GODEBUG=netdns=1 prints a one-time strategy decision. (cgo or go DNS lookups)
GODEBUG=netdns=2 prints the per-lookup strategy as a function of the hostname.
The new "netcgo" build tag forces cgo DNS lookups.
GODEBUG=netdns=go (or existing build tag "netgo") forces Go DNS resolution.
GODEBUG=netdns=cgo (or new build tag "netcgo") forces libc DNS resolution.
Options can be combined with e.g. GODEBUG=netdns=go+1 or GODEBUG=netdns=2+cgo.
Fixes#11322Fixes#11450
Change-Id: I7a67e9f759fd0a02320e7803f9ded1638b19e861
Reviewed-on: https://go-review.googlesource.com/11584
Reviewed-by: Russ Cox <rsc@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
The recent https://golang.org/cl/11810 is reportedly a bit too
aggressive.
Apparently some HTTP requests in the wild do contain both a
Transfer-Encoding along with a bogus Content-Length. Instead of
returning a 400 Bad Request error, we should just ignore the
Content-Length like we did before.
Change-Id: I0001be90d09f8293a34f04691f608342875ff5c4
Reviewed-on: https://go-review.googlesource.com/11962
Reviewed-by: Andrew Gerrand <adg@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Go's continuous build system depends on HTTP trailers for the buildlet
interface.
Andrew rewrote the makerelease tool to work in terms of Go's builder
system (now at x/build/cmd/release), but it previously could only
create GCE-based buildlets, which meant x/build/cmd/release couldn't
build the release for Darwin.
https://golang.org/cl/11901 added support for proxying buildlet
connections via the coordinator, but that exposed the fact that
httputil.ReverseProxy couldn't proxy Trailers. A fork of that code
also wasn't possible because net/http needlessly deleted the "Trailer"
response header in the Transport code. This mistake goes back to
"release-branch.r56" and earlier but was never noticed because nobody
ever uses Trailers, and servers via ResponseWriter never had the
ability to even set trailers before this Go 1.5. Note that setting
trailers requires pre-declaring (in the response header) which
trailers you'll set later (after the response body). Because you could
never set them, before this release you could also never proxy them.
Change-Id: I2410a099921790dcd391675ae8610300efa19108
Reviewed-on: https://go-review.googlesource.com/11940
Reviewed-by: Andrew Gerrand <adg@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
This reverts commit 6f7961da28.
Russ suggests changing the frozon syscall package and obviously it's a
better solution. Perhaps he will also let me know the way how to get the
project owners to agree later.
Fixes#11492.
Change-Id: I98f9f366b72b85db54b4acfc3a604b62fb6d783c
Reviewed-on: https://go-review.googlesource.com/11854
Run-TryBot: Mikio Hara <mikioh.mikioh@gmail.com>
Reviewed-by: Rob Pike <r@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
This was originally done in https://codereview.appspot.com/5690059
(Feb 2012) to deal with bad response headers coming back from webcams,
but it presents a potential security problem with HTTP request
smuggling for request headers containing "Content Length" instead of
"Content-Length".
Part of overall HTTP hardening for request smuggling. See RFC 7230.
Thanks to Régis Leroy for the report.
Change-Id: I92b17fb637c9171c5774ea1437979ae2c17ca88a
Reviewed-on: https://go-review.googlesource.com/11772
Reviewed-by: Russ Cox <rsc@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
RawPath is a hint to the desired encoding of Path.
It is ignored when it is not a valid encoding of Path,
such as when Path has been changed but RawPath has not.
It is not ignored but also not useful when it matches
the url package's natural choice of encoding.
In this latter case, set it to the empty string.
This should help drive home the point that clients
cannot in general depend on it being present and
that they should use the EncodedPath method instead.
This also reduces the impact of the change on tests,
especially tests that use reflect.DeepEqual on parsed URLs.
Change-Id: I437c51a33b85439a31c307caf1436118508ea196
Reviewed-on: https://go-review.googlesource.com/11760
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
This allows for "race free" cancellation, in the sense discussed in
issue #11013: in contrast to Transport.CancelRequest, the cancellation
will not be lost if the user cancels before the request is put into the
transport's internal map.
Fixes#11013.
Change-Id: I0b5e7181231bdd65d900e343f764b4d1d7c422cd
Reviewed-on: https://go-review.googlesource.com/11601
Run-TryBot: David Symonds <dsymonds@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
TestTransportAndServerSharedBodyRace got flaky after
issue #9662 was fixed by https://golang.org/cl/11412, which made
servers hang up on clients when a Handler stopped reading its body
early.
This test was affected by a race between the the two goroutines in the
test both only reading part of the request, which was an unnecessary
detail for what the test was trying to test (concurrent Read/Close
races on an *http.body)
Also remove an unused remnant from an old test from which this one was
derived. And make the test not deadlock when it fails. (which was why
the test was showing up as 2m timeouts on the dashboard)
Fixes#11418
Change-Id: Ic83d18aef7e09a9cd56ac15e22ebed75713026cb
Reviewed-on: https://go-review.googlesource.com/11610
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Andrew Gerrand <adg@golang.org>
The race occurs rarely, but by putting some delays and more reads/writes
of prePendingDial/postPendingDial in the handlePendingDial function I
could reproduce it.
Fixes#11136
Change-Id: I8da9e66c88fbda049eaaaaffa2717264ef327768
Reviewed-on: https://go-review.googlesource.com/11250
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
Linux allows to have a peer IP address on IP interface over ethernet
link encapsulation, though it only installs a static route with the peer
address as an on-link nexthop.
Fixes#11338.
Change-Id: Ie2583737e4c7cec39baabb89dd732463d3f10a61
Reviewed-on: https://go-review.googlesource.com/11352
Reviewed-by: Russ Cox <rsc@golang.org>
Includes a new net/http test too.
