Elias Naur [Sat, 7 May 2016 05:24:39 +0000 (07:24 +0200)]
cmd/go: add -shared to darwin/arm{,64} default build mode
Buildmode c-archive now supports position independent code for
darwin/arm (in addition to darwin/arm64). Make PIC (-shared) the
default for both platforms in the default buildmode.
Without this change, gomobile will go install the standard library
into its separate package directory without PIC support.
Also add -shared to darwin/arm64 in buildmode c-archive, for
symmetry (darwin/arm64 always generates position independent code).
Fixes #15519
Change-Id: If27d2cbea8f40982e14df25da2703cbba572b5c6
Reviewed-on: https://go-review.googlesource.com/22920 Reviewed-by: David Crawshaw <crawshaw@golang.org>
Run-TryBot: David Crawshaw <crawshaw@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Tal Shprecher [Thu, 5 May 2016 22:14:08 +0000 (15:14 -0700)]
cmd/compile: properly handle map assignments for OAS2DOTTYPE
The boolean destination in an OAS2DOTTYPE expression craps out during
compilation when trying to assign to a map entry because, unlike slice entries,
map entries are not directly addressable in memory. The solution is to
properly order the boolean destination node so that map entries are set
via autotmp variables.
Fixes #14678
Change-Id: If344e8f232b5bdac1b53c0f0d21eeb43ab17d3de
Reviewed-on: https://go-review.googlesource.com/22833 Reviewed-by: Matthew Dempsky <mdempsky@google.com>
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Elias Naur [Wed, 9 Mar 2016 09:00:12 +0000 (10:00 +0100)]
runtime: use entire address space on 32 bit
In issue #13992, Russ mentioned that the heap bitmap footprint was
halved but that the bitmap size calculation hadn't been updated. This
presents the opportunity to either halve the bitmap size or double
the addressable virtual space. This CL doubles the addressable virtual
space. On 32 bit this can be tweaked further to allow the bitmap to
cover the entire 4GB virtual address space, removing a failure mode
if the kernel hands out memory with a too low address.
First, fix the calculation and double _MaxArena32 to cover 4GB virtual
memory space with the same bitmap size (256 MB).
Then, allow the fallback mode for the initial memory reservation
on 32 bit (or 64 bit with too little available virtual memory) to not
include space for the arena. mheap.sysAlloc will automatically reserve
additional space when the existing arena is full.
Finally, set arena_start to 0 in 32 bit mode, so that any address is
acceptable for subsequent (additional) reservations.
Before, the bitmap was always located just before arena_start, so
fix the two places relying on that assumption: Point the otherwise unused
mheap.bitmap to one byte after the end of the bitmap, and use it for
bitmap addressing instead of arena_start.
With arena_start set to 0 on 32 bit, the cgoInRange check is no longer a
sufficient check for Go pointers. Introduce and call inHeapOrStack to
check whether a pointer is to the Go heap or stack.
While we're here, remove sysReserveHigh which seems to be unused.
Brad Fitzpatrick [Fri, 6 May 2016 18:46:35 +0000 (18:46 +0000)]
net: skip more flaky net tests on flaky net builders
e.g. https://storage.googleapis.com/go-build-log/9b937dd8/linux-arm_df54a25a.log
Change-Id: Ic5864c7bd840b4f0c6341f919fcbcd5c708b14e7
Reviewed-on: https://go-review.googlesource.com/22881 Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
Ian Lance Taylor [Fri, 6 May 2016 18:32:18 +0000 (11:32 -0700)]
misc/cgo/testcarchive: avoid possible pthread_create race
The old code assumed that the thread ID set by pthread_create would be
available in the newly created thread. While that is clearly true
eventually, it is not necessarily true immediately. Rather than try to
pass down the thread ID, just call pthread_self in the created thread.
Fixes #15576 (I hope).
Change-Id: Ic07086b00e4fd5676c04719a299c583320da64a1
Reviewed-on: https://go-review.googlesource.com/22880
Run-TryBot: Ian Lance Taylor <iant@golang.org> Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Brad Fitzpatrick [Fri, 6 May 2016 18:21:22 +0000 (18:21 +0000)]
net/http: wait longer for subprocess to startup in test
Might deflake the occasional linux-amd64-race failures.
