TestGdbAutotmpTypes takes more than one minute due to gdb performances.
Therefore, it must be skipped in short mode.
Change-Id: I253ebce62264cc7367c9b0f6ce9c5088a9994641
Reviewed-on: https://go-review.googlesource.com/c/go/+/164339
Reviewed-by: Bryan C. Mills <bcmills@google.com>
Run-TryBot: Bryan C. Mills <bcmills@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Fixes#28558
Change-Id: I0ecd9c47fb017cf4bd44725a83a0016c7bb94633
Reviewed-on: https://go-review.googlesource.com/c/go/+/156478
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
This commit add port of runtime/cgo for aix/ppc64.
AIX assembly is different from Linux assembly, therefore gcc_ppc64.S
must be redone for AIX.
Change-Id: I780ebab4ef9c4ce912f4c4d521d8c135b1eebf6e
Reviewed-on: https://go-review.googlesource.com/c/go/+/164002
Run-TryBot: Bryan C. Mills <bcmills@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
This adds benchmarks to test file uploads using PUT requests.
It's designed to complement changes https://golang.org/cl/163599 and
https://golang.org/cl/163737, allowing an easy comparison of
performance before and after these changes are applied.
Updates #30377.
Co-authored-by: Emmanuel Odeke <emm.odeke@gmail.com>
Change-Id: Ib8e692c61e1f7957d88c7101669d4f7fb8110c65
GitHub-Last-Rev: 242622b4fc
GitHub-Pull-Request: golang/go#30424
Reviewed-on: https://go-review.googlesource.com/c/go/+/163862
Run-TryBot: Emmanuel Odeke <emm.odeke@gmail.com>
Reviewed-by: Emmanuel Odeke <emm.odeke@gmail.com>
And explain that it does this. A minor change probably worth mentioning,
although (#28782) I'd still like to freeze this document against any substantial
changes.
Fix#30568.
Change-Id: I74c56744871cfaf00dc52a9b480ca61d3ed19a6b
Reviewed-on: https://go-review.googlesource.com/c/go/+/165597
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Currently, lineWrapper does not detect if it is printing a line comment or not.
Hence, while wrapping a comment, the new line does not get prefixed with a //.
We add logic to lineWrapper to detect this case and add // accordingly. Block
comments do not need any such handling.
Added tests for both cases.
Fixes#20929
Change-Id: I656037c2d865f31dd853cf9195f43ab7c6e6fc53
Reviewed-on: https://go-review.googlesource.com/c/go/+/163578
Run-TryBot: Robert Griesemer <gri@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Robert Griesemer <gri@golang.org>
Currently, the pacer assumes the goal growth ratio is always exactly
GOGC/100. But sometimes this isn't the case, like when the heap is
very small (limited by heapminimum). So to placate the pacer, we lie
about the value of heap_marked in such situations.
Right now, these two lies make a truth, but GOGC is about to get more
complicated with the introduction of heap limits.
Rather than introduce more lies into the system to handle this,
introduce the concept of an "effective GOGC", which is the GOGC we're
actually using for pacing (we'll need this concept anyway with heap
limits). This commit changes the pacer to use the effective GOGC
rather than the user-set GOGC. This way, we no longer need to lie
about heap_marked because its true value is incorporated into the
effective GOGC.
Change-Id: I5b005258f937ab184ffcb5e76053abd798d542bd
Reviewed-on: https://go-review.googlesource.com/c/go/+/66092
Run-TryBot: Austin Clements <austin@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Michael Knyszek <mknyszek@google.com>
Reviewed-by: Rick Hudson <rlh@golang.org>
Currently, the minimum sweep distance is 1MB * GOGC/100. It's been
this way since it was introduced in CL 13043 with no justification.
Since there seems to be no good reason to scale the minimum sweep
distance by GOGC, make the minimum sweep distance just 1MB.
Change-Id: I5320574a23c0eec641e346282aab08a3bbb057da
Reviewed-on: https://go-review.googlesource.com/c/go/+/66091
Run-TryBot: Austin Clements <austin@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Michael Knyszek <mknyszek@google.com>
Reviewed-by: Rick Hudson <rlh@golang.org>
This slightly rearranges gcSetTriggerRatio to compute the goal before
computing the other controls. This will simplify implementing the heap
limit, which needs to control the absolute goal and flow the rest of
the control parameters from this.
For #16843.