Fixes#11202
Change-Id: I61edc594f4de8eb6780b8dfa221269dd482e8f35
Reviewed-on: https://go-review.googlesource.com/11492
Reviewed-by: Dmitry Vyukov <dvyukov@google.com>
Reviewed-by: Russ Cox <rsc@golang.org>
All other exported errors in net/http are commented. This change adds
documentation to ErrNoCookie and ErrNoLocation to explain where they are
returned, and why.
Change-Id: I21fa0d070dd35256681ad0714000f238477d4af1
Reviewed-on: https://go-review.googlesource.com/11044
Reviewed-by: Russ Cox <rsc@golang.org>
The consecutive calls to Head would sometimes get different
connections depending on if the readLoop had finished executing
and placed its connection on the idle list or not. This change
ensures that readLoop completes before we make our second connection.
Fixes#11250
Change-Id: Ibdbc4d3d0aba0162452f6dec5928355a37dda70a
Reviewed-on: https://go-review.googlesource.com/11170
Reviewed-by: Russ Cox <rsc@golang.org>
Some old buggy browsers sent extra CRLF(s) after POST bodies. Skip
over them before reading subsequent requests.
Fixes#10876
Change-Id: I62eacf2b3e985caffa85aee3de39d8cd3548130b
Reviewed-on: https://go-review.googlesource.com/11491
Reviewed-by: Andrew Gerrand <adg@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
If a client sent a POST with a huge request body, calling
req.Body.Close in the handler (which is implicit at the end of a
request) would end up consuming it all.
Put a cap on that, using the same threshold used elsewhere for similar
cases.
Fixes#9662
Change-Id: I26628413aa5f623a96ef7c2609a8d03c746669e5
Reviewed-on: https://go-review.googlesource.com/11412
Reviewed-by: Andrew Gerrand <adg@golang.org>
Instead of ambiguously referring to "the Client's CheckRedirect
function" in Head, describe the default behavior like for Get as users
aren't expected to change DefaultClient.CheckRedirect.
While here, use consistent punctuation for the Get and Head Client
method documentation.
Change-Id: I9e7046c73b0d0bc4de002234924d9e7c59aceb41
Reviewed-on: https://go-review.googlesource.com/11362
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
The ListenAndServerTLS function still requires the certFile and
keyFile, but the Server.ListenAndServerTLS method doesn't need to
require the certFile and keyFile if the Server.TLSConfig.Certificates
are already populated.
Fixes#8599
Change-Id: Id2e3433732f93e2619bfd78891f775d89f1d651e
Reviewed-on: https://go-review.googlesource.com/11413
Reviewed-by: Andrew Gerrand <adg@golang.org>
To date, the behavior has depended on whether we're using cgo and
in turn what the host resolver does. Most host resolvers will "resolve"
IP addresses, but the non-cgo pure Go path has not.
This CL makes resolution of IP addresses always work, even if we're not using cgo
and even if the host resolver does not "resolve" IP addresses.
Fixes#11335.
Change-Id: I19e82be968154d94904bb2f72e9c17893019a909
Reviewed-on: https://go-review.googlesource.com/11420
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
This appears to be some legacy which is no longer used.
Change-Id: I469beb59a90853e8de910158f179b32f1aa14c7d
Reviewed-on: https://go-review.googlesource.com/11304
Reviewed-by: Andrew Gerrand <adg@golang.org>
Run-TryBot: Andrew Gerrand <adg@golang.org>
Historically we have declined to try to provide real support for URLs
that contain %2F in the path, but they seem to be popping up more
often, especially in (arguably ill-considered) REST APIs that shoehorn
entire paths into individual path elements.
The obvious thing to do is to introduce a URL.RawPath field that
records the original encoding of Path and then consult it during
URL.String and URL.RequestURI. The problem with the obvious thing
is that it breaks backward compatibility: if someone parses a URL
into u, modifies u.Path, and calls u.String, they expect the result
to use the modified u.Path and not the original raw encoding.
Split the difference by treating u.RawPath as a hint: the observation
is that there are many valid encodings of u.Path. If u.RawPath is one
of them, use it. Otherwise compute the encoding of u.Path as before.
If a client does not use RawPath, the only change will be that String
selects a different valid encoding sometimes (the original passed
to Parse).
This ensures that, for example, HTTP requests use the exact
encoding passed to http.Get (or http.NewRequest, etc).
Also add new URL.EscapedPath method for access to the actual
escaped path. Clients should use EscapedPath instead of
reading RawPath directly.
All the old workarounds remain valid.
Fixes#5777.
Might help #9859.
Fixes#7356.
Fixes#8767.
Fixes#8292.
Fixes#8450.
Fixes#4860.
Fixes#10887.
Fixes#3659.
Fixes#8248.
Fixes#6658.
Reduces need for #2782.
Change-Id: I77b88f14631883a7d74b72d1cf19b0073d4f5473
Reviewed-on: https://go-review.googlesource.com/11302
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
A header of ": value" results in an empty key. Do not add
it to the headers, because RFC7230 (section 3.2) says that
field-names are tokens, which are one or more characters.
Fixes#11205.
Change-Id: I883be89da1489dc84f98523786b019d1d0169d46
Reviewed-on: https://go-review.googlesource.com/11242
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
This change allows LookupAddr to use getnameinfo through cgo for working
together with various name services other than DNS.
Fixes#7855.
Change-Id: I5b3b4aefe3d1b904541c3350865734d8cbb1c1c4
Reviewed-on: https://go-review.googlesource.com/3420
Reviewed-by: Ian Lance Taylor <iant@golang.org>
The motivation of TestLookupHost was to test codepaths on LookupHost,
LookupIP when we set CGO_ENABLED=1. Now we have serveral tests on those
APIs and their codepaths such as TestLookupGooglePublicDNSAddr,
TestCgoLookupIP, TestGoLookupIP, and the test using the ambiguous source
"localhost" is unnecessary.
