% of statements
ok math/cmplx 0.020s coverage: 66.5% of statements
ok math/rand 0.031s coverage: 69.9% of statements
ok mime 0.022s coverage: 83.0% of statements
ok mime/multipart 0.389s coverage: 76.1% of statements
ok net 2.219s coverage: 58.0% of statements
ok net/http 4.744s coverage: 82.9% of statements
ok net/http/cgi 0.593s coverage: 68.5% of statements
ok net/http/cookiejar 0.038s coverage: 90.3% of statements
ok net/http/fcgi 0.047s coverage: 37.6% of statements
ok net/http/httptest 0.068s coverage: 68.9% of statements
ok net/http/httputil 0.058s coverage: 52.8% of statements
? net/http/pprof [no test files]
ok net/mail 0.025s coverage: 80.3% of statements
ok net/rpc 0.063s coverage: 71.5% of statements
ok net/rpc/jsonrpc 0.047s coverage: 81.3% of statements
ok net/smtp 0.032s coverage: 74.1% of statements
ok net/textproto 0.023s coverage: 66.0% of statements
ok net/url 0.020s coverage: 78.2% of statements
ok os 4.729s coverage: 73.3% of statements
ok os/exec 39.620s coverage: 65.1% of statements
ok os/signal 0.541s coverage: 89.9% of statements
ok os/user 0.022s coverage: 62.2% of statements
ok path 0.018s coverage: 90.8% of statements
ok path/filepath 10.834s coverage: 88.4% of statements
ok reflect 0.055s coverage: 83.2% of statements
ok regexp 0.084s coverage: 75.5% of statements
ok regexp/syntax 0.547s coverage: 85.2% of statements
ok runtime 4.755s coverage: 75.9% of statements
? runtime/cgo [no test files]
FAIL runtime/debug 0.018s
ok runtime/pprof 0.368s coverage: 8.5% of statements
? runtime/race [no test files]
ok sort 0.059s coverage: 97.7% of statements
ok strconv 0.315s coverage: 95.6% of statements
ok strings 0.147s coverage: 96.1% of statements
ok sync 0.083s coverage: 56.7% of statements
ok sync/atomic 0.035s coverage: 0.0% of statements
ok syscall 0.043s coverage: 24.0% of statements
ok testing 0.018s coverage: 24.0% of statements
? testing/iotest [no test files]
ok testing/quick 0.062s coverage: 83.2% of statements
ok text/scanner 0.020s coverage: 91.5% of statements
ok text/tabwriter 0.021s coverage: 90.4% of statements
ok text/template 0.052s coverage: 81.1% of statements
ok text/template/parse 0.024s coverage: 86.1% of statements
ok time 2.431s coverage: 88.8% of statements
ok unicode 0.024s coverage: 92.1% of statements
ok unicode/utf16 0.017s coverage: 97.3% of statements
ok unicode/utf8 0.019s coverage: 97.4% of statements
? unsafe [no test files]
R=golang-dev, r
CC=golang-dev
https://golang.org/cl/10586043
Makes it easy to ask the simple question, what is the hash of this data?
Also mark block as non-escaping.
R=golang-dev, agl
CC=golang-dev
https://golang.org/cl/10624044
Before, some packages disappear silently if the package cannot be imported,
such as if the import statement is unparseable.
Before:
% ls src
foo issue
% go list ./...
_/home/r/bug/src/foo
%
After:
% go list ./...
src/issue/issue.go:3:5: expected 'STRING', found newline
_/home/r/bug/src/foo
%
R=rsc
CC=golang-dev
https://golang.org/cl/10568043
race is more important than arch (moreover race implies x64)
don't know how to test it
R=golang-dev, dave, r
CC=golang-dev
https://golang.org/cl/10484046
Functions without bodies were excluded from the ordering logic,
because when I wrote the ordering logic there was no reason to
analyze them.
But then we added //go:noescape tags that need analysis, and we
didn't update the ordering logic.
So in the absence of good ordering, //go:noescape only worked
if it appeared before the use in the source code.
Fixes#5773.
R=golang-dev, r
CC=golang-dev
https://golang.org/cl/10570043
USEFIELD is a special kind of NOP, so treat it like a NOP
when generating the pc-ln table.
There are more invasive fixes that could be applied here.
I am going for minimum number of lines changed.
The smallest test case we know of is five distinct Go files
in four packages, and the bug only happens with
GOEXPERIMENT=fieldtrack enabled, which we don't
normally build with, so the test would never run
meaningfully anyway.
Fixes#5762.
R=golang-dev, bradfitz
CC=golang-dev
https://golang.org/cl/10495044
If you hg update your client to an earlier CL, then
hg sync will move you back to tip if it pulls anything in,
but it will leave you where you are if it doesn't pull anything in.
That's confusing: make hg sync always update to tip.
R=golang-dev, bradfitz, r, dsymonds
CC=golang-dev
https://golang.org/cl/10456044
Built after adding -Wconversion to the list of compiler
arguments used when building. I believe these are all OK
assuming we will not change the API. There is no effort to
detect overflow due to very long strings.