Change-Id: I46b7c1f8b6e4edbee78930fb093b60bd1a03d75e
Reviewed-on: https://go-review.googlesource.com/c/go/+/46750
Run-TryBot: Austin Clements <austin@google.com>
Reviewed-by: Michael Knyszek <mknyszek@google.com>
Reviewed-by: Rick Hudson <rlh@golang.org>
This was used during the implementation of concurrent runtime.GC() but
now there's nothing that triggers GC unconditionally. Remove this
trigger type and simplify (gcTrigger).test() accordingly.
Change-Id: I17a893c2ed1f661b8146d7783d529f71735c9105
Reviewed-on: https://go-review.googlesource.com/c/go/+/66090
Run-TryBot: Austin Clements <austin@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Michael Knyszek <mknyszek@google.com>
Reviewed-by: Rick Hudson <rlh@golang.org>
The following corner cases for readCookiesTests are tested now:
- An extra cookie delimiter ";" at the end of a Cookie header
- An empty Cookie header
Change-Id: Id8028b448e4182613fb261bf0903efc98cbf4997
Reviewed-on: https://go-review.googlesource.com/c/go/+/164702
Reviewed-by: Volker Dobler <dr.volker.dobler@gmail.com>
Reviewed-by: Emmanuel Odeke <emm.odeke@gmail.com>
Run-TryBot: Emmanuel Odeke <emm.odeke@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
With cgo, some syscalls will be called with g == nil or m == nil.
SyscallX functions cannot handle them so they call an equivalent
function in sys_aix_ppc64.s which will directly call this syscall.
Change-Id: I6508ec772b304111330e6833e7db729200af547c
Reviewed-on: https://go-review.googlesource.com/c/go/+/164001
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
I had been finding these over a year or so, but none were big enough
changes to warrant CLs. They're a handful now, so clean them all up in a
single commit.
The smaller bodies get a bit simpler, but most importantly, the larger
bodies get unindented.
Change-Id: I5707a6fee27d4c9ff9efd3d363af575d7a4bf2aa
Reviewed-on: https://go-review.googlesource.com/c/go/+/165340
Run-TryBot: Daniel Martí <mvdan@mvdan.cc>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Base64-encoding 32 bytes results in a 44-byte string.
While in general a 44-byte string might decode to 33 bytes,
if you take a 44-byte string that actually only encodes 32 bytes,
and you try to decode it into 32 bytes, that should succeed.
Instead it fails trying to do a useless dst[33:] slice operation.
Delete that slice operation.
Noticed while preparing CL 156322.
Change-Id: I8024bf28a65e2638675b980732b2ff91c66c62cf
Reviewed-on: https://go-review.googlesource.com/c/go/+/164628
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Fix builder breakage from CL 148958.
This is an inlining test that should be skipped on -N -l.
The inlining also doesn't happen on arm and wasm, so skip the test
there too.
Fixes the noopt builder, the linux-arm builder, and the wasm builder.
Updates #30605
Change-Id: I06b90d595be7185df61db039dd225dc90d6f678f
Reviewed-on: https://go-review.googlesource.com/c/go/+/165339
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Two additions are faster than two multiplications and one addition. The
code seems simpler to me too, as it's more obvious that we advance two
destination bytes for each source byte.
name old time/op new time/op delta
Encode/256-8 374ns ± 0% 331ns ± 0% -11.44% (p=0.008 n=5+5)
Encode/1024-8 1.47µs ± 0% 1.29µs ± 0% -11.89% (p=0.004 n=6+5)
Encode/4096-8 5.85µs ± 1% 5.15µs ± 0% -11.89% (p=0.004 n=6+5)
Encode/16384-8 23.3µs ± 0% 20.6µs ± 0% -11.68% (p=0.004 n=6+5)
Change-Id: Iabc63616c1d9fded55fa668ff41dd49efeaa2ea4
Reviewed-on: https://go-review.googlesource.com/c/go/+/151198
Run-TryBot: Daniel Martí <mvdan@mvdan.cc>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: roger peppe <rogpeppe@gmail.com>
After UserCacheDir and UserHomeDir, the only remaining piece which is
commonly needed and portable is a per-user directory to store persistent
files.
For that purpose, UserCacheDir is wrong, as it's meant only for
temporary files. UserHomeDir is also far from ideal, as that clutters
the user's home directory.