Fixes#11182.
Change-Id: I397c823e1648114d91a229b316477bff2948b4f9
Reviewed-on: https://go-review.googlesource.com/11057
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Unfortunately there's no simple, easy way to make Dial{TCP,UDP} fail
consistently across all platforms. Fow now we skip the test on Solaris.
Change-Id: Ib3c55f670ac6a174fe9ea682dac7aab96b1e9dfb
Reviewed-on: https://go-review.googlesource.com/11058
Reviewed-by: Ian Lance Taylor <iant@golang.org>
dialSerial connects to a list of addresses in sequence. If a
timeout is specified, then each address gets an equal fraction of the
remaining time, with a magic constant (2 seconds) to prevent
"dial a million addresses" from allotting zero time to each.
Normally, net.Dial passes the DNS stub resolver's output to dialSerial.
If an error occurs (like destination/port unreachable), it quickly skips
to the next address, but a blackhole in the network will cause the
connection to hang until the timeout elapses. This is how UNIXy clients
traditionally behave, and is usually sufficient for non-broken networks.
The DualStack flag enables dialParallel, which implements Happy Eyeballs
by racing two dialSerial goroutines, giving the preferred family a
head start (300ms by default). This allows clients to avoid long
timeouts when the network blackholes IPv4 xor IPv6.
Fixes#8453Fixes#8455Fixes#8847
Change-Id: Ie415809c9226a1f7342b0217dcdd8f224ae19058
Reviewed-on: https://go-review.googlesource.com/8768
Reviewed-by: Mikio Hara <mikioh.mikioh@gmail.com>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
These were found by grepping the comments from the go code and feeding
the output to aspell.
Change-Id: Id734d6c8d1938ec3c36bd94a4dbbad577e3ad395
Reviewed-on: https://go-review.googlesource.com/10941
Reviewed-by: Aamir Khan <syst3m.w0rm@gmail.com>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
The Error function is a potential XSS vector if a user can control the
error message.
For example, an http.FileServer when given a request for this path
/<script>alert("xss!")</script>
may return a response with a body like this
open <script>alert("xss!")</script>: no such file or directory
Browsers that sniff the content may interpret this as HTML and execute
the script. The nosniff header added by this CL should help, but we
should also try santizing the output entirely.
Change-Id: I447f701531329a2fc8ffee2df2f8fa69d546f893
Reviewed-on: https://go-review.googlesource.com/10640
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
This change fixes incorrect parsing of literal IP addresses in local
database when the addresses contain IPv6 zone identifiers, are in
dotted-decimal notation or in colon-hexadecimal notation with leading
zeros.
https://golang.org/cl/5851 already fixed the code path using getaddrinfo
via cgo. This change fixes the remaining non-cgo code path.
Fixes#8243.
Fixes#8996.
Change-Id: I48443611cbabed0d69667cc73911ba3de396fd44
Reviewed-on: https://go-review.googlesource.com/10306
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Fixes an error where the compiler did not spot that the shadowed err
value was set again after being read. That second assignment was lost
as the value was redeclared in an inner scope.
Spotted by Gordon Klass, https://groups.google.com/forum/#!topic/golang-nuts/MdDLbvOjb4o
Change-Id: I28f2da6f98c52afcbb45e17d2b4f36c586598f98
Reviewed-on: https://go-review.googlesource.com/10600
Reviewed-by: Mikio Hara <mikioh.mikioh@gmail.com>
Add the AddressParser type to allow decoding any charset in
mail addresses.
Fixes#7079
Change-Id: Ic34efb3e3d804a4e17149a6c38cfd73f5f275ab7
Reviewed-on: https://go-review.googlesource.com/10392
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
As mentioned in
http://gitweb.dragonflybsd.org/dragonfly.git/commit/727ccde8cce813911d885b7f6ed749dcea68a886,
DragonFly BSD is dropping support for IPv6 IPv4-mapped address.
Unfortunately, on some released versions we see the kernels pretend to
support the feature but actually not (unless tweaking some kernel states
via sysctl.)
To avoid unpredictable behavior, the net package assumes that all
DragonFly BSD kernels don't support IPv6 IPv4-mapped address.
Fixes#10764.
Change-Id: Ic7af3651e0372ec03774432fbb6b2eb0c455e994
Reviewed-on: https://go-review.googlesource.com/10071
Reviewed-by: Ian Lance Taylor <iant@golang.org>
This change ensures that the test does not try to close dual stack
listeners which have not yet been opened.
Spotted in crash here
http://build.golang.org/log/e5843777df400868ce708b7f00c50dc32c2ec478
Change-Id: I79d513e166effb3e018e2b9dfc23751d92fcbe4b
Reviewed-on: https://go-review.googlesource.com/10371
Reviewed-by: Mikio Hara <mikioh.mikioh@gmail.com>
Run-TryBot: Dave Cheney <dave@cheney.net>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Also mentions golang.org/x/net/ipv4 and golang.org/x/net/ipv6.
Change-Id: I653deac7a5e2b129237655a72d6c91207f1b1685
Reviewed-on: https://go-review.googlesource.com/9779
Reviewed-by: Ian Lance Taylor <iant@golang.org>
On Windows, we need to make sure that the node under test has external
connectivity.
Fixes#10795.
Change-Id: I99f2336180c7b56474fa90a4a6cdd5a6c4dd3805
Reviewed-on: https://go-review.googlesource.com/10006
Reviewed-by: Ian Lance Taylor <iant@golang.org>
The previous implementation spawned an extra goroutine to handle
rechecking resolv.conf for changes.