R=golang-dev, dave, rsc, r
CC=golang-dev
https://golang.org/cl/10195044
Currently more than 1 gorutine can execute raceWrite() in Wait()
in the following scenario:
1. goroutine 1 executes first check of wg.counter, sees that it's == 0
2. goroutine 2 executes first check of wg.counter, sees that it's == 0
3. goroutine 2 locks the mutex, sees that he is the first waiter and executes raceWrite()
4. goroutine 2 block on the semaphore
5. goroutine 3 executes Done() and unblocks goroutine 2
6. goroutine 1 lock the mutex, sees that he is the first waiter and executes raceWrite()
It produces the following false report:
WARNING: DATA RACE
Write by goroutine 35:
sync.raceWrite()
src/pkg/sync/race.go:41 +0x33
sync.(*WaitGroup).Wait()
src/pkg/sync/waitgroup.go:103 +0xae
command-line-arguments_test.TestNoRaceWaitGroupMultipleWait2()
src/pkg/runtime/race/testdata/waitgroup_test.go:156 +0x19a
testing.tRunner()
src/pkg/testing/testing.go:361 +0x108
Previous write by goroutine 36:
sync.raceWrite()
src/pkg/sync/race.go:41 +0x33
sync.(*WaitGroup).Wait()
src/pkg/sync/waitgroup.go:103 +0xae
command-line-arguments_test.func·012()
src/pkg/runtime/race/testdata/waitgroup_test.go:148 +0x4d
R=golang-dev, r
CC=golang-dev
https://golang.org/cl/10424043
Resubmit 3c2cddfbdaec now that windows callbacks
are not generated during runtime.
Fixes#5494
R=golang-dev, minux.ma, rsc
CC=golang-dev
https://golang.org/cl/10487043
- change runtime_pollWait so it does not return
closed or timeout if IO is ready - windows must
know if IO has completed or not even after
interruption;
- add (*pollDesc).Prepare(mode int) that can be
used for both read and write, same for Wait;
- introduce runtime_pollWaitCanceled and expose
it in net as (*pollDesc).WaitCanceled(mode int);
Full windows netpoll changes are
here https://golang.org/cl/8670044/.
R=golang-dev, dvyukov
CC=golang-dev
https://golang.org/cl/10485043
The compiler still gets the escape analysis wrong, but the annotation here is correct.
R=golang-dev, dave, bradfitz
CC=golang-dev
https://golang.org/cl/10514046
No semantic change.
I found the wording distracting in a couple of instances and was moved to improve it.
R=golang-dev, bradfitz
CC=golang-dev
https://golang.org/cl/10478048
If a server response contains a Content-Length and the body is short,
the Transport should end in io.ErrUnexpectedEOF, not io.EOF.
Fixes#5738
R=golang-dev, kevlar, r
CC=golang-dev
https://golang.org/cl/10237050
In general the description should describe what is added or fixed,
not how it was done (the code does this), but in this case the cause
was delete was missing, so the fix is to add it.
Fixes issue 5765.
R=golang-dev, iant, r
CC=golang-dev
https://golang.org/cl/10496043
until we decide what to do with issues 5659/5736.
Profiling with race detector is not very useful in general,
and now it makes race builders red.
R=golang-dev, rsc
CC=golang-dev
https://golang.org/cl/10523043
Time is a tiny struct, so the compiler copies a Time by
copying each of the three fields.
The layout of a time on amd64 is [ptr int32 gap32 ptr].
Copying a Time onto a location that formerly held a pointer in the
second word changes only the low 32 bits, creating a different
but still plausible pointer. This confuses the garbage collector
when it appears in argument or result frames.
To avoid this problem, declare nsec as uintptr, so that there is
no gap on amd64 anymore, and therefore no partial pointers.
Note that rearranging the fields to put the int32 last still leaves
a gap - [ptr ptr int32 gap32] - because Time must have a total
size that is ptr-width aligned.
Update #5749
This CL is enough to fix the problem, but we should still do
the other actions listed in the initial report. We're not too far
from completely precise collection.
R=golang-dev, dvyukov, r
CC=golang-dev
https://golang.org/cl/10504043
In general the description should describe what is added or fixed,
not how it was done (the code does this), but in this
case the cause were "/ /= ;" was missing,
so the fix is to add it.
Fixes#5761.
R=golang-dev, adg
CC=golang-dev
https://golang.org/cl/10479043
In general the description should describe what is added or fixed, not how it was done (the code does this), but in this case the cause was delete was missing, so the fix is to add it.
Fixes#5759.
R=adg
CC=dave, dominik.honnef, gobot, golang-dev
https://golang.org/cl/10476043
The tradition is to use _posix when the platform extends beyond unix variants. As windows has its own file, rename to the more usual _unix.
R=golang-dev, r
CC=golang-dev
https://golang.org/cl/10320043
If time.Sub results in a value that won't fit in a Duration (int64),
return either the min or max int64 value as appropriate.
Fixes#5011.
R=golang-dev, bradfitz, r, rsc
CC=golang-dev
https://golang.org/cl/10328043
The existing compilers convert empty strings to empty
but non-nil byte and rune slices. The spec required
a nil byte and rune slice in those cases. That seems
an odd additional requirement. Adjust the spec to
match the reality.
Also, removed over-specification for conversions of
nil []byte and []rune: such nil slices already act
like empty slices and thus don't need extra language.
Added extra examples instead.
Fixes#5704.
R=rsc, r, iant
CC=golang-dev
https://golang.org/cl/10440045
trivial: it is not a serious problem to leak a fd in a short lived process, but it was obscuring my investigation of issue 5593.
R=golang-dev, iant, bradfitz
CC=golang-dev
https://golang.org/cl/10391043
