]> Cypherpunks repositories - gostls13.git/log
gostls13.git
10 years agotime: deflake TestAfterQueuing
Brad Fitzpatrick [Wed, 6 May 2015 17:54:20 +0000 (10:54 -0700)]
time: deflake TestAfterQueuing

Ramp up the delay on subsequent attempts. Fast builders have the same delay.

Not a perfect fix, but should make it better. And this easy.

Fixes #9903 maybe
Fixes #10680 maybe

Change-Id: I967380c2cb8196e6da9a71116961229d37b36335
Reviewed-on: https://go-review.googlesource.com/9795
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
10 years agoos: eradicate smallpox after test
Brad Fitzpatrick [Wed, 6 May 2015 16:53:48 +0000 (09:53 -0700)]
os: eradicate smallpox after test

Otherwise:

$ go test  -short -cpu=1,1,2,2
--- FAIL: TestLookupEnv (0.00s)
        env_test.go:102: SMALLPOX="virus"
--- FAIL: TestLookupEnv-2 (0.00s)
        env_test.go:102: SMALLPOX="virus"
--- FAIL: TestLookupEnv-2 (0.00s)
        env_test.go:102: SMALLPOX="virus"

Change-Id: Ic1f6dd1bae3c79c4f7da02bc8c30b5e599627a82
Reviewed-on: https://go-review.googlesource.com/9794
Reviewed-by: Rob Pike <r@golang.org>
10 years agonet: always use cgo for DNS on Android
Brad Fitzpatrick [Wed, 6 May 2015 16:32:11 +0000 (09:32 -0700)]
net: always use cgo for DNS on Android

Android has (had?) its own local DNS resolver daemon, also my fault:

https://android.googlesource.com/platform/system/netd/+/007e987fee7e815e0c4bc820f434a632b7a69a9d

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>

10 years agodoc/go1.5.txt: LookupEnv
Rob Pike [Wed, 6 May 2015 16:30:52 +0000 (09:30 -0700)]
doc/go1.5.txt: LookupEnv

Change-Id: Iff27fa0ca50fe9e41d811d30df41fc2d3057aa1d
Reviewed-on: https://go-review.googlesource.com/9792
Reviewed-by: Rob Pike <r@golang.org>
10 years agocmd/doc: if no top-level symbols match, look for methods
Rob Pike [Mon, 4 May 2015 17:11:27 +0000 (10:11 -0700)]
cmd/doc: if no top-level symbols match, look for methods

Improving the usability further.

Before:

$ go doc bytes.Read
doc: symbol Read not present in package bytes installed in "bytes"
$

After:
$ go doc bytes.Read
func (b *Buffer) Read(p []byte) (n int, err error)
    Read reads the next len(p) bytes from the buffer or until the buffer is drained.
    The return value n is the number of bytes read. If the buffer has no data to
    return, err is io.EOF (unless len(p) is zero); otherwise it is nil.
func (r *Reader) Read(b []byte) (n int, err error)
$

Change-Id: I646511fada138bd09e9b39820da01a5ccef4a90f
Reviewed-on: https://go-review.googlesource.com/9656
Reviewed-by: Russ Cox <rsc@golang.org>
10 years agoCONTRIBUTORS: add Burcu Dogan's personal mail
Burcu Dogan [Wed, 6 May 2015 14:41:12 +0000 (10:41 -0400)]
CONTRIBUTORS: add Burcu Dogan's personal mail

Change-Id: I2bc92f6d33db44f96df4219e6144393d5150fe0f
Reviewed-on: https://go-review.googlesource.com/9785
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
10 years agoos: rewrite LookupEnv's test
Rob Pike [Wed, 6 May 2015 15:26:51 +0000 (08:26 -0700)]
os: rewrite LookupEnv's test

GOROOT is not dependably set.

When I first wrote this test, I thought it was a waste of time
because the function can't fail if the other environment functions
work, but I didn't want to add functionality without testing it.
Of course, the test broke, and I learned something: GOROOT is not
set on iOS or, to put it more broadly, the world continues to
surprise me with its complexity and horror, such as a version of
cat with syntax coloring.

In that vein, I built this test around smallpox.

Change-Id: Ifa6c218a927399d05c47954fdcaea1015e558fb6
Reviewed-on: https://go-review.googlesource.com/9791
Reviewed-by: Russ Cox <rsc@golang.org>
10 years agoruntime: remove unused GC times from api.next
Rick Hudson [Wed, 6 May 2015 13:10:27 +0000 (09:10 -0400)]
runtime: remove unused GC times from api.next

Updates api boilerplate in seperate CL see commit 1845314 for code changes.
Fixes #10462

Change-Id: I4e28dbdcdd693688835bcd1d4b0224454aa7154d
Reviewed-on: https://go-review.googlesource.com/9784
Reviewed-by: Austin Clements <austin@google.com>
10 years agoruntime: remove unused GC timers
Rick Hudson [Mon, 4 May 2015 18:03:07 +0000 (14:03 -0400)]
runtime: remove unused GC timers

During development some tracing routines were added that are not
needed in the release. These included GCstarttimes, GCendtimes, and
GCprinttimes.
Fixes #10462

Change-Id: I0788e6409d61038571a5ae0cbbab793102df0a65
Reviewed-on: https://go-review.googlesource.com/9689
Reviewed-by: Austin Clements <austin@google.com>
10 years agonet: fix inconsistent error values on Read for solaris
Mikio Hara [Wed, 6 May 2015 12:34:29 +0000 (21:34 +0900)]
net: fix inconsistent error values on Read for solaris

Updates #4856.

Change-Id: Ia04e24fb1fe57e244d7b1cd417f7f419ad610acd
Reviewed-on: https://go-review.googlesource.com/9776
Reviewed-by: Aram Hăvărneanu <aram@mgk.ro>
10 years agonet: fix comment in sendFile
Aram Hăvărneanu [Mon, 23 Mar 2015 20:33:08 +0000 (21:33 +0100)]
net: fix comment in sendFile

Change-Id: Iacee13150b283f9d2867a7ca98f805900f7cbe50
Reviewed-on: https://go-review.googlesource.com/7943
Reviewed-by: Minux Ma <minux@golang.org>
10 years agonet: link with networking libraries when net package is in use
Aram Hăvărneanu [Mon, 23 Mar 2015 12:52:46 +0000 (13:52 +0100)]
net: link with networking libraries when net package is in use

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>
10 years agosyscall, net: use sendfile on Solaris
Aram Hăvărneanu [Mon, 23 Mar 2015 12:51:12 +0000 (13:51 +0100)]
syscall, net: use sendfile on Solaris

Updates #5847.

Change-Id: Ic93f2e5f9a6aa3bd49cf75b16474ec5e897d17e1
Reviewed-on: https://go-review.googlesource.com/7940
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Reviewed-by: Minux Ma <minux@golang.org>
Reviewed-by: Rob Pike <r@golang.org>
10 years agonet: try to fix setKeepAlivePeriod on Solaris
Aram Hăvărneanu [Tue, 17 Mar 2015 12:50:40 +0000 (13:50 +0100)]
net: try to fix setKeepAlivePeriod on Solaris

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>
10 years agoruntime, syscall: link Solaris binaries directly instead of using dlopen/dlsym
Aram Hăvărneanu [Tue, 21 Apr 2015 13:12:29 +0000 (15:12 +0200)]
runtime, syscall: link Solaris binaries directly instead of using dlopen/dlsym

Before CL 8214 (use .plt instead of .got on Solaris) Solaris used a
dynamic linking scheme that didn't permit lazy binding. To speed program
startup, Go binaries only used it for a small number of symbols required
by the runtime. Other symbols were resolved on demand on first use, and
were cached for subsequent use. This required some moderately complex
code in the syscall package.