This change eliminates the extra goroutine, and has rechecking
done as part of a lookup. A side effect of this change is that the
first lookup after a resolv.conf change will now succeed, whereas
previously it would have failed. It also fixes rechecking logic to
ignore resolv.conf parsing errors as it should.
Fixes#8652Fixes#10576Fixes#10649Fixes#10650Fixes#10845
Change-Id: I502b587c445fa8eca5207ca4f2c8ec8c339fec7f
Reviewed-on: https://go-review.googlesource.com/9991
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
Reviewed-by: Mikio Hara <mikioh.mikioh@gmail.com>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
The Transport's writer to the remote server is wrapped in a
bufio.Writer to suppress many small writes while writing headers and
trailers. However, when writing the request body, the buffering may get
in the way if the request body is arriving slowly.
Because the io.Copy from the Request.Body to the writer is already
buffered, the outer bufio.Writer is unnecessary and prevents small
Request.Body.Reads from going to the server right away. (and the
io.Reader contract does say to return when you've got something,
instead of blocking waiting for more). After the body is finished, the
Transport's bufio.Writer is still used for any trailers following.
A previous attempted fix for this made the chunk writer always flush
if the underlying type was a bufio.Writer, but that is not quite
correct. This CL instead makes it opt-in by using a private sentinel
type (wrapping a *bufio.Writer) to the chunk writer that requests
Flushes after each chunk body (the chunk header & chunk body are still
buffered together into one write).
Fixes#6574
Change-Id: Icefcdf17130c9e285c80b69af295bfd3e72c3a70
Reviewed-on: https://go-review.googlesource.com/10021
Reviewed-by: Andrew Gerrand <adg@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
FileConn and FilePacketConn APIs accept user-configured socket
descriptors to make them work together with runtime-integrated network
poller, but there's a limitation. The APIs reject protocol sockets that
are not supported by standard library. It's very hard for the net,
syscall packages to look after all platform, feature-specific sockets.
This change allows various platform, feature-specific socket descriptors
to use runtime-integrated network poller by using SocketConn,
SocketPacketConn APIs that bridge between the net, syscall packages and
platforms.
New exposed APIs:
pkg net, func SocketConn(*os.File, SocketAddr) (Conn, error)
pkg net, func SocketPacketConn(*os.File, SocketAddr) (PacketConn, error)
pkg net, type SocketAddr interface { Addr, Raw }
pkg net, type SocketAddr interface, Addr([]uint8) Addr
pkg net, type SocketAddr interface, Raw(Addr) []uint8
Fixes#10565.
Change-Id: Iec57499b3d84bb5cb0bcf3f664330c535eec11e3
Reviewed-on: https://go-review.googlesource.com/9275
Reviewed-by: Ian Lance Taylor <iant@golang.org>
TestAcceptIgnoreSomeErrors was created to test that network
accept function ignores some errors. But conditions created
by the test also affects network reads. Change the test to
ignore these read errors when acceptable.
Fixes#10785
Change-Id: I3da85cb55bd3e78c1980ad949e53e82391f9b41e
Reviewed-on: https://go-review.googlesource.com/9942
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
When running the client header timeout test, there is a race between
us timing out and waiting on the remaining requests to be serviced. If
the client times out before the server blocks on the channel in the
handler, we will be simultaneously adding to a waitgroup with the
value 0 and waiting on it when we call TestServer.Close().
This is largely a theoretical race. We have to time out before we
enter the handler and the only reason we would time out if we're
blocked on the channel. Nevertheless, make the race detector happy
by turning the close into a channel send. This turns the defer call
into a synchronization point and we can be sure that we've entered
the handler before we close the server.
Fixes#10780
Change-Id: Id73b017d1eb7503e446aa51538712ef49f2f5c9e
Reviewed-on: https://go-review.googlesource.com/9905
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
On darwin/arm, the test sometimes fails with:
Process 557 resuming
--- FAIL: TestWriteTimeoutFluctuation (1.64s)
timeout_test.go:706: Write took over 1s; expected 0.1s
FAIL
Process 557 exited with status = 1 (0x00000001)
go_darwin_arm_exec: timeout running tests
This change increaes timeout on iOS builders from 1s to 3s as a
temporarily fix.
Updates #10775.
Change-Id: Ifdaf99cf5b8582c1a636a0f7d5cc66bb276efd72
Reviewed-on: https://go-review.googlesource.com/9915
Reviewed-by: Minux Ma <minux@golang.org>
No code changes, but the test passes here.
And TryBots are happy.
Fixes#8662 maybe
Change-Id: Id37380f72a951c9ad7cf96c0db153c05167e62ed
Reviewed-on: https://go-review.googlesource.com/9778
Reviewed-by: Minux Ma <minux@golang.org>
This change doesn't work perfectly on IPv6-only kernels including CLAT
enabled kernels, but works enough on IPv4-only kernels.
Fixes#10721.
Updates #10729.
Change-Id: I7db0e572e252aa0a9f9f54c8e557955077b72e44
Reviewed-on: https://go-review.googlesource.com/9777
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Also adds missing temporary file deletion.
Change-Id: Ia644b0898022e05d2f5232af38f51d55e40c6fb5
Reviewed-on: https://go-review.googlesource.com/9772
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Android has (had?) its own local DNS resolver daemon, also my fault:
007e987fee
And you access that via libc, not DNS.
Fixes#10714
Change-Id: Iaff752872ce19bb5c7771ab048fd50e3f72cb73c
Reviewed-on: https://go-review.googlesource.com/9793
Reviewed-by: David Crawshaw <crawshaw@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
Fixes#10221.
Change-Id: Ib23805494d8af1946360bfea767f9727e2504dc5
Reviewed-on: https://go-review.googlesource.com/7941
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Reviewed-by: Minux Ma <minux@golang.org>
Unfortunately Oracle Solaris does not have TCP_KEEPIDLE and
TCP_KEEPINTVL. TCP_KEEPIDLE is equivalent to TCP_KEEPALIVE_THRESHOLD,
but TCP_KEEPINTVL does not have a direct equivalent, so we don't set
TCP_KEEPINTVL any more.