Add UserConfigDir, which is implemented in a similar manner to
UserConfigDir.
Fixes#29960.
Change-Id: I7d7a56615103cf76e2b5e2bab2029a6b09d19f0b
Reviewed-on: https://go-review.googlesource.com/c/go/+/160877
Run-TryBot: Daniel Martí <mvdan@mvdan.cc>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
To retrieve MTU on aix/ppc64, a socket must be created. Previously, this
socket was recreated for each interface and not close at all, causing a
fd leak on software using interface API.
Change-Id: Ib573e234bfce58964935831b68d007bfbd923476
Reviewed-on: https://go-review.googlesource.com/c/go/+/165397
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Adds a magic header value that is translated to the
Fetch API redirect option, following existing practices.
Updates #26769
Change-Id: Iaf1c9f710de63ea941a360b73f1b4bb725331a35
Reviewed-on: https://go-review.googlesource.com/c/go/+/164666
Reviewed-by: Richard Musiol <neelance@gmail.com>
Reviewed-by: Agniva De Sarker <agniva.quicksilver@gmail.com>
Run-TryBot: Agniva De Sarker <agniva.quicksilver@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
When parsing nested object, UnmarshalTypeError does not contain actual
path to nested field in original JSON.
This commit change Field to contain the full path to that field. One
can get the Field name by stripping all the leading path elements.
Fixes#22369
Change-Id: I6969cc08abe8387a351e3fb2944adfaa0dccad2a
Reviewed-on: https://go-review.googlesource.com/c/go/+/145218
Reviewed-by: Daniel Martí <mvdan@mvdan.cc>
Run-TryBot: Daniel Martí <mvdan@mvdan.cc>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Replace addu with a sll instruction with a definite behavior (sll will discard the upper 32 bits of the 64 bits, then do sign extensions, with certain behavior). It won't have any UNPREDICTABLE expectation.
Fixes#30459
Change-Id: Id79085c28c5cc4f86939b4ef08ef4bff46077c45
GitHub-Last-Rev: 03569796a9
GitHub-Pull-Request: golang/go#30461
Reviewed-on: https://go-review.googlesource.com/c/go/+/164758
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Android O seems to require Pipe to call the pipe2 system call.
But kernel version 2.6.23 only supports pipe, not pipe2.
So try pipe2 first, then fall back to pipe.
Fixes#30549
Change-Id: I3c5d86e8e945a5ec8a0ecea7853302952c0476c4
Reviewed-on: https://go-review.googlesource.com/c/go/+/165217
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
On aix/ppc64, ar tool must always have -X64 argument if it aims to
create 64 bits archives.
This commit also adds the -D flag handler when calling ar with
gccgotoolchain, to match gccgo version.
Change-Id: I1f5750f8f64a7073780d283567f0b60fc7fa5b97
Reviewed-on: https://go-review.googlesource.com/c/go/+/164417
Reviewed-by: Ian Lance Taylor <iant@golang.org>
The commit fixes asmcgocall in order to use the AIX C ABI.
Change-Id: I2a44914a65557a841ea1e12991938af26ad7fd1d
Reviewed-on: https://go-review.googlesource.com/c/go/+/164000
Run-TryBot: Bryan C. Mills <bcmills@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
EvalSymlinks was mishandling cases like "/x/../../y" or "../../../x"
where there is an extra ".." that goes past the start of the path.
Fixes#30520
Change-Id: I07525575f83009032fa1a99aa270c8d42007d276
Reviewed-on: https://go-review.googlesource.com/c/go/+/164762
Reviewed-by: Bryan C. Mills <bcmills@google.com>
Fixes StartTimer's doc with the verb 'be'
that was previously missing in 'can also used'.
Change-Id: I4b3e6103fbf62d676056d32fcce4618536b7c05c
Reviewed-on: https://go-review.googlesource.com/c/go/+/165117
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Currently on darwin we use MADV_FREE, which unfortunately doesn't result
in a change in the process's RSS until pages actually get kicked out,
which the OS is free to do lazily (e.g. until it finds itself under
memory pressure).
To remedy this, we instead use MADV_FREE_REUSABLE which has similar
semantics, except that it also sets a reusable bit on each page so the
process's RSS gets reported more accurately. The one caveat is for every
time we call MADV_FREE_REUSABLE on a region we must call MADV_FREE_REUSE
to keep the kernel's accounting updated.