CL 8214 changed the way dynamic linking is implemented, and now lazy
binding is supported. As now all symbols are resolved lazily by the
dynamic loader, there is no need for the complex code in the syscall
package that did the same. This CL makes Go programs link directly
with the necessary shared libraries and deletes the lazy-loading code
implemented in Go.

Change-Id: Ifd7275db72de61b70647242e7056dd303b1aee9e
Reviewed-on: https://go-review.googlesource.com/9184
Reviewed-by: Minux Ma <minux@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
10 years agogo/build: enable cgo by default on solaris/amd64
Aram Hăvărneanu [Mon, 30 Mar 2015 21:15:51 +0000 (23:15 +0200)]
go/build: enable cgo by default on solaris/amd64

Change-Id: I0110b01fe4c64851ac2cfb5a92c31ce156831bc8
Reviewed-on: https://go-review.googlesource.com/8265
Reviewed-by: Minux Ma <minux@golang.org>
10 years agodoc/progs: disable cgo tests that use C.Stdout on Solaris
Aram Hăvărneanu [Mon, 30 Mar 2015 21:12:04 +0000 (23:12 +0200)]
doc/progs: disable cgo tests that use C.Stdout on Solaris

Solaris, like Windows, NetBSD and OpenBSD, uses macros for stdin, stdout,
and stderr. Cgo can't access them without getters/setters written in
C. Because of this we disable affected tests like for the other platforms.

Updates #10715.

Change-Id: I3d33a5554b5ba209273dbdff992925a38a281b42
Reviewed-on: https://go-review.googlesource.com/8264
Reviewed-by: Minux Ma <minux@golang.org>
Reviewed-by: Russ Cox <rsc@golang.org>
10 years agomisc/cgo: add various solaris build lines
Aram Hăvărneanu [Mon, 30 Mar 2015 21:11:36 +0000 (23:11 +0200)]
misc/cgo: add various solaris build lines

Change-Id: Ifd9ac7f5300232fb83c6350a787b5803adb96b48
Reviewed-on: https://go-review.googlesource.com/8263
Reviewed-by: Minux Ma <minux@golang.org>
10 years agonet: add cgo support for Solaris
Aram Hăvărneanu [Mon, 30 Mar 2015 21:07:18 +0000 (23:07 +0200)]
net: add cgo support for Solaris

Change-Id: Ib66bebd418d97f38956970f93e69aa41e7c55523
Reviewed-on: https://go-review.googlesource.com/8262
Reviewed-by: Minux Ma <minux@golang.org>
Reviewed-by: Mikio Hara <mikioh.mikioh@gmail.com>
10 years agoos/user: small fixes for Solaris
Aram Hăvărneanu [Mon, 30 Mar 2015 21:06:06 +0000 (23:06 +0200)]
os/user: small fixes for Solaris

Change-Id: I56149ef6607fb4d9baff9047cb3a47d71cad6fa6
Reviewed-on: https://go-review.googlesource.com/8261
Reviewed-by: Minux Ma <minux@golang.org>
10 years agoruntime/cgo: add cgo support for solaris/amd64
Aram Hăvărneanu [Mon, 30 Mar 2015 21:03:09 +0000 (23:03 +0200)]
runtime/cgo: add cgo support for solaris/amd64

Change-Id: Ic9744c7716cdd53f27c6e5874230963e5fff0333
Reviewed-on: https://go-review.googlesource.com/8260
Reviewed-by: Minux Ma <minux@golang.org>
10 years agocmd/ld: make .rela and .rela.plt sections contiguous
Aram Hăvărneanu [Mon, 30 Mar 2015 13:45:33 +0000 (15:45 +0200)]
cmd/ld: make .rela and .rela.plt sections contiguous

ELF normally requires this and Solaris runtime loader will crash if we
don't do it.

Fixes Solaris build.

Change-Id: I0482eed890aff2d346136ae7f9caf8f094f502ed
Reviewed-on: https://go-review.googlesource.com/8216
Reviewed-by: Ian Lance Taylor <iant@golang.org>
10 years agoruntime: always load address of libcFunc on Solaris
Aram Hăvărneanu [Mon, 30 Mar 2015 11:52:07 +0000 (13:52 +0200)]
runtime: always load address of libcFunc on Solaris

The linker always uses .plt for externals, so libcFunc is now an actual
external symbol instead of a pointer to one.

Fixes most of the breakage introduced in previous CL.

Change-Id: I64b8c96f93127f2d13b5289b024677fd3ea7dbea
Reviewed-on: https://go-review.googlesource.com/8215
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Reviewed-by: Minux Ma <minux@golang.org>
10 years agocmd/6l: use .plt instead of .got on Solaris
Aram Hăvărneanu [Mon, 30 Mar 2015 11:46:28 +0000 (13:46 +0200)]
cmd/6l: use .plt instead of .got on Solaris

Solaris requires all external procedures to be accessed through the
PLT. If 6l won't do it, /bin/ld will, so all the code written with .GOT
in mind won't work with the external linker.

This CL makes external linking work, opening the path to cgo support
on Solaris.

This CL breaks the Solaris build, this is fixed in subsequent CLs in
this series.

Change-Id: If370a79f49fdbe66d28b89fa463b4f3e91685f69
Reviewed-on: https://go-review.googlesource.com/8214
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Reviewed-by: Minux Ma <minux@golang.org>
10 years agonet: simplify error messages in tests
Mikio Hara [Fri, 1 May 2015 03:38:42 +0000 (12:38 +0900)]
net: simplify error messages in tests

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>

10 years agonet: deflake listener tests
Mikio Hara [Thu, 30 Apr 2015 22:49:12 +0000 (07:49 +0900)]
net: deflake listener tests

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>
10 years agodoc/go1.5.txt: mention GOARCH reservation change
Shenghou Ma [Wed, 6 May 2015 01:17:24 +0000 (21:17 -0400)]
doc/go1.5.txt: mention GOARCH reservation change

Change-Id: Ie0ba7a5cb860e6fff68ec3042764c7e026b1234c
Reviewed-on: https://go-review.googlesource.com/9781
Reviewed-by: Minux Ma <minux@golang.org>
10 years agoimage/gif: allow encoding a single-frame image whose top-left corner
Nigel Tao [Tue, 5 May 2015 07:39:09 +0000 (17:39 +1000)]
image/gif: allow encoding a single-frame image whose top-left corner
isn't (0, 0).

Also fix a s/b.Min.X/b.Max.X/ typo in bounds checking.

Fixes #10676

Change-Id: Ie5ff7ec20ca30367a8e65d32061959a2d8e089e9
Reviewed-on: https://go-review.googlesource.com/9712
Reviewed-by: Rob Pike <r@golang.org>
10 years agocmd/internal/ld: recompute resoff after changing elfreserve
Michael Hudson-Doyle [Wed, 6 May 2015 00:35:19 +0000 (12:35 +1200)]
cmd/internal/ld: recompute resoff after changing elfreserve

My last update to https://golang.org/cl/9710 missed this.