Old Darwin versions also lack TCP_KEEPINTVL, but the code tries to set
it anyway so that it works on newer versions. We can't do that because
Oracle might assign the number illumos uses for TCP_KEEPINTVL to a
constant with a different meaning.
Unfortunately there's nothing we can do if we want to support both
illumos and Oracle Solaris with the same GOOS.
Updates #9614.
Change-Id: Id39eb5147f7afa8e951f886c0bf529d00f0e1bd4
Reviewed-on: https://go-review.googlesource.com/7690
Reviewed-by: Minux Ma <minux@golang.org>
Reviewed-by: Mikio Hara <mikioh.mikioh@gmail.com>
This change simplifies unnecessarily redundant error messages in tests.
There's no need to worry any more because package APIs now return
consistent, self-descriptive error values.
Alos renames ambiguous test functions and makes use of test tables.
Change-Id: I7b61027607c4ae2a3cf605d08d58cf449fa27eb2
Reviewed-on: https://go-review.googlesource.com/9662
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Mikio Hara <mikioh.mikioh@gmail.com>
This change makes TestDualStack{TCP,UDP}Listener work more properly by
attempting to book an available service port before testing.
Also simplifies error messages in tests.
Fixes#5001.
Change-Id: If13b0d0039878c9bd32061a0440664e4fa7abaf7
Reviewed-on: https://go-review.googlesource.com/9661
Reviewed-by: Ian Lance Taylor <iant@golang.org>
The test was measuring something, assuming other goroutines had
already scheduled.
Fixes#10427
Change-Id: I2a4d3906f9d4b5ea44b57d972e303bbe2b0b1cde
Reviewed-on: https://go-review.googlesource.com/9561
Reviewed-by: David Crawshaw <crawshaw@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
Per resolv.conf man page, "If this file does not exist, only the name
server on the local machine will be queried."
This behavior also occurs if file is present but unreadable,
or if no nameservers are listed.
Fixes#10566
Change-Id: Id5716da0eae534d5ebfafea111bbc657f302e307
Reviewed-on: https://go-review.googlesource.com/9380
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
This change deflakes timeout, deadline tests, especially fixes socket
and goroutine leaks. Also adds a few missing tests that use features
introduced after go1 release.
Change-Id: Ibf73a4859f8d4a0ee494ca2fd180cbce72a7a2c7
Reviewed-on: https://go-review.googlesource.com/9464
Reviewed-by: Ian Lance Taylor <iant@golang.org>
This change merges unicast_posix_test.go and multicast_test.go into
listen_test.go before deflaking tests for Listen functions.
No code changes.
Change-Id: Ic4cd6531b95dfb5b6e6e254241692eca61a71e94
Reviewed-on: https://go-review.googlesource.com/9460
Reviewed-by: Ian Lance Taylor <iant@golang.org>
This change adds missing CloseRead test and Close tests on Conn,
Listener and PacketConn with various networks.
Change-Id: Iadf99eaf526a323f853d203edc7c8d0577f67972
Reviewed-on: https://go-review.googlesource.com/9469
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Fixes an issue where Response.Write writes out a Content-Length: -1
header when the corresponding Request is a POST or PUT and the
ContentLength was not previously set.
This was encountered when using httputil.DumpResponse
to write out the response from a server that responded to a PUT
request with no Content-Length header. The dumped output is
thus invalid.
Change-Id: I52c6ae8ef3443f1f9de92aeee9f9581dabb05991
Reviewed-on: https://go-review.googlesource.com/9496
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
Not only by network, transport-layer intermediaries but by
virtualization stuff in a node, it is hard to identify the root cause of
weird faults without information of packet flows after disaster
happened.
This change adds Source field to OpError to be able to represent a
5-tuple of internet transport protocols for helping dealing with
complicated systems.
Also clarifies the usage of Source and Addr fields.
Updates #4856.
Change-Id: I96a523fe391ed14406bfb21604c461d4aac2fa19
Reviewed-on: https://go-review.googlesource.com/9231
Reviewed-by: Ian Lance Taylor <iant@golang.org>
RFC 1035 3.3.14 allows a TXT record to contain one or more <character-string>s.
The current implementation returns a "no such host" error if there is more
than one <character-string> in the TXT record.
Fixes#10482
Change-Id: I0ded258005e6b7ba45f687fecd10afa2b321bb77
Reviewed-on: https://go-review.googlesource.com/8966
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
Also, please be informed that the Write method on both connected and
unconnected-mode sockets may return a positive number of bytes written
with timeout or use of closed network connection error.
Change-Id: I2e2e6192e29cef4e9389eb0422c605c6d12e6a3c
Reviewed-on: https://go-review.googlesource.com/9466
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
In the followup changes, tests that require external facilities such as
DNS servers and RRs will move into external_test.go.
Change-Id: Ib460b0c51961159830357652dbf5430e1ba01514
Reviewed-on: https://go-review.googlesource.com/9461
Reviewed-by: Dave Cheney <dave@cheney.net>
Fixes#10366 (how to set custom headers)
Fixes#9836 (PATCH in PostForm)
Fixes#9276 (generating a server-side Request for testing)
Update #8991 (clarify Response.Write for now; export ReverseProxy's copy later?)
Change-Id: I95a11bf3bb3eeeeb72775b6ebfbc761641addc35
Reviewed-on: https://go-review.googlesource.com/9410
Reviewed-by: David Crawshaw <crawshaw@golang.org>
If the machine's network configuration files (resolv.conf,
nsswitch.conf) don't have any unsupported options, prefer Go's DNS
resolver, which doesn't have the cgo & thread over.