Also, because this change requires adding new constants that only exist
on darwin, it splits mem_bsd.go into mem_bsd.go and mem_darwin.go.
Fixes#29844.
Change-Id: Idb6421698511138a430807bcbbd1516cd57557c8
Reviewed-on: https://go-review.googlesource.com/c/go/+/159117
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Run-TryBot: Michael Knyszek <mknyszek@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
All hook files are automatically set up when any git-codereview command is run.
And since the contribution guidelines point to installing git-codereview,
this file does not serve any purpose any more.
Change-Id: I165f6905ca03fd3d512c59e2654ef79e76de934c
Reviewed-on: https://go-review.googlesource.com/c/go/+/158677
Reviewed-by: Andrew Bonventre <andybons@golang.org>
Path should now appear with the correct slash, depending on which
platform install document is being viewed - keeping in line with the
rest of the document.
Fixes#30160
Change-Id: Ib10e5a4adf366c700bff6f8d246bd5e3111ed61c
Reviewed-on: https://go-review.googlesource.com/c/go/+/162918
Reviewed-by: Andrew Bonventre <andybons@golang.org>
When Stdin, Stdout, and Stderr are nil, there are no goroutines to keep
track of, so we don't need a channel.
And in startProcess, preallocate the right size for sysattr.Files,
saving a bit of space and a couple of slice growth allocs.
name old time/op new time/op delta
ExecHostname-8 419µs ± 0% 417µs ± 1% ~ (p=0.093 n=6+6)
name old alloc/op new alloc/op delta
ExecHostname-8 6.40kB ± 0% 6.28kB ± 0% -1.86% (p=0.002 n=6+6)
name old allocs/op new allocs/op delta
ExecHostname-8 34.0 ± 0% 31.0 ± 0% -8.82% (p=0.002 n=6+6)
Change-Id: Ic1d617f29e9c6431cdcadc7f9bb992750a6d5f48
Reviewed-on: https://go-review.googlesource.com/c/164801
Run-TryBot: Daniel Martí <mvdan@mvdan.cc>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Add documentation that individual Write calls are buffered and
copy documentation from bufio.Writer notifying the user to call
Flush and Error when all writes are complete. Remove reference
to "file" since the implementation is general and allows any
io.Writer.
Fixes#30045
Change-Id: I50165470e548f296494e764707fbabe36c665015
Reviewed-on: https://go-review.googlesource.com/c/160680
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Converting a syscall.Errno to an interface is
a significant source of allocations in os/exec.
Elsewhere in the tree, we have pre-allocated errors
for common errno values. Use the same trick here.
This CL makes yet another copy of this code.
The problem is that there isn't really a great place to share it.
The existing copies are in:
cmd/vendor/golang.org/x/sys/unix
cmd/vendor/golang.org/x/sys/windows
cmd/vendor/golang.org/x/sys/windows/registry
internal/syscall/windows
internal/syscall/windows/registry
syscall
internal/poll can't import from cmd/vendor, and cmd/vendor
can't import from internal/*, so we can ignore cmd/vendor.
We could put the unix version in internal/syscall/unix
and then have a platform-independent wrapper in internal/syscall.
But syscall couldn't use it; internal/syscall/* depends on syscall.
So that only allows code re-use with internal/syscall/windows/*.
We could create a new very low level internal package, internal/errno.
But syscall couldn't use it, because it has to import syscall
to get access to syscall.Errno.
So that only allows code re-use with internal/syscall/windows/*.
It's not clear that that any of these options pulls its weight.
The obvious and "correct" place for this is syscall.
But we can't export syscall's version, because package syscall is frozen.
So just copy the code. There's not much of it.
name old alloc/op new alloc/op delta
ExecHostname-8 6.15kB ± 0% 6.13kB ± 0% -0.38% (p=0.000 n=20+19)
name old allocs/op new allocs/op delta
ExecHostname-8 34.0 ± 0% 31.0 ± 0% -8.82% (p=0.000 n=20+20)
Fixes#30535
Change-Id: Idd31c7cced6e15387acc698ffc011e1b7b479903
Reviewed-on: https://go-review.googlesource.com/c/164971
Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
The names of some instructions have been updated in the WebAssembly
specification to be more consistent, see
994591e51c.
This change to the spec is possible because it is still in a draft
state.
Go's support for WebAssembly is still experimental and thus excempt from
the compatibility promise. Being consistent with the spec should
warrant this breaking change to the assembly instruction names.