Change-Id: Ie042032ca307e1065fcf6c402a166f9ff74027e7
Reviewed-on: https://go-review.googlesource.com/9771
Reviewed-by: Ian Lance Taylor <iant@golang.org>
10 years agonet: enable unixpacket tests on openbsd
Mikio Hara [Sun, 3 May 2015 14:29:34 +0000 (23:29 +0900)]
net: enable unixpacket tests on openbsd

Change-Id: I0e1519d429a295faa70013687d6faf2f5ce0be24
Reviewed-on: https://go-review.googlesource.com/9713
Reviewed-by: Ian Lance Taylor <iant@golang.org>
10 years agosyscall: fix TestSCMCredentials
Mikio Hara [Tue, 5 May 2015 15:02:30 +0000 (00:02 +0900)]
syscall: fix TestSCMCredentials

Fixes #10703.

Change-Id: I55c0f07625a0847fb27defa9891af6db6eb21f82
Reviewed-on: https://go-review.googlesource.com/9714
Reviewed-by: Ian Lance Taylor <iant@golang.org>
10 years agocmd/6l, cmd/internal/ld: handle R_PCREL to function in other shared library
Michael Hudson-Doyle [Tue, 5 May 2015 04:10:12 +0000 (16:10 +1200)]
cmd/6l, cmd/internal/ld: handle R_PCREL to function in other shared library

An ELF linker handles a PC-relative reference to an STT_FUNC defined in a
shared library by building a PLT entry and referring to that, so do the
same in 6l.

Fixes #10690

Change-Id: I061a96fd4400d957e301d0ac86760ce256910e1d
Reviewed-on: https://go-review.googlesource.com/9711
Run-TryBot: Michael Hudson-Doyle <michael.hudson@canonical.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
10 years agonet/http: don't send implicit Content-Length if Transfer-Encoding is set
Brad Fitzpatrick [Fri, 1 May 2015 14:02:36 +0000 (07:02 -0700)]
net/http: don't send implicit Content-Length if Transfer-Encoding is set

Fixes #9987

Change-Id: Ibebd105a2bcdc1741f3b41aa78cb986f3f518b53
Reviewed-on: https://go-review.googlesource.com/9638
Reviewed-by: Andrew Gerrand <adg@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>

10 years agotext/template: check for malformed pipelines
Didier Spezia [Tue, 5 May 2015 06:45:49 +0000 (06:45 +0000)]
text/template: check for malformed pipelines

Catch some malformed pipelines at parsing time.
The current code accepts pipelines such as:

{{12|.}}
{{"hello"|print|false}}
{{.|"blah blah"}}

Such pipelines generate panic in html/template at execution time.

Add an extra check to verify all the commands of the pipeline are executable
(except for the first one).

Fixes #10610

Change-Id: Id72236ba8f76a59fa284fe3d4c2cb073e50b51f1
Reviewed-on: https://go-review.googlesource.com/9626
Reviewed-by: Rob Pike <r@golang.org>
10 years agocmd/internal/ld: reserve space for package list note when -buildmode=shared
Michael Hudson-Doyle [Tue, 5 May 2015 02:17:07 +0000 (14:17 +1200)]
cmd/internal/ld: reserve space for package list note when -buildmode=shared

This makes the intermediate object file a little bigger but it doesn't waste
any space in the final shared library.

Fixes #10691

Change-Id: Ic51a571d60291f1ac2dad1b50dba4679643168ae
Reviewed-on: https://go-review.googlesource.com/9710
Reviewed-by: Ian Lance Taylor <iant@golang.org>
10 years agocmd/go: rebuild stale shared objects before linking against them.
Michael Hudson-Doyle [Tue, 14 Apr 2015 10:03:21 +0000 (12:03 +0200)]
cmd/go: rebuild stale shared objects before linking against them.

This changes the action graph when shared libraries are involved to always have
an action for the shared library (which does nothing when the shared library
is up to date).

Change-Id: Ibbc70fd01cbb3f4e8c0ef96e62a151002d446144
Reviewed-on: https://go-review.googlesource.com/8934
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

10 years agocmd/go: fix linux-amd64-clang builder
Dave Cheney [Fri, 1 May 2015 00:49:36 +0000 (10:49 +1000)]
cmd/go: fix linux-amd64-clang builder

Fixes #10660

Fix the clang only builder by passing -extld down to the linker when needed.
The build passed on most hosts because gcc is almost always present. The bug
was verified by symlinking bin/false in place of gcc in my $PATH and running
the build.

Also, resolve a TODO and move the support logic into its own function.

Tested manually

    env CC=clang-3.5 ./all.bash # linux/amd64
    env CC=gcc-4.8 ./all.bash # linux/amd64
    ./all.bash # linux/amd64
    ./all.bash # darwin/amd64

Change-Id: I4e27a1119356e295500a0d19ad7a4ec14207bf10
Reviewed-on: https://go-review.googlesource.com/9526
Run-TryBot: Dave Cheney <dave@cheney.net>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
10 years agofmt: prevent panic from %.[]
Fabrizio (Misto) Milo [Sun, 3 May 2015 21:22:33 +0000 (14:22 -0700)]
fmt: prevent panic from %.[]

Fixes #10675

Change-Id: Ia057427ce3e81d35f1ba6c354868a0ad6cc9abf2
Reviewed-on: https://go-review.googlesource.com/9636
Reviewed-by: Rob Pike <r@golang.org>
10 years agoos: add LookupEnv, like Getenv but reports presence
Rob Pike [Tue, 5 May 2015 18:17:08 +0000 (11:17 -0700)]
os: add LookupEnv, like Getenv but reports presence

Fixes #9676.

Change-Id: I32fe474cdfa09aff91daa4b10ac4df28ffdaa649
Reviewed-on: https://go-review.googlesource.com/9741
Reviewed-by: Ian Lance Taylor <iant@golang.org>
10 years agomisc/ios: fix plist indentation and whitespace
Burcu Dogan [Sun, 3 May 2015 07:13:46 +0000 (00:13 -0700)]
misc/ios: fix plist indentation and whitespace

Change-Id: Ida727edb592e77918ca5511b41456786d57c97b2
Reviewed-on: https://go-review.googlesource.com/9634
Reviewed-by: David Crawshaw <crawshaw@golang.org>
10 years agoRevert "cmd/internal: disable OSQRT on GOARM=5"
Minux Ma [Tue, 5 May 2015 07:35:08 +0000 (07:35 +0000)]
Revert "cmd/internal: disable OSQRT on GOARM=5"

I just submitted the real fix for #10641.

This reverts commit 3120adc21294829c230005a89e9bb0ad63de606b.

Change-Id: I55051515f697e27ca887ed21c2ac985f0b9b062b
Reviewed-on: https://go-review.googlesource.com/9720
Reviewed-by: Joel Sing <jsing@google.com>
10 years agofmt: document that Scanf returns an error the same as Scan
Rob Pike [Tue, 5 May 2015 18:05:35 +0000 (11:05 -0700)]
fmt: document that Scanf returns an error the same as Scan

No semantic change.

Fixes #8708.

Change-Id: Ieda04a86a19bb69bfc2519d381a2f025e7cb8279
Reviewed-on: https://go-review.googlesource.com/9740
Reviewed-by: Ian Lance Taylor <iant@golang.org>
10 years agoruntime: rename ptrsize to ptrdata
Russ Cox [Tue, 5 May 2015 01:43:30 +0000 (21:43 -0400)]
runtime: rename ptrsize to ptrdata

I forgot there is already a ptrSize constant.
Rename field to avoid some confusion.

Change-Id: I098fdcc8afc947d6c02c41c6e6de24624cc1c8ff
Reviewed-on: https://go-review.googlesource.com/9700
Reviewed-by: Austin Clements <austin@google.com>
10 years agofmt: change the overflow test for large numbers in verbs
Rob Pike [Tue, 5 May 2015 15:59:18 +0000 (08:59 -0700)]
fmt: change the overflow test for large numbers in verbs

The old one was inferior.