It means users can have more than 500 DNS requests outstanding (our
current limit for cgo lookups) and not have one blocked thread per
outstanding request.
Discussed in thread https://groups.google.com/d/msg/golang-dev/2ZUi792oztM/Q0rg_DkF5HMJ
Change-Id: I3f685d70aff6b47bec30b63e9fba674b20507f95
Reviewed-on: https://go-review.googlesource.com/8945
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
In the brief window between getConn and persistConn.roundTrip,
a cancel could end up going missing.
Fix by making it possible to inspect if a cancel function was cleared
and checking if we were canceled before entering roundTrip.
Fixes#10511
Change-Id: If6513e63fbc2edb703e36d6356ccc95a1dc33144
Reviewed-on: https://go-review.googlesource.com/9181
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Previously all errors were 404 errors, even if the real error had
nothing to do with a file being non-existent.
Fixes#10283
Change-Id: I5b08b471a9064c347510cfcf8557373704eef7c0
Reviewed-on: https://go-review.googlesource.com/9200
Reviewed-by: Daniel Morsing <daniel.morsing@gmail.com>
There used to be a small window where if a server declared it would do
a keep-alive connection but then actually closed the connection before
the roundTrip goroutine scheduled after being sent a response from the
readLoop goroutine, then the readLoop goroutine would loop around and
block forever reading from a channel because the numExpectedResponses
accounting was done too late.
Fixes#10457
Change-Id: Icbae937ffe83c792c295b7f4fb929c6a24a4f759
Reviewed-on: https://go-review.googlesource.com/9169
Reviewed-by: Daniel Morsing <daniel.morsing@gmail.com>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
This change replaces server tests with new ones that require features
introduced after go1 release, such as runtime-integrated network poller,
Dialer, etc.
Change-Id: Icf1f94f08f33caacd499cfccbe74cda8d05eed30
Reviewed-on: https://go-review.googlesource.com/9195
Reviewed-by: Ian Lance Taylor <iant@golang.org>
This change deflakes zero byte read/write tests on datagram sockets, and
enables them by default.
Change-Id: I52f1a76f8ff379d90f40a07bb352fae9343ea41a
Reviewed-on: https://go-review.googlesource.com/9194
Reviewed-by: Ian Lance Taylor <iant@golang.org>
This change excludes internal UDP header size from a result of number of
bytes written on WriteTo.
Change-Id: I847d57f7f195657b6f14efdf1b4cfab13d4490dd
Reviewed-on: https://go-review.googlesource.com/9196
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Reviewed-by: David du Colombier <0intro@gmail.com>
This change fixes inconsistent error values on
Lookup{Addr,CNAME,Host,IP.MX,NS,Port,SRV,TXT}.
Updates #4856.
Change-Id: I059bc8ffb96ee74dff8a8c4e8e6ae3e4a462a7ef
Reviewed-on: https://go-review.googlesource.com/9108
Reviewed-by: Ian Lance Taylor <iant@golang.org>
This change fixes inconsistent error values on Interfaces,
InterfaceAddrs, InterfaceBy{Index,Name}, and Addrs and MulticastAddrs
methods of Interface.
Updates #4856.
Change-Id: I09e65522a22f45c641792d774ebf7a0081b874ad
Reviewed-on: https://go-review.googlesource.com/9140
Reviewed-by: Ian Lance Taylor <iant@golang.org>
This change fixes inconsistent error values on
Set{Deadline,ReadDeadline,WriteDeadline,ReadBuffer,WriteBuffer} for
Conn, Listener and PacketConn, and
Set{KeepAlive,KeepAlivePeriod,Linger,NoDelay} for TCPConn.
Updates #4856.
Change-Id: I34ca5e98f6de72863f85b2527478b20d8d5394dd
Reviewed-on: https://go-review.googlesource.com/9109
Reviewed-by: Ian Lance Taylor <iant@golang.org>
This change fixes inconsistent error values on
File{Conn,Listener,PacketConn} and File method of Conn, Listener.
Updates #4856.
Change-Id: I3197b9277bef0e034427e3a44fa77523acaa2520
Reviewed-on: https://go-review.googlesource.com/9101
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Also moves a few server test helpers into mockserver_test.go.
Change-Id: I5a95c9bc6f0c4683751bcca77e26a8586a377466
Reviewed-on: https://go-review.googlesource.com/9106
Reviewed-by: Ian Lance Taylor <iant@golang.org>
This change fixes inconsistent error values on Accept{,TCP,Unix}.
Updates #4856.
Change-Id: Ie3bb534c19a724cacb3ea3f3656e46c810b2123f
Reviewed-on: https://go-review.googlesource.com/8996
Reviewed-by: Ian Lance Taylor <iant@golang.org>
This change fixes inconsistent error values on Close, CloseRead and
CloseWrite.
Updates #4856.
Change-Id: I3c4d46ccd7d6e1a2f52d8e75b512f62c533a368d
Reviewed-on: https://go-review.googlesource.com/8994
Reviewed-by: Ian Lance Taylor <iant@golang.org>
This change fixes inconsistent error values on Write,
WriteTo{,UDP,IP,Unix} and WriteMsg{UDP,IP,Unix}.
Updates #4856.
Change-Id: I4208ab6a0650455ad7d70a80a2d6169351d6055f
Reviewed-on: https://go-review.googlesource.com/8993
Reviewed-by: Ian Lance Taylor <iant@golang.org>
This change fixes inconsistent error values on Read,
ReadFrom{,UDP,IP,Unix} and ReadMsg{UDP,IP,Unix}.
Updates #4856.
Change-Id: I7de5663094e09be2d78cdb18ce6f1e7ec260888d
Reviewed-on: https://go-review.googlesource.com/8992
Reviewed-by: Ian Lance Taylor <iant@golang.org>
This change adds a type addrinfoErrno to represent getaddrinfo,
getnameinfo-specific errors, and uses it in cgo-based lookup functions.