Change-Id: I273b0e32bb92236168eb99887b166e079799c1f1
Reviewed-on: https://go-review.googlesource.com/22858 Reviewed-by: Ian Lance Taylor <iant@golang.org>
Brad Fitzpatrick [Fri, 6 May 2016 16:07:11 +0000 (16:07 +0000)]
os: rename remaining four os1_*.go files to os_*.go
Change-Id: Ice9c234960adc7857c8370b777a0b18e29d59281
Reviewed-on: https://go-review.googlesource.com/22853 Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Brad Fitzpatrick [Fri, 6 May 2016 16:24:57 +0000 (16:24 +0000)]
net: ignore network failures on some builders
We run the external network tests on builders, but some of our
builders have less-than-ideal DNS connectivity. This change continues
to run the tests on all builders, but marks certain builders as flaky
(network-wise), and only validates their DNS results if they got DNS
results.
Change-Id: I826dc2a6f6da55add89ae9c6db892b3b2f7b526b
Reviewed-on: https://go-review.googlesource.com/22852 Reviewed-by: Ian Lance Taylor <iant@golang.org>
Brad Fitzpatrick [Fri, 6 May 2016 16:06:02 +0000 (16:06 +0000)]
runtime: delete empty files
I meant to delete these in CL 22850, actually.
Change-Id: I0c286efd2b9f1caf0221aa88e3bcc03649c89517
Reviewed-on: https://go-review.googlesource.com/22851 Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Embed noLock struct into the following types, so `go vet -copylocks` catches
their copying additionally to types containing sync.Mutex:
- sync.Cond
- sync.WaitGroup
- sync.Pool
- atomic.Value
Richard Miller [Fri, 6 May 2016 13:21:52 +0000 (14:21 +0100)]
syscall,os,net: don't use ForkLock in plan9
This is the follow-on to CL 22610: now that it's the child instead of
the parent which lists unwanted fds to close in syscall.StartProcess,
plan9 no longer needs the ForkLock to protect the list from changing.
The readdupdevice function is also now unused and can be removed.
Change-Id: I904c8bbf5dbaa7022b0f1a1de0862cd3064ca8c7
Reviewed-on: https://go-review.googlesource.com/22842 Reviewed-by: David du Colombier <0intro@gmail.com>
Run-TryBot: David du Colombier <0intro@gmail.com> Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Robert Griesemer [Fri, 6 May 2016 00:46:58 +0000 (17:46 -0700)]
test: update test for issue 15548
Accidentally checked in the version of file c.go that doesn't
exhibit the bug - hence the test was not testing the bug fix.
Double-checked that this version exposes the bug w/o the fix.
Change-Id: Ie4dc455229d1ac802a80164b5d549c2ad4d971f5
Reviewed-on: https://go-review.googlesource.com/22837
Run-TryBot: Robert Griesemer <gri@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Minux Ma <minux@golang.org>
Brad Fitzpatrick [Fri, 6 May 2016 00:52:37 +0000 (17:52 -0700)]
cmd/go: fail with nice error message on bad GOOS/GOARCH pair
Fixes #12272
Change-Id: I2115ec62ed4061084c482eb385a583a1c1909888
Reviewed-on: https://go-review.googlesource.com/22838 Reviewed-by: Ian Lance Taylor <iant@golang.org> Reviewed-by: Minux Ma <minux@golang.org>
Joe Tsai [Thu, 17 Sep 2015 07:22:56 +0000 (00:22 -0700)]
archive/tar: centralize all information about tar header format
The Reader and Writer have hard-coded constants regarding the
offsets and lengths of certain fields in the tar format sprinkled
all over. This makes it harder to verify that the offsets are
correct since a reviewer would need to search for them throughout
the code. Instead, all information about the layout of header
fields should be centralized in one single file. This has the
advantage of being both centralized, and also acting as a form
of documentation about the header struct format.