Fixes #10695.

Change-Id: Ia7fb88c9ceb1b10197b77a54f729865385288d98
Reviewed-on: https://go-review.googlesource.com/9709
Reviewed-by: Dmitry Vyukov <dvyukov@google.com>
10 years agotext/template: shut down lexing goroutine on error
Rob Pike [Mon, 4 May 2015 20:09:31 +0000 (13:09 -0700)]
text/template: shut down lexing goroutine on error

When a parse error occurred, the lexing goroutine would lay idle.
It's not likely a problem but if the program is for some reason
accepting badly formed data repeatedly, it's wasteful.

The solution is easy: Just drain the input on error. We know this
will succeed because the input is always a string and is therefore
guaranteed finite.

With debugging prints in the package tests I've shown this is effective,
shutting down 79 goroutines that would otherwise linger, out of 123 total.

Fixes #10574.

Change-Id: I8aa536e327b219189a7e7f604a116fa562ae1c39
Reviewed-on: https://go-review.googlesource.com/9658
Reviewed-by: Russ Cox <rsc@golang.org>
10 years agoruntime: let freezetheworld work even when gomaxprocs=1
Keith Randall [Thu, 30 Apr 2015 21:56:35 +0000 (14:56 -0700)]
runtime: let freezetheworld work even when gomaxprocs=1

Freezetheworld still has stuff to do when gomaxprocs=1.
In particular, signals can come in on other Ms (like the GC M, say)
and the single user M is still running.

Fixes #10546

Change-Id: I2f07f17d1c81e93cf905df2cb087112d436ca7e7
Reviewed-on: https://go-review.googlesource.com/9551
Reviewed-by: Dmitry Vyukov <dvyukov@google.com>
10 years agonet: fix inconsistent errors
Mikio Hara [Tue, 21 Apr 2015 12:20:15 +0000 (21:20 +0900)]
net: fix inconsistent errors

These a series of changes fix inconsistent errors on the package net
APIs. Now almost all the APIs return OpError as a common error type
except Lookup, Resolve and Parse APIs. The Lookup, Resolve and Parse
APIs return more specific errors such as DNSError, AddrError or
ParseError.

An OpError may contain nested error information. For example, Dial may
return an OpError containing a DNSError, AddrError, unexposed type/value
or other package's type/value like the following:
OpError{/* dial info */, Err: &DNSError{}}
OpError{/* dial info */, Err: &AddrError{}}
OpError{/* dial info */, Err: <unexposed type or value>}
OpError{/* dial info */, Err: <other package's type or value>}

and Read and Write may return an OpError containing other OpError when
an application uses io.Copy or similar:
OpError{/* for io.Reader */, Err: &OpError{/* for io.Writer */}}

When an endpoint is created for connection-oriented byte-stream
protocols, Read may return an io.EOF when the connection is closed by
remote endpoint.

Fixes #4856.

A series of changes:
- net: fix inconsistent error values on Dial, Listen partially
  https://go.googlesource.com/go/+/89b7c66d0d14462fd7893be4290bdfe5f9063ae1
- net: fix inconsistent error values on Read
  https://go.googlesource.com/go/+/ec1144423f45e010c72363fe59291d43214b6e31
- net: fix inconsistent error values on Write
  https://go.googlesource.com/go/+/11b5f98bf0d5eb8854f735cc332c912725070214
- net: fix inconsistent error values on Close
  https://go.googlesource.com/go/+/310db63c5bc121e7bfccb494c01a6b91a257e7fc
- net: fix inconsistent error values on Accept
  https://go.googlesource.com/go/+/4540e162b1aefda8157372764ad3d290a414ef1d
- net: fix inconsistent error values on File
  https://go.googlesource.com/go/+/885111365ba0a74421059bfbd18f4c57c1e70332
- net: fix inconsistent error values on setters
  https://go.googlesource.com/go/+/2173a27903897c481b0a0daf3ca3e0a0685701db
- net: fix inconsistent error values on Interface
  https://go.googlesource.com/go/+/456cf0f22c93e1a6654980f4a48a564555f6c8a2
- net: fix inconsistent error values on Lookup
  https://go.googlesource.com/go/+/0fc582e87942b2e52bed751b6c56660ba99e9a7d
- net: add Source field to OpError
  https://go.googlesource.com/go/+/afd2d2b6df3ebfe99faf347030f15adfdf422fa0

Change-Id: Id678e369088dc9fbe9073cfe7ff8a8754a57d61f
Reviewed-on: https://go-review.googlesource.com/9236
Reviewed-by: Ian Lance Taylor <iant@golang.org>
10 years agonet: add missing ReadFrom, WriteTo deadline tests
Mikio Hara [Sat, 2 May 2015 08:58:06 +0000 (17:58 +0900)]
net: add missing ReadFrom, WriteTo deadline tests

Change-Id: If84edfaec361ca2fbb75707c4ad30e4ce64f7013
Reviewed-on: https://go-review.googlesource.com/9664
Reviewed-by: Ian Lance Taylor <iant@golang.org>
10 years agoruntime: fix software FP regs corruption when emulating SQRT on ARM
Shenghou Ma [Thu, 30 Apr 2015 23:03:31 +0000 (19:03 -0400)]
runtime: fix software FP regs corruption when emulating SQRT on ARM

When emulating ARM FSQRT instruction, the sqrt function itself
should not use any floating point arithmetics, otherwise it will
clobber the user software FP registers.

Fortunately, the sqrt function only uses floating point instructions
to test for corner cases, so it's easy to make that function does
all it job using pure integer arithmetic only. I've verified that
after this change, runtime.stepflt and runtime.sqrt doesn't contain
any call to _sfloat. (Perhaps we should add //go:nosfloat to make
the compiler enforce this?)

Fixes #10641.

Change-Id: Ida4742c49000fae4fea4649f28afde630ce4c576
Signed-off-by: Shenghou Ma <minux@golang.org>
Reviewed-on: https://go-review.googlesource.com/9570
Reviewed-by: Dave Cheney <dave@cheney.net>
Reviewed-by: Keith Randall <khr@golang.org>
10 years agoimage/gif: be consistent wrt "color map" or "color table" names.
Nigel Tao [Mon, 4 May 2015 06:47:10 +0000 (16:47 +1000)]
image/gif: be consistent wrt "color map" or "color table" names.

The spec at http://www.w3.org/Graphics/GIF/spec-gif89a.txt always says
"color table" and not "color map".

Change-Id: I4c172e3ade15618cbd616629822ce7d109a200af
Reviewed-on: https://go-review.googlesource.com/9668
Reviewed-by: Rob Pike <r@golang.org>
10 years agocmd/internal: disable OSQRT on GOARM=5
Joel Sing [Thu, 30 Apr 2015 18:19:34 +0000 (04:19 +1000)]
cmd/internal: disable OSQRT on GOARM=5

OSQRT currently produces incorrect results when used on arm with softfloat.
Disable it on GOARM=5 until the actual problem is found and fixed.

Updates #10641

Change-Id: Ia6f6879fbbb05cb24399c2feee93c1be21113e73
Reviewed-on: https://go-review.googlesource.com/9524
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Reviewed-by: Dave Cheney <dave@cheney.net>
10 years agogo/build: reserve GOARCH values for all common architectures
Shenghou Ma [Sat, 2 May 2015 02:09:54 +0000 (22:09 -0400)]
go/build: reserve GOARCH values for all common architectures

Whenever we introduce a new GOARCH, older Go releases won't
recognize them and this causes trouble for both our users and
us (we need to add unnecessary build tags).