Also retags cgo files for clarification and does minor cleanup.
Change-Id: I6db7130ad7bf35bbd4e8839a97759e1364c43828
Reviewed-on: https://go-review.googlesource.com/9020
Reviewed-by: Ian Lance Taylor <iant@golang.org>
This change makes TestDialError, TestListenError work without any
external dependency, enables them by default, and removes unnecessary
-run_error_test flag for fixing #4856.
Also fixes inconsistent error values on Dial, Listen partially as a
first stab.
Updates #4856.
Change-Id: Ie10c151ae06759085f352c7db2ca45107a81914f
Reviewed-on: https://go-review.googlesource.com/8903
Reviewed-by: Ian Lance Taylor <iant@golang.org>
In followup changes, we'll move OpError around from the netFD layer to
the Conn layer for fixing #4856. Before doing that, this change makes
netFD of Plan 9 match netFD for POSIX platforms to avoid conflict.
Change-Id: Iea7632716d48722a1758e52effefec964a3a9442
Reviewed-on: https://go-review.googlesource.com/8990
Reviewed-by: David du Colombier <0intro@gmail.com>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Under some dial tests that require external network connectivity, we
must prevent application traffic but must not interfere with control
plane traffic such as DNS message exchange. But test helper function
disableSocketConnect prevents both application and control plane traffic
unconditionally and makes some dial tests with -ipv6 fail when
CGO_ENABLED=0.
This change makes disableSocketConnect take a look at not only address
family but socket type for fixing some dial tests with -ipv6 when
CGO_ENBALED=0.
Change-Id: I32241d9592d31483424bb5e69cb4d56f3fc20312
Reviewed-on: https://go-review.googlesource.com/8743
Reviewed-by: Ian Lance Taylor <iant@golang.org>
The memory model has been clarified since. This is legal and doesn't
need justification.
Change-Id: I60f9938503f86f52bb568ca1a99ac721ee72cee5
Reviewed-on: https://go-review.googlesource.com/8913
Reviewed-by: Mikio Hara <mikioh.mikioh@gmail.com>
Just like darwin/arm.
Change-Id: Ib9a32bb0aed5f08b27de11a93aaf273cacdf5779
Reviewed-on: https://go-review.googlesource.com/8819
Reviewed-by: Minux Ma <minux@golang.org>
Previously, running
$ go get -u -v golang.org/x/tools/cmd/godoc
would results in dozens of HTTP requests for
https://golang.org/x/tools?go-get=1
once per package under x/tools.
Now it caches the results. We still end up doing one HTTP request for
all the packages under x/tools, but this reduces the total number of
HTTP requests in ~half.
This also moves the singleflight package back into an internal
package. singleflight was originally elsewhere as a package, then got
copied into "net" (without its tests). But now that we have internal,
put it in its own package, and restore its test.
Fixes#9249
Change-Id: Ieb5cf04fc4d0a0c188cb957efdc7ea3068c34e3f
Reviewed-on: https://go-review.googlesource.com/8727
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
Remove the "netaddr" type, which ambiguously represented either one
address, or a list of addresses. Instead, use "addrList" wherever
multiple addresses are supported.
The "first" method returns the first address matching some condition
(e.g. "is it IPv4?"), primarily to support legacy code that can't handle
multiple addresses.
The "partition" method splits an addrList into two categories, as
defined by some strategy function. This is useful for implementing
Happy Eyeballs, and similar two-channel algorithms.
Finally, internetAddrList (formerly resolveInternetAddr) no longer
mangles the ordering defined by getaddrinfo. In the future, this may
be used by a sequential Dial implementation.
Updates #8453, #8455.
Change-Id: I7375f4c34481580ab40e31d33002a4073a0474f3
Reviewed-on: https://go-review.googlesource.com/8360
Reviewed-by: Mikio Hara <mikioh.mikioh@gmail.com>
Run-TryBot: Mikio Hara <mikioh.mikioh@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
There was a logical race in Transport.RoundTrip where a roundtrip with
a pending response would race with the channel for the connection
closing. This usually happened for responses with connection: close
and no body.
We handled this race by reading the close channel, setting a timer
for 100ms and if no response was returned before then, we would then
return an error.
This put a lower bound on how fast a connection could fail. We couldn't
fail a request faster than 100ms.
Reordering the channel operations gets rid of the logical race. If
the readLoop causes the connection to be closed, it would have put
its response into the return channel already and we can fetch it with
a non-blocking receive.
Change-Id: Idf09e48d7a0453d7de0120d3055d0ce5893a5428
Reviewed-on: https://go-review.googlesource.com/1787
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
When making a request to an IPv6 address with a zone identifier, for
exmaple [fe80::1%en0], RFC 6874 says HTTP clients must remove the zone
identifier "%en0" before writing the request for security reason.
This change removes any IPv6 zone identifer attached to URI in the Host
header field in requests.
Fixes#9544.
Change-Id: I7406bd0aa961d260d96f1f887c2e45854e921452
Reviewed-on: https://go-review.googlesource.com/3111
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Using IPv6 link-local addresses to make connections between on-link
nodes is useful for small distributed applications but it requires zone
identifiers to distinguish a correct IP link. It's the same for
transports using URI for destination discovery such as HTTP, WebSocket.
This change allows Parse, ParseRequestURI functions and String method of
URL to parse/return a literal IPv6 address followed by a zone identifier
within a URI as described in RFC 6874.
Fixes#6530.
Change-Id: I2936ea65c1446994770cf2ee2c28a1c73faaa0ca
Reviewed-on: https://go-review.googlesource.com/2431
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
This change deflakes TestDialerDualStackFDLeak, TestDialerDualStack,
TestResolve{TCP,UDP,IP}Addr by removing external dependencies.