Change-Id: Iafb8b18ee7f55dd0e23c6c7824aa1fad43117ef1
Reviewed-on: https://go-review.googlesource.com/c/153797
Run-TryBot: Richard Musiol <neelance@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Most notably, it's missing on Windows machines. For example,
windows-amd64-race started failing consistently:
--- FAIL: BenchmarkExecEcho
bench_test.go:15: could not find echo: exec: "echo": executable file not found in %PATH%
We can also reproduce this from Linux with Wine:
$ GOOS=windows go test -bench=. -benchtime=1x -run=- -exec wine
--- FAIL: BenchmarkExecEcho
bench_test.go:15: could not find echo: exec: "echo": executable file not found in %PATH%
Instead, use the "hostname" program, which is available on Windows too.
Interestingly enough, it's also slightly faster than "echo". Any program
is fine as long as it's close enough to a no-op, though.
name old time/op new time/op delta
ExecEcho-8 422µs ± 0% 395µs ± 0% -6.39% (p=0.004 n=6+5)
name old alloc/op new alloc/op delta
ExecEcho-8 6.39kB ± 0% 6.42kB ± 0% +0.53% (p=0.002 n=6+6)
name old allocs/op new allocs/op delta
ExecEcho-8 36.0 ± 0% 36.0 ± 0% ~ (all equal)
Change-Id: I772864d69979172b5cf807552c84d0e165e73051
Reviewed-on: https://go-review.googlesource.com/c/164704
Run-TryBot: Daniel Martí <mvdan@mvdan.cc>
Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
This change adds support for using wasm with Electron. It refactors
environment detection to a more modular approach instead of explicitly
testing for Node.js.
Fixes#29404
Change-Id: I882a9c56523744e7fd7cb2013d158df91cf91d14
Reviewed-on: https://go-review.googlesource.com/c/164665
Run-TryBot: Richard Musiol <neelance@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
There's a "lib/time" sub-section in the Go 1.12 relase notes that
points to a non-existent golang.org/pkg/lib/time page.
The note is about a change in the tz database in the src/lib/time
directory, but the section's title (and the link) should probably just
refer to the time package.
Change-Id: Ibf9dacd710e72886f14ad0b7415fea1e8d25b83a
Reviewed-on: https://go-review.googlesource.com/c/164977
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
We're always going to add stdin, stdout, and stderr to childFiles, so
its length will be at least three. The final length will be those three
elements plus however many files were given via ExtraFiles.
Allocate for that final length directly, saving two slice growth allocs
in the common case where ExtraFiles is empty.
name old time/op new time/op delta
ExecEcho-8 435µs ± 0% 435µs ± 0% ~ (p=0.394 n=6+6)
name old alloc/op new alloc/op delta
ExecEcho-8 6.39kB ± 0% 6.37kB ± 0% -0.39% (p=0.002 n=6+6)
name old allocs/op new allocs/op delta
ExecEcho-8 36.0 ± 0% 34.0 ± 0% -5.56% (p=0.002 n=6+6)
Change-Id: Ib702c0da1e43f0a55ed937af6d45fca6a170e8f3
Reviewed-on: https://go-review.googlesource.com/c/164898
Run-TryBot: Daniel Martí <mvdan@mvdan.cc>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Most of the encoding time is spent in the first Encode loop, since the
rest of the function only deals with the few remaining bytes. Any
unnecessary work done in that loop body matters tremendously.
One such unnecessary bottleneck was the use of the enc.encode table.
Since enc is a pointer receiver, and the field is first used within the
loop, the encoder must perform a nil check at every iteration.
Add a dummy use of the field before the start of the loop, to move the
nil check there. After that line, the compiler now knows that enc can't
be nil, and thus the hot loop is free of nil checks.
name old time/op new time/op delta
EncodeToString-4 14.7µs ± 0% 13.7µs ± 1% -6.53% (p=0.000 n=10+10)
name old speed new speed delta
EncodeToString-4 559MB/s ± 0% 598MB/s ± 1% +6.99% (p=0.000 n=10+10)
Updates #20206.
Change-Id: Icbb523a7bd9e470a8be0a448d1d78ade97ed4ff6
Reviewed-on: https://go-review.googlesource.com/c/151158
Run-TryBot: Daniel Martí <mvdan@mvdan.cc>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