Go 1.5 has introduced three new GOARCHes so far: arm64 ppc64
ppc64le, we can take the time to introduce GOARCHes for all
common architectures that Go might support in the future to
avoid the problem.

Fixes #10165.

Change-Id: Ida4f9112897cfb1e85b06538db79125955ad0f4c
Reviewed-on: https://go-review.googlesource.com/9644
Reviewed-by: Ian Lance Taylor <iant@golang.org>
10 years agocmd/internal/gc: Use shifts for powers-of-two indexing
Keith Randall [Thu, 30 Apr 2015 17:41:57 +0000 (10:41 -0700)]
cmd/internal/gc: Use shifts for powers-of-two indexing

Fixes #10638

Change-Id: I7bbaad7e5a599aa94d1d158e903596231c7e9897
Reviewed-on: https://go-review.googlesource.com/9535
Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
10 years agocmd/internal/ld: delete Biobuf
Dave Cheney [Sat, 2 May 2015 02:44:49 +0000 (12:44 +1000)]
cmd/internal/ld: delete Biobuf

Update #10652

This proposal deletes cmd/internal/ld.Biobuf and replaces all uses with
cmd/internal/obj.Biobuf. As cmd/internal/ld already imported cmd/internal/obj
there are no additional dependencies created.

Notes:

- ld.Boffset included more checks, so it was merged into obj.Boffset
- obj.Bflush was removed in 8d16253c90ae, so replaced all calls to
  ld.Bflush, with obj.Biobuf.Flush.
- Almost all of this change was prepared with sed.

Change-Id: I814854d52f5729a5a40c523c8188e465246b88da
Reviewed-on: https://go-review.googlesource.com/9660
Reviewed-by: Keith Randall <khr@golang.org>
Run-TryBot: Dave Cheney <dave@cheney.net>

10 years agogo/internal/gcimporter, go/types: also skip tests on nacl/arm
Shenghou Ma [Sun, 3 May 2015 05:29:52 +0000 (01:29 -0400)]
go/internal/gcimporter, go/types: also skip tests on nacl/arm

Change-Id: I3e839587626832da069d95a7d7389ea6bb2318da
Reviewed-on: https://go-review.googlesource.com/9674
Reviewed-by: Ian Lance Taylor <iant@golang.org>
10 years agoarchive/tar: fix error message
David du Colombier [Fri, 24 Apr 2015 13:37:53 +0000 (15:37 +0200)]
archive/tar: fix error message

Write should return ErrWriteAfterClose instead
of ErrWriteTooLong when called after Close.

Change-Id: If5ec4ef924e4c56489e0d426976f7e5fad79be9b
Reviewed-on: https://go-review.googlesource.com/9259
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
10 years agoruntime: add pointer size to type structure
Austin Clements [Mon, 4 May 2015 18:37:45 +0000 (14:37 -0400)]
runtime: add pointer size to type structure

This adds a field to the runtime type structure that records the size
of the prefix of objects of that type containing pointers. Any data
after this offset is scalar data.

This is necessary for shrinking the type bitmaps to 1 bit and will
help the garbage collector efficiently estimate the amount of heap
that needs to be scanned.

Change-Id: I1318d79e6360dca0ac980245016c562e61f52ff5
Reviewed-on: https://go-review.googlesource.com/9691
Reviewed-by: Russ Cox <rsc@golang.org>
Run-TryBot: Austin Clements <austin@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

10 years agocmd/internal/ld: put all type decoding in decodesym.go
Austin Clements [Mon, 4 May 2015 18:35:35 +0000 (14:35 -0400)]
cmd/internal/ld: put all type decoding in decodesym.go

Move the one instance of type structure decoding in the linker that
doesn't live decodesym.go in to decodesym.go.

Change-Id: Ic6a23500deb72f0e9c8227ab611511e9781fac70
Reviewed-on: https://go-review.googlesource.com/9690
Reviewed-by: Russ Cox <rsc@golang.org>
10 years agofmt: catch overflow in width and prec calculations
Rob Pike [Mon, 4 May 2015 18:28:51 +0000 (11:28 -0700)]
fmt: catch overflow in width and prec calculations

Fixes #10674.

Change-Id: If3fae3244d87aeaa70815f499105c264394aa7ad
Reviewed-on: https://go-review.googlesource.com/9657
Reviewed-by: Ian Lance Taylor <iant@golang.org>
10 years agoruntime: Reduce calls to shouldtriggergc
Rick Hudson [Tue, 28 Apr 2015 13:07:51 +0000 (09:07 -0400)]
runtime: Reduce calls to shouldtriggergc

shouldtriggergc is slightly expensive due to the call overhead
and the use of an atomic. This CL reduces the number of time
one checks if a GC should be done from one at each allocation
to once when a span is allocated. Since shouldtriggergc is an
important abstraction simply hand inlining it, along with its
atomic instruction would lose the abstraction.

Change-Id: Ia3210655b4b3d433f77064a21ecb54e4d9d435f7
Reviewed-on: https://go-review.googlesource.com/9403
Reviewed-by: Austin Clements <austin@google.com>
10 years agotext/template: ensures code consistency in lexer
Aymerick [Wed, 29 Apr 2015 09:33:41 +0000 (11:33 +0200)]
text/template: ensures code consistency in lexer

At the end of lexInsideAction(), we return lexInsideAction: this is the default
behaviour when we are still parsing an action. But some switch branches return
lexInsideAction too.

So let's ensure code consistency by always reaching the end of the
lexInsideAction function when needed.

Change-Id: I7e9d8d6e51f29ecd6db6bdd63b36017845d95368
Reviewed-on: https://go-review.googlesource.com/9441
Reviewed-by: Rob Pike <r@golang.org>
10 years agotime: Fix ordering of slots in AfterQueueing test
Keith Randall [Mon, 4 May 2015 16:17:53 +0000 (09:17 -0700)]
time: Fix ordering of slots in AfterQueueing test

We shouldn't sort the slots array, as it is used each time the
test is run.  Tests after the first should continue to use the
unsorted ordering.

Note that this doesn't fix the flaky test.  Just a bug I saw
while investigating.

Change-Id: Ic03cca637829d569d50d3a2278d19410d4dedba9
Reviewed-on: https://go-review.googlesource.com/9637
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
10 years agoimage/gif: don't encode local color tables if they're the same as the
Nigel Tao [Wed, 29 Apr 2015 06:40:24 +0000 (16:40 +1000)]
image/gif: don't encode local color tables if they're the same as the
global color table.

Change-Id: Ia38f75708ed5e5b430680a1eecafb4fc8047269c
Reviewed-on: https://go-review.googlesource.com/9467
Reviewed-by: Rob Pike <r@golang.org>
10 years agocmd/go: Update alldocs.go header to point to the correct shell script.
David Symonds [Mon, 4 May 2015 00:37:34 +0000 (10:37 +1000)]
cmd/go: Update alldocs.go header to point to the correct shell script.

The script was renamed in b3000b6.

Change-Id: I45ecafff7400e4bff14f31906278609abf2bcb9f
Reviewed-on: https://go-review.googlesource.com/9667
Reviewed-by: Andrew Gerrand <adg@golang.org>
10 years agohash/crc32: move reverse representation docs to an example
Shenghou Ma [Sun, 3 May 2015 21:53:19 +0000 (17:53 -0400)]
hash/crc32: move reverse representation docs to an example

Updates #8229.