Fixes#8764.
Change-Id: I5cca0a93776cf05652e0e6a4a4ff4af392ccb885
Reviewed-on: https://go-review.googlesource.com/8485
Reviewed-by: Ian Lance Taylor <iant@golang.org>
This change tries to stop various tester goroutines at the end of each
scope for avoiding interference between test cases including benchmarks.
Not yet finished completely but enough to land upcoming changes to Dial
functions. The rest will be fixed later.
Change-Id: Ic38b8681a3a2ddbcd69ba3696f24a61d418a0346
Reviewed-on: https://go-review.googlesource.com/8398
Reviewed-by: Ian Lance Taylor <iant@golang.org>
This change adds testHookLookIP to enable to inject DNS name to IP
address mappings for Happ{y,yish,ier} Eyeballs dial testing.
Change-Id: I8ac04a594e1e2bd77909528df0552889914a7790
Reviewed-on: https://go-review.googlesource.com/8399
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Now, only a zero deadline is interpreted as noDeadline. Any other time
in the past yields an immediate timeout.
TestConnectDeadlineInThePast already covers this case. We just need to
un-skip it for plan9, where dialChannel is used.
Change-Id: I995fd1a632c31f8004dac772c3d7c43a2a5853b0
Reviewed-on: https://go-review.googlesource.com/8435
Reviewed-by: Mikio Hara <mikioh.mikioh@gmail.com>
This change makes use of the socktest package instead of the non-thread
safe variable syscall.SocketDisableIPv6 for simulating unreachable
external networks.
Also adds -ipv4 flag, -ipv6 flag already exists, as a control knob for
testing on each of IPv4-only, IPv6-only and dual IP stack kernels.
Fixes#7687.
Change-Id: I82002007fd526e8cf4de207f935e721df049a22f
Reviewed-on: https://go-review.googlesource.com/8390
Reviewed-by: Ian Lance Taylor <iant@golang.org>
This change makes TestDialTimeoutFDLeak work on almost all the supported
platforms.
Updates #4384.
Change-Id: I3608f438003003f9b7cfa17c9e5fe7077700fd60
Reviewed-on: https://go-review.googlesource.com/8392
Reviewed-by: Ian Lance Taylor <iant@golang.org>
It's a single, package-wide init function specified for the net package.
Change-Id: Id5894d65e1a92297cc16803cc5e4d4eef0b4b099
Reviewed-on: https://go-review.googlesource.com/8391
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Also updates the comments on test flags.
Change-Id: I8dbd90270e08728ab309ab88a3030e0f8e547175
Reviewed-on: https://go-review.googlesource.com/8394
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
This change makes TestDialTimeout work on almost all the supported
platforms.
Updates #3016.
Updates #3307.
Updates #3867.
Updates #5380.
Updates #5349.
Change-Id: Iacf0ebea23cdd8f6c0333d70c667a5a5f5eb0ed2
Reviewed-on: https://go-review.googlesource.com/8220
Reviewed-by: Ian Lance Taylor <iant@golang.org>
This change adds socket system call hooks to existing test cases for
simulating a bit complicated network conditions to help making timeout
and dual IP stack test cases work more properly in followup changes.
Also test cases print debugging information in non-short mode like the
following:
Leaked goroutines:
net.TestWriteTimeout.func2(0xc20802a5a0, 0xc20801d000, 0x1000, 0x1000, 0xc2081d2ae0)
/go/src/net/timeout_test.go:170 +0x98
created by net.TestWriteTimeout
/go/src/net/timeout_test.go:173 +0x745
net.runDatagramPacketConnServer(0xc2080730e0, 0x2bd270, 0x3, 0x2c1770, 0xb, 0xc2081d2ba0, 0xc2081d2c00)
/go/src/net/server_test.go:398 +0x667
created by net.TestTimeoutUDP
/go/src/net/timeout_test.go:247 +0xc9
(snip)
Leaked sockets:
3: {Cookie:615726511685632 Err:<nil> SocketErr:0}
5: {Cookie:7934075906097152 Err:<nil> SocketErr:0}
Socket statistical information:
{Family:1 Type:805306370 Protocol:0 Opened:17 Accepted:0 Connected:5 Closed:17}
{Family:2 Type:805306369 Protocol:0 Opened:450 Accepted:234 Connected:279 Closed:636}
{Family:1 Type:805306369 Protocol:0 Opened:11 Accepted:5 Connected:5 Closed:16}
{Family:28 Type:805306369 Protocol:0 Opened:95 Accepted:22 Connected:16 Closed:116}
{Family:2 Type:805306370 Protocol:0 Opened:84 Accepted:0 Connected:34 Closed:83}
{Family:28 Type:805306370 Protocol:0 Opened:52 Accepted:0 Connected:4 Closed:52}
Change-Id: I0e84be59a0699bc31245c78e2249423459b8cdda
Reviewed-on: https://go-review.googlesource.com/6390
Reviewed-by: Ian Lance Taylor <iant@golang.org>
This change consolidates test helpers that test platform capabilities.
testNetwork, testAddress and testListenArgs report whether given
ariguments are testable on the current platform configuration to
mitigate to receive weird test results.
Change-Id: Ie1ed568a1f9cc50f3155945ea01562904bc2c389
Reviewed-on: https://go-review.googlesource.com/8076
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Package socktest provides utilities for socket testing.
This package allows test cases in the net package to simulate
complicated network conditions such as that a destination address is
resolvable/discoverable but is not routable/reachable at network layer.
Those conditions are required for testing functionality of timeout,
multiple address families.
Change-Id: Idbe32bcc3319b41b0cecac3d058014a93e13288b
Reviewed-on: https://go-review.googlesource.com/6090
Reviewed-by: Ian Lance Taylor <iant@golang.org>