This method was chosen over using "encoding/binary" since that
method would cause an allocation of a header struct every time
binary.Read was called. This method causes zero allocations and
its logic is no longer than if structs were declared.
Russ Cox [Wed, 6 Jan 2016 17:45:23 +0000 (12:45 -0500)]
runtime: stop traceback at foreign function
This can only happen when profiling and there is foreign code
at the top of the g0 stack but we're not in cgo.
That in turn only happens with the race detector.
Fixes #13568.
Change-Id: I23775132c9c1a3a3aaae191b318539f368adf25e
Reviewed-on: https://go-review.googlesource.com/18322 Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Joe Tsai [Tue, 5 Apr 2016 18:22:53 +0000 (11:22 -0700)]
all: use SeekStart, SeekCurrent, SeekEnd
CL/19862 (f79b50b8d5bc159561c1dcf7c17e2a0db96a9a11) recently introduced the constants
SeekStart, SeekCurrent, and SeekEnd to the io package. We should use these constants
consistently throughout the code base.
Richard Miller [Fri, 29 Apr 2016 16:39:33 +0000 (17:39 +0100)]
syscall: simplify closing of extra fds in plan9 StartProcess
Reviving earlier work by @ality in https://golang.org/cl/57890043
to make the closing of extra file descriptors in syscall.StartProcess
less race-prone. Instead of making a list of open fds in the parent
before forking, the child can read through the list of open fds and
close the ones not explicitly requested. Also eliminate the
complication of keeping open any extra fds which were inherited by
the parent when it started.
This CL will be followed by one to eliminate the ForkLock in plan9,
which is now redundant.
Fixes #5605
Change-Id: I6b4b942001baa54248b656c52dced3b62021c486
Reviewed-on: https://go-review.googlesource.com/22610
Run-TryBot: David du Colombier <0intro@gmail.com> Reviewed-by: David du Colombier <0intro@gmail.com>
Emmanuel Odeke [Wed, 4 May 2016 07:42:13 +0000 (01:42 -0600)]
runtime: print signal name in panic, if name is known
Adds a small function signame that infers a signal name
from the signal table, otherwise will fallback to using
hex(sig) as previously. No signal table is present for
Windows hence it will always print the hex value.
```shell
$ go run main.go &
$ kill -11 <pid>
fatal error: unexpected signal during runtime execution
[signal SIGSEGV: segmentation violation code=0x1 addr=0xb01dfacedebac1e
pc=0xc71db]
...
```
Fixes #13969
Change-Id: Ie6be312eb766661f1cea9afec352b73270f27f9d
Reviewed-on: https://go-review.googlesource.com/22753 Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
The following performance improvements have been made to the
low-level atomic functions for ppc64le & ppc64:
- For those cases containing a lwarx and stwcx (or other sizes):
sync, lwarx, maybe something, stwcx, loop to sync, sync, isync
The sync is moved before (outside) the lwarx/stwcx loop, and the
sync after is removed, so it becomes:
sync, lwarx, maybe something, stwcx, loop to lwarx, isync
- For the Or8 and And8, the shifting and manipulation of the
address to the word aligned version were removed and the
instructions were changed to use lbarx, stbcx instead of
register shifting, xor, then lwarx, stwcx.
- New instructions LWSYNC, LBAR, STBCC were tested and added.
runtime/atomic_ppc64x.s was changed to use the LWSYNC opcode
instead of the WORD encoding.
Fixes #15469
Ran some of the benchmarks in the runtime and sync directories.
Some results varied from run to run but the trend was improvement
based on best times for base and new:
If b has exactly one predecessor, as happens
frequently with static calls, we can make
lookupVarOutgoing generate less garbage.
Instead of generating a value that is just
going to be an OpCopy and then get eliminated,
loop. This can lead to lots of looping.
However, this loop is way cheaper than generating
lots of ssa.Values and then eliminating them.
For a subset of the code in #15537:
Before:
28.31 real 36.17 user 1.68 sys 2282450944 maximum resident set size
After:
9.63 real 11.66 user 0.51 sys 638144512 maximum resident set size
Updates #15537.