Change-Id: I3e691479d3659ed1b3ff8ebbb71b4fc03f2e67af
Reviewed-on: https://go-review.googlesource.com/9680
Reviewed-by: Rob Pike <r@golang.org>
10 years agotest: fix nosplit test for noopt build
Keith Randall [Fri, 1 May 2015 16:36:18 +0000 (09:36 -0700)]
test: fix nosplit test for noopt build

Noopt builds get a larger stack guard.  This test must take that into account.

Change-Id: I1b5cbafdbbfee8c369ae1bebd0b900524ebf0d7d
Reviewed-on: https://go-review.googlesource.com/9610
Reviewed-by: Russ Cox <rsc@golang.org>
10 years agodoc/go1.5.txt: mention openbsd/arm port
Shenghou Ma [Sun, 3 May 2015 05:33:09 +0000 (01:33 -0400)]
doc/go1.5.txt: mention openbsd/arm port

Change-Id: Idc1aacddb79a9270265dd71fa6175539a5bed2c8
Reviewed-on: https://go-review.googlesource.com/9675
Reviewed-by: Minux Ma <minux@golang.org>
10 years agosyscall: fix little-endian assumptions
Shenghou Ma [Sun, 3 May 2015 02:44:49 +0000 (22:44 -0400)]
syscall: fix little-endian assumptions

Change-Id: Ia38256998e544d620a342dabedd9289d61fb0551
Reviewed-on: https://go-review.googlesource.com/9672
Reviewed-by: Dave Cheney <dave@cheney.net>
Run-TryBot: Minux Ma <minux@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

10 years agodoc/go1.5.txt: huge integers are parse errors in text/template
Rob Pike [Sun, 3 May 2015 00:51:58 +0000 (17:51 -0700)]
doc/go1.5.txt: huge integers are parse errors in text/template

Change-Id: Ie9331d102224290833c96a1535cdb96102a7fe9e
Reviewed-on: https://go-review.googlesource.com/9633
Reviewed-by: Rob Pike <r@golang.org>
10 years agotext/template/parse: huge integers are not floats
Rob Pike [Fri, 1 May 2015 22:33:08 +0000 (15:33 -0700)]
text/template/parse: huge integers are not floats

Ideal constants in the template package are a little different from Go.
This is a case that slipped through the cracks: A huge integer number
was accepted as a floating-point number, but this loses precision
and is confusing. Also, the code in the template package (as opposed
to the parse package) wasn't expecting it.

Root this out at the source: If an integer doesn't fit an int64 or uint64,
complain right away.

Change-Id: I375621e6f5333c4d53f053a3c84a9af051711b7a
Reviewed-on: https://go-review.googlesource.com/9651
Reviewed-by: Russ Cox <rsc@golang.org>
10 years agocmd/internal/gc,ld: use new flag syntax
Rob Pike [Fri, 1 May 2015 23:14:50 +0000 (16:14 -0700)]
cmd/internal/gc,ld: use new flag syntax

Followup to CL 9505

Change-Id: I1817b672723bd3d853283f388bc7cbaae2765acc
Reviewed-on: https://go-review.googlesource.com/9652
Reviewed-by: Russ Cox <rsc@golang.org>
10 years agocmd/internal/gc: fix build on big endian systems
Shenghou Ma [Sat, 2 May 2015 08:36:53 +0000 (04:36 -0400)]
cmd/internal/gc: fix build on big endian systems

The siz argument to both runtime.newproc and runtime.deferproc is
int32, not uintptr. This problem won't manifest on little-endian
systems because that stack slot is uintptr sized anyway. However,
on big-endian systems, it will make a difference.

Change-Id: I2351d1ec81839abe25375cff95e327b80764c2b5
Reviewed-on: https://go-review.googlesource.com/9647
Run-TryBot: Minux Ma <minux@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
10 years agotext/template: check for literals in chain of terms
Didier Spezia [Sat, 2 May 2015 11:03:35 +0000 (11:03 +0000)]
text/template: check for literals in chain of terms

The current parser ignores obvious errors such as:
{{0.1.E}}
{{true.any}}
{{"hello".wrong}}
{{nil.E}}

The common problem is that a chain is built from
a literal value. It then panics at execution time.

Furthermore, a double dot triggers the same behavior:
{{..E}}

Addresses a TODO left in Tree.operand to catch these
errors at parsing time.

Note that identifiers can include a '.', and pipelines
could return an object which a field can be derived
from (like a variable), so they are excluded from the check.

Fixes #10615

Change-Id: I903706d1c17861b5a8354632c291e73c9c0bc4e1
Reviewed-on: https://go-review.googlesource.com/9621
Reviewed-by: Rob Pike <r@golang.org>
10 years agogo/ast: fix formatting of error message
Dmitry Vyukov [Sat, 2 May 2015 11:47:27 +0000 (14:47 +0300)]
go/ast: fix formatting of error message

There are three problems:
1. There is no CR at the end of the message.
2. The message is unconditionally printed.
3. The message is printed to stdout.

Change-Id: Ib2d880eea03348e8a69720aad7752302a75bd277
Reviewed-on: https://go-review.googlesource.com/9622
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
10 years agonacltest.bash: remove syscall/fstest_nacl.go after test
Shenghou Ma [Sat, 2 May 2015 01:56:28 +0000 (21:56 -0400)]
nacltest.bash: remove syscall/fstest_nacl.go after test

Fixes #9232.

Change-Id: I11a7fb7691d9e7473620db1b7fa29fec359c73d3
Reviewed-on: https://go-review.googlesource.com/9642
Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com>

10 years agoruntime: fix stackDebug comment
Alex Brainman [Fri, 1 May 2015 05:53:45 +0000 (15:53 +1000)]
runtime: fix stackDebug comment

Change-Id: Ia9191bd7ecdf7bd5ee7d69ae23aa71760f379aa8
Reviewed-on: https://go-review.googlesource.com/9590
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
10 years agocmd/internal/obj: remove Biobuf unget
Dave Cheney [Fri, 1 May 2015 05:29:11 +0000 (15:29 +1000)]
cmd/internal/obj: remove Biobuf unget

This change applies CL 9365 to the copy of Biobuf in cmd/internal/obj.

In the process I discovered that some of the methods that should have been
checking the unget buffer before reading were not and it was probably just
dumb luck that we handn't hit these issues before; Bungetc is only used in
one place in cmd/internal/gc and only an unlikely code path.

Change-Id: Ifa0c5c08442e9fe951a5078c6e9ec77a8a4dc2ff
Reviewed-on: https://go-review.googlesource.com/9529
Reviewed-by: Daniel Morsing <daniel.morsing@gmail.com>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Reviewed-by: Dave Cheney <dave@cheney.net>
Run-TryBot: Dave Cheney <dave@cheney.net>

10 years agonet/http: fix scheduling race resulting in flaky test
Brad Fitzpatrick [Thu, 30 Apr 2015 22:00:51 +0000 (15:00 -0700)]
net/http: fix scheduling race resulting in flaky test

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>

10 years agotext/template: detect unmatched else at parsing time
Didier Spezia [Fri, 1 May 2015 18:20:31 +0000 (18:20 +0000)]
text/template: detect unmatched else at parsing time

An unmatched {{else}} should trigger a parsing error.

The top level parser is able to issue an error in case
of unmatched {{end}}. It does it a posteriori (i.e. after having
parsed the action).

Extend this behavior to also check for unmatched {{else}}

Fixes #10611

Change-Id: I1d4f433cc64e11bea5f4d61419ccc707ac01bb1d
Reviewed-on: https://go-review.googlesource.com/9620
Reviewed-by: Rob Pike <r@golang.org>
10 years agocmd/internal/gc,ld: use new flag argument syntax
Didier Spezia [Thu, 30 Apr 2015 11:56:32 +0000 (11:56 +0000)]
cmd/internal/gc,ld: use new flag argument syntax

The usage messages for the flags in gc and ld are using the old
flag argument syntax:
   "arg: description using arg"

Update them to the Go 1.5 flag package's syntax:
   "description using arg"

Fixes #10505

Change-Id: Ifa54ff91e1fd644cfc9a3b41e10176eac3654137
Reviewed-on: https://go-review.googlesource.com/9505
Reviewed-by: Rob Pike <r@golang.org>
10 years agoruntime: detailed debug output of controller state
Austin Clements [Mon, 6 Apr 2015 18:30:03 +0000 (14:30 -0400)]
runtime: detailed debug output of controller state

This adds a detailed debug dump of the state of the GC controller and
a GODEBUG flag to enable it.

Change-Id: I562fed7981691a84ddf0f9e6fcd9f089f497ac13
Reviewed-on: https://go-review.googlesource.com/9640
Reviewed-by: Russ Cox <rsc@golang.org>
10 years agoruntime: correct accounting of scan work and bytes marked
Russ Cox [Fri, 1 May 2015 15:52:56 +0000 (11:52 -0400)]
runtime: correct accounting of scan work and bytes marked

(1) Count pointer-free objects found during scanning roots
as marked bytes, by not zeroing the mark total after scanning roots.

(2) Don't count the bytes for the roots themselves, by not adding
them to the mark total in scanblock (the zeroing removed by (1)
was aimed at that add but hitting more).

Combined, (1) and (2) fix the calculation of the marked heap size.
This makes the GC trigger much less often in the Go 1 benchmarks,
which have a global []byte pointing at 256 MB of data.
That 256 MB allocation was not being included in the heap size
in the current code, but was included in Go 1.4.
This is the source of much of the relative slowdown in that directory.

(3) Count the bytes for the roots as scanned work, by not zeroing
the scan total after scanning roots. There is no strict justification
for this, and it probably doesn't matter much either way,
but it was always combined with another buggy zeroing
(removed in (1)), so guilty by association.

Austin noticed this.

name                                    old mean                new mean        delta
BenchmarkBinaryTree17              13.1s × (0.97,1.03)      5.9s × (0.97,1.05)  -55.19% (p=0.000)
BenchmarkFannkuch11                4.35s × (0.99,1.01)     4.37s × (1.00,1.01)  +0.47% (p=0.032)
BenchmarkFmtFprintfEmpty          84.6ns × (0.95,1.14)    85.7ns × (0.94,1.05)  ~ (p=0.521)
BenchmarkFmtFprintfString          320ns × (0.95,1.06)     283ns × (0.99,1.02)  -11.48% (p=0.000)
BenchmarkFmtFprintfInt             311ns × (0.98,1.03)     288ns × (0.99,1.02)  -7.26% (p=0.000)
BenchmarkFmtFprintfIntInt          554ns × (0.96,1.05)     478ns × (0.99,1.02)  -13.70% (p=0.000)
BenchmarkFmtFprintfPrefixedInt     434ns × (0.96,1.06)     393ns × (0.98,1.04)  -9.60% (p=0.000)
BenchmarkFmtFprintfFloat           620ns × (0.99,1.03)     584ns × (0.99,1.01)  -5.73% (p=0.000)
BenchmarkFmtManyArgs              2.19µs × (0.98,1.03)    1.94µs × (0.99,1.01)  -11.62% (p=0.000)
BenchmarkGobDecode                21.2ms × (0.97,1.06)    15.2ms × (0.99,1.01)  -28.17% (p=0.000)
BenchmarkGobEncode                18.1ms × (0.94,1.06)    11.8ms × (0.99,1.01)  -35.00% (p=0.000)
BenchmarkGzip                      650ms × (0.98,1.01)     649ms × (0.99,1.02)  ~ (p=0.802)
BenchmarkGunzip                    143ms × (1.00,1.01)     143ms × (1.00,1.01)  ~ (p=0.438)
BenchmarkHTTPClientServer          110µs × (0.98,1.04)     101µs × (0.98,1.02)  -8.79% (p=0.000)
BenchmarkJSONEncode               40.3ms × (0.97,1.03)    31.8ms × (0.98,1.03)  -20.92% (p=0.000)
BenchmarkJSONDecode                119ms × (0.97,1.02)     108ms × (0.99,1.02)  -9.15% (p=0.000)
BenchmarkMandelbrot200            6.03ms × (1.00,1.01)    6.03ms × (0.99,1.01)  ~ (p=0.750)
BenchmarkGoParse                  8.58ms × (0.89,1.10)    6.80ms × (1.00,1.00)  -20.71% (p=0.000)
BenchmarkRegexpMatchEasy0_32       162ns × (1.00,1.01)     162ns × (0.99,1.02)  ~ (p=0.131)
BenchmarkRegexpMatchEasy0_1K       540ns × (0.99,1.02)     559ns × (0.99,1.02)  +3.58% (p=0.000)
BenchmarkRegexpMatchEasy1_32       139ns × (0.98,1.04)     139ns × (1.00,1.00)  ~ (p=0.466)
BenchmarkRegexpMatchEasy1_1K       889ns × (0.99,1.01)     885ns × (0.99,1.01)  -0.50% (p=0.022)
BenchmarkRegexpMatchMedium_32      252ns × (0.99,1.02)     252ns × (0.99,1.01)  ~ (p=0.469)
BenchmarkRegexpMatchMedium_1K     72.9µs × (0.99,1.01)    73.6µs × (0.99,1.03)  ~ (p=0.168)
BenchmarkRegexpMatchHard_32       3.87µs × (1.00,1.01)    3.86µs × (1.00,1.00)  ~ (p=0.055)
BenchmarkRegexpMatchHard_1K        118µs × (0.99,1.01)     117µs × (0.99,1.00)  ~ (p=0.133)
BenchmarkRevcomp                   995ms × (0.94,1.10)     949ms × (0.99,1.01)  -4.64% (p=0.000)
BenchmarkTemplate                  141ms × (0.97,1.02)     127ms × (0.99,1.01)  -10.00% (p=0.000)
BenchmarkTimeParse                 641ns × (0.99,1.01)     623ns × (0.99,1.01)  -2.79% (p=0.000)
BenchmarkTimeFormat                729ns × (0.98,1.03)     679ns × (0.99,1.00)  -6.93% (p=0.000)

Change-Id: I839bd7356630d18377989a0748763414e15ed057
Reviewed-on: https://go-review.googlesource.com/9602
Reviewed-by: Austin Clements <austin@google.com>
10 years agocmd/internal/gc, runtime: use 1-bit bitmap for stack frames, data, bss
Russ Cox [Tue, 28 Apr 2015 02:45:57 +0000 (22:45 -0400)]
cmd/internal/gc, runtime: use 1-bit bitmap for stack frames, data, bss

The bitmaps were 2 bits per pointer because we needed to distinguish
scalar, pointer, multiword, and we used the leftover value to distinguish
uninitialized from scalar, even though the garbage collector (GC) didn't care.

Now that there are no multiword structures from the GC's point of view,
cut the bitmaps down to 1 bit per pointer, recording just live pointer vs not.

The GC assumes the same layout for stack frames and for the maps
describing the global data and bss sections, so change them all in one CL.

The code still refers to 4-bit heap bitmaps and 2-bit "type bitmaps", since
the 2-bit representation lives (at least for now) in some of the reflect data.