Excitingly, it appears that this also helps
regular code:
Ian Lance Taylor [Wed, 4 May 2016 20:27:27 +0000 (13:27 -0700)]
runtime: put tracebackctxt C functions in .c file
Since tracebackctxt.go uses //export functions, the C functions can't be
externally visible in the C comment. The code was using attributes to
work around that, but that failed on Windows.
Change-Id: If4449fd8209a8998b4f6855ea89e5db1471b2981
Reviewed-on: https://go-review.googlesource.com/22786 Reviewed-by: Minux Ma <minux@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Alex Brainman [Tue, 3 May 2016 06:37:33 +0000 (16:37 +1000)]
debug/pe: unexport newly introduced identifiers
CLs 22181, 22332 and 22336 intorduced new functionality to be used
in cmd/link (see issue #15345 for details). But we didn't have chance
to use new functionality yet. Unexport newly introduced identifiers,
so we don't have to commit to the API until we actually tried it.
Rename File.COFFSymbols into File._COFFSymbols,
COFFSymbol.FullName into COFFSymbol._FullName,
Section.Relocs into Section._Relocs,
Reloc into _Relocs,
File.StringTable into File._StringTable and
StringTable into _StringTable.
Updates #15345
Change-Id: I770eeb61f855de85e0c175225d5d1c006869b9ec
Reviewed-on: https://go-review.googlesource.com/22720 Reviewed-by: David Crawshaw <crawshaw@golang.org>
Run-TryBot: Alex Brainman <alex.brainman@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Michael Munday [Wed, 4 May 2016 18:26:46 +0000 (14:26 -0400)]
cmd/compile: fix uint64 to float casts on ppc64
Adds the FCFIDU instruction and uses it instead of the FCFID
instruction for unsigned integer to float casts. This change means
that unsigned integers do not have to be cast to signed integers
before being cast to a floating point value. Therefore it is no
longer necessary to insert instructions to detect and fix
values that overflow int64.
The previous code generating the uint64 to int64 cast handled
overflow by truncating the uint64 value. This truncation can
change the result of the rounding performed by the integer to
float cast.
The FCFIDU instruction was added in Power ISA 2.06B.
Cherry Zhang [Tue, 3 May 2016 20:49:54 +0000 (13:49 -0700)]
runtime/cgo: add context argument to crosscall2 on mips64
Change-Id: Id018516075842afd8af12fbf207763a851d5a851
Reviewed-on: https://go-review.googlesource.com/22754 Reviewed-by: Ian Lance Taylor <iant@golang.org>
Martin Möhrmann [Tue, 3 May 2016 09:10:26 +0000 (11:10 +0200)]
misc/cgo/fortran: fix gfortran compile test
Fixes #14544
Change-Id: I58b0b164ebbfeafe4ab32039a063df53e3018a6d
Reviewed-on: https://go-review.googlesource.com/22730 Reviewed-by: Ian Lance Taylor <iant@golang.org> Reviewed-by: Sean Lake <odysseus9672@gmail.com>
Michael Hudson-Doyle [Tue, 3 May 2016 23:23:24 +0000 (11:23 +1200)]
cmd/link: always read type data for dynimport symbols
Consider three shared libraries:
libBase.so -- defines a type T
lib2.so -- references type T
lib3.so -- also references type T, and something from lib2
lib2.so will contain a type symbol for T in its symbol table, but no
definition. If, when linking lib3.so the linker reads the symbols from lib2.so
before libBase.so, the linker didn't read the type data and later crashed.
The fix is trivial but the test change is a bit messy because the order the
linker reads the shared libraries in ends up depending on the order of the
import statements in the file so I had to rename one of the test packages so
that gofmt doesn't fix the test by accident...
Fixes #15516
Change-Id: I124b058f782c900a3a54c15ed66a0d91d0cde5ce
Reviewed-on: https://go-review.googlesource.com/22744
Run-TryBot: Michael Hudson-Doyle <michael.hudson@canonical.com>
TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Ian Lance Taylor <iant@golang.org>
Vishvananda Ishaya [Wed, 17 Feb 2016 01:58:11 +0000 (17:58 -0800)]
net: allow netgo to use lookup from nsswitch.conf
Change https://golang.org/cl/8945 allowed Go to use its own DNS resolver
instead of libc in a number of cases. The code parses nsswitch.conf and
attempts to resolve things in the same order. Unfortunately, builds with
netgo completely ignore this parsing and always search via
hostLookupFilesDNS.
This commit modifies the logic to allow binaries built with netgo to
parse nsswitch.conf and attempt to resolve using the order specified
there. If the parsing results in hostLookupCGo, it falls back to the
original hostLookupFilesDNS. Tests are also added to ensure that both
the parsing and the fallback work properly.
Fixes #14354
Change-Id: Ib079ad03d7036a4ec57f18352a15ba55d933f261
Reviewed-on: https://go-review.googlesource.com/19523
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Matthew Dempsky <mdempsky@google.com>
Robert Griesemer [Tue, 3 May 2016 00:03:36 +0000 (17:03 -0700)]
cmd/compile: use correct packages when exporting/importing _ (blank) names
1) Blank parameters cannot be accessed so the package doesn't matter.
Do not export it, and consistently use localpkg when importing a
blank parameter.
2) More accurately replicate fmt.go and parser.go logic when importing
a blank struct field. Blank struct fields get exported without
package qualification.
(This is actually incorrect, even with the old textual export format,
but we will fix that in a separate change. See also issue 15514.)
Fixes #15491.
Change-Id: I7978e8de163eb9965964942aee27f13bf94a7c3c
Reviewed-on: https://go-review.googlesource.com/22714 Reviewed-by: Matthew Dempsky <mdempsky@google.com>
1.7 traces embed symbol info and we now generate symbolized pprof profiles,
so we don't need the binary. Make binary argument optional as 1.5 traces
still need it.
Change-Id: I65eb13e3d20ec765acf85c42d42a8d7aae09854c
Reviewed-on: https://go-review.googlesource.com/22410 Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com> Reviewed-by: Austin Clements <austin@google.com>
Dmitry Vyukov [Fri, 26 Feb 2016 20:57:16 +0000 (21:57 +0100)]
runtime: per-P contexts for race detector
Race runtime also needs local malloc caches and currently uses
a mix of per-OS-thread and per-goroutine caches. This leads to
increased memory consumption. But more importantly cache of
synchronization objects is per-goroutine and we don't always
have goroutine context when feeing memory in GC. As the result
synchronization object descriptors leak (more precisely, they
can be reused if another synchronization object is recreated
at the same address, but it does not always help). For example,
the added BenchmarkSyncLeak has effectively runaway memory
consumption (based on a real long running server).
This change updates race runtime with support for per-P contexts.
BenchmarkSyncLeak now stabilizes at ~1GB memory consumption.
Long term, this will allow us to remove race runtime dependency
on glibc (as malloc is the main cornerstone).
I've also implemented a different scheme to pass P context to
race runtime: scheduler notified race runtime about association
between G and P by calling procwire(g, p)/procunwire(g, p).
But it turned out to be very messy as we have lots of places
where the association changes (e.g. syscalls). So I dropped it
in favor of the current scheme: race runtime asks scheduler
about the current P.
Fixes #14533
Change-Id: Iad10d2f816a44affae1b9fed446b3580eafd8c69
Reviewed-on: https://go-review.googlesource.com/19970 Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Dmitry Vyukov <dvyukov@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Dmitry Vyukov [Fri, 18 Mar 2016 15:34:11 +0000 (16:34 +0100)]
runtime: fix CPU underutilization
Runqempty is a critical predicate for scheduler. If runqempty spuriously
returns true, then scheduler can fail to schedule arbitrary number of
runnable goroutines on idle Ps for arbitrary long time. With the addition
of runnext runqempty predicate become broken (can spuriously return true).
Consider that runnext is not nil and the main array is empty. Runqempty
observes that the array is empty, then it is descheduled for some time.
Then queue owner pushes another element to the queue evicting runnext
into the array. Then queue owner pops runnext. Then runqempty resumes
and observes runnext is nil and returns true. But there were no point
in time when the queue was empty.
Fix runqempty predicate to not return true spuriously.
Michael Hudson-Doyle [Mon, 2 May 2016 02:46:40 +0000 (14:46 +1200)]
cmd/cgo: an approach to tsan that works with gcc
GCC, unlike clang, does not provide any way for code being compiled to tell if
-fsanitize-thread was passed. But cgo can look to see if that flag is being
passed and generate different code in that case.
Fixes #14602
Change-Id: I86cb5318c2e35501ae399618c05af461d1252d2d
Reviewed-on: https://go-review.googlesource.com/22688
Run-TryBot: Michael Hudson-Doyle <michael.hudson@canonical.com>
TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Ian Lance Taylor <iant@golang.org>
Change-Id: I4f235cee0a62ec435f9e8540a1ec08ae03b1a75f
Reviewed-on: https://go-review.googlesource.com/21819 Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Ian Lance Taylor [Mon, 2 May 2016 00:03:46 +0000 (17:03 -0700)]
cmd/cgo, misc/cgo/test: make -Wdeclaration-after-statement clean
I got a complaint that cgo output triggers warnings with
-Wdeclaration-after-statement. I don't think it's worth testing for
this--C has permitted declarations after statements since C99--but it is
easy enough to fix. It may break again; so it goes.
This CL also fixes errno handling to avoid getting confused if the tsan
functions happen to change the global errno variable.
Change-Id: I0ec7c63a6be5653ef44799d134c8d27cb5efa441
Reviewed-on: https://go-review.googlesource.com/22686
Run-TryBot: Ian Lance Taylor <iant@golang.org> Reviewed-by: Minux Ma <minux@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Mikio Hara [Sun, 1 May 2016 19:44:46 +0000 (04:44 +0900)]
net/http: gofmt -w -s
Change-Id: I7e07888e90c7449f119e74b97995efcd7feef76e
Reviewed-on: https://go-review.googlesource.com/22682 Reviewed-by: Ian Lance Taylor <iant@golang.org>
The Transport's automatic gzip uncompression lost information in the
process (the compressed Content-Length, if known). Normally that's
okay, but it's not okay for reverse proxies which have to be able to
generate a valid HTTP response from the Transport's provided
*Response.
Reverse proxies should normally be disabling compression anyway and
just piping the compressed pipes though and not wasting CPU cycles
decompressing them. So also document that on the new Uncompressed
field.
Then, using the new field, fix Response.Write to not inject a bogus
"Connection: close" header when it doesn't see a transfer encoding or
content-length.
Updates #15366 (the http2 side remains, once this is submitted)
Brad Fitzpatrick [Sun, 1 May 2016 04:11:26 +0000 (21:11 -0700)]
net/http: provide access to the listener address an HTTP request arrived on
This adds a context key named LocalAddrContextKey (for now, see #15229) to
let users access the net.Addr of the net.Listener that accepted the connection
that sent an HTTP request. This is similar to ServerContextKey which provides
access to the *Server. (A Server may have multiple Listeners)
Brad Fitzpatrick [Sun, 1 May 2016 02:11:42 +0000 (21:11 -0500)]
net/http: add Transport.IdleConnTimeout
Don't keep idle HTTP client connections open forever. Add a new knob,
Transport.IdleConnTimeout, and make the default be 90 seconds. I
figure 90 seconds is more than a minute, and less than infinite, and I
figure enough code has things waking up once a minute polling APIs.
This also removes the Transport's idleCount field which was unused and
redundant with the size of the idleLRU map (which was actually used).
The B/op number is effectively meaningless. There
is a surprisingly large one-time cost that gets
divided by the number of iterations that your
machine can get through in a second.
This CL discards the first run, which helps.
It is not a panacea. Running with -benchtime=10s
will allow the sync.Pool to be emptied,
which brings the problem back.
However, since there are more iterations to divide
the cost through, it’s not quite as bad,
and running with a high benchtime is rare.
This CL changes the meaning of the B/op number,
which is unfortunate, since it won’t have the
same order of magnitude as previous Go versions.
But it wasn’t really comparable before anyway,
since it didn’t have any reliable meaning at all.