Because these stack frame bitmaps are stored directly in the rodata in
the binary, this CL reduces the size of the 6g binary by about 1.1%.

Performance change is basically a wash, but using less memory,
and smaller binaries, and enables other bitmap reductions.

name                                      old mean                new mean        delta
BenchmarkBinaryTree17                13.2s × (0.97,1.03)     13.0s × (0.99,1.01)  -0.93% (p=0.005)
BenchmarkBinaryTree17-2              9.69s × (0.96,1.05)     9.51s × (0.96,1.03)  -1.86% (p=0.001)
BenchmarkBinaryTree17-4              10.1s × (0.97,1.05)     10.0s × (0.96,1.05)  ~ (p=0.141)
BenchmarkFannkuch11                  4.35s × (0.99,1.01)     4.43s × (0.98,1.04)  +1.75% (p=0.001)
BenchmarkFannkuch11-2                4.31s × (0.99,1.03)     4.32s × (1.00,1.00)  ~ (p=0.095)
BenchmarkFannkuch11-4                4.32s × (0.99,1.02)     4.38s × (0.98,1.04)  +1.38% (p=0.008)
BenchmarkFmtFprintfEmpty            83.5ns × (0.97,1.10)    87.3ns × (0.92,1.11)  +4.55% (p=0.014)
BenchmarkFmtFprintfEmpty-2          81.8ns × (0.98,1.04)    82.5ns × (0.97,1.08)  ~ (p=0.364)
BenchmarkFmtFprintfEmpty-4          80.9ns × (0.99,1.01)    82.6ns × (0.97,1.08)  +2.12% (p=0.010)
BenchmarkFmtFprintfString            320ns × (0.95,1.04)     322ns × (0.97,1.05)  ~ (p=0.368)
BenchmarkFmtFprintfString-2          303ns × (0.97,1.04)     304ns × (0.97,1.04)  ~ (p=0.484)
BenchmarkFmtFprintfString-4          305ns × (0.97,1.05)     306ns × (0.98,1.05)  ~ (p=0.543)
BenchmarkFmtFprintfInt               311ns × (0.98,1.03)     319ns × (0.97,1.03)  +2.63% (p=0.000)
BenchmarkFmtFprintfInt-2             297ns × (0.98,1.04)     301ns × (0.97,1.04)  +1.19% (p=0.023)
BenchmarkFmtFprintfInt-4             302ns × (0.98,1.02)     304ns × (0.97,1.03)  ~ (p=0.126)
BenchmarkFmtFprintfIntInt            554ns × (0.96,1.05)     554ns × (0.97,1.03)  ~ (p=0.975)
BenchmarkFmtFprintfIntInt-2          520ns × (0.98,1.03)     517ns × (0.98,1.02)  ~ (p=0.153)
BenchmarkFmtFprintfIntInt-4          524ns × (0.98,1.02)     525ns × (0.98,1.03)  ~ (p=0.597)
BenchmarkFmtFprintfPrefixedInt       433ns × (0.97,1.06)     434ns × (0.97,1.06)  ~ (p=0.804)
BenchmarkFmtFprintfPrefixedInt-2     413ns × (0.98,1.04)     413ns × (0.98,1.03)  ~ (p=0.881)
BenchmarkFmtFprintfPrefixedInt-4     420ns × (0.97,1.03)     421ns × (0.97,1.03)  ~ (p=0.561)
BenchmarkFmtFprintfFloat             620ns × (0.99,1.03)     636ns × (0.97,1.03)  +2.57% (p=0.000)
BenchmarkFmtFprintfFloat-2           601ns × (0.98,1.02)     617ns × (0.98,1.03)  +2.58% (p=0.000)
BenchmarkFmtFprintfFloat-4           613ns × (0.98,1.03)     626ns × (0.98,1.02)  +2.15% (p=0.000)
BenchmarkFmtManyArgs                2.19µs × (0.96,1.04)    2.23µs × (0.97,1.02)  +1.65% (p=0.000)
BenchmarkFmtManyArgs-2              2.08µs × (0.98,1.03)    2.10µs × (0.99,1.02)  +0.79% (p=0.019)
BenchmarkFmtManyArgs-4              2.10µs × (0.98,1.02)    2.13µs × (0.98,1.02)  +1.72% (p=0.000)
BenchmarkGobDecode                  21.3ms × (0.97,1.05)    21.1ms × (0.97,1.04)  -1.36% (p=0.025)
BenchmarkGobDecode-2                20.0ms × (0.97,1.03)    19.2ms × (0.97,1.03)  -4.00% (p=0.000)
BenchmarkGobDecode-4                19.5ms × (0.99,1.02)    19.0ms × (0.99,1.01)  -2.39% (p=0.000)
BenchmarkGobEncode                  18.3ms × (0.95,1.07)    18.1ms × (0.96,1.08)  ~ (p=0.305)
BenchmarkGobEncode-2                16.8ms × (0.97,1.02)    16.4ms × (0.98,1.02)  -2.79% (p=0.000)
BenchmarkGobEncode-4                15.4ms × (0.98,1.02)    15.4ms × (0.98,1.02)  ~ (p=0.465)
BenchmarkGzip                        650ms × (0.98,1.03)     655ms × (0.97,1.04)  ~ (p=0.075)
BenchmarkGzip-2                      652ms × (0.98,1.03)     655ms × (0.98,1.02)  ~ (p=0.337)
BenchmarkGzip-4                      656ms × (0.98,1.04)     653ms × (0.98,1.03)  ~ (p=0.291)
BenchmarkGunzip                      143ms × (1.00,1.01)     143ms × (1.00,1.01)  ~ (p=0.507)
BenchmarkGunzip-2                    143ms × (1.00,1.01)     143ms × (1.00,1.01)  ~ (p=0.313)
BenchmarkGunzip-4                    143ms × (1.00,1.01)     143ms × (1.00,1.01)  ~ (p=0.312)
BenchmarkHTTPClientServer            110µs × (0.98,1.03)     109µs × (0.99,1.02)  -1.40% (p=0.000)
BenchmarkHTTPClientServer-2          154µs × (0.90,1.08)     149µs × (0.90,1.08)  -3.43% (p=0.007)
BenchmarkHTTPClientServer-4          138µs × (0.97,1.04)     138µs × (0.96,1.04)  ~ (p=0.670)
BenchmarkJSONEncode                 40.2ms × (0.98,1.02)    40.2ms × (0.98,1.05)  ~ (p=0.828)
BenchmarkJSONEncode-2               35.1ms × (0.99,1.02)    35.2ms × (0.98,1.03)  ~ (p=0.392)
BenchmarkJSONEncode-4               35.3ms × (0.98,1.03)    35.3ms × (0.98,1.02)  ~ (p=0.813)
BenchmarkJSONDecode                  119ms × (0.97,1.02)     117ms × (0.98,1.02)  -1.80% (p=0.000)
BenchmarkJSONDecode-2                115ms × (0.99,1.02)     114ms × (0.98,1.02)  -1.18% (p=0.000)
BenchmarkJSONDecode-4                116ms × (0.98,1.02)     114ms × (0.98,1.02)  -1.43% (p=0.000)
BenchmarkMandelbrot200              6.03ms × (1.00,1.01)    6.03ms × (1.00,1.01)  ~ (p=0.985)
BenchmarkMandelbrot200-2            6.03ms × (1.00,1.01)    6.02ms × (1.00,1.01)  ~ (p=0.320)
BenchmarkMandelbrot200-4            6.03ms × (1.00,1.01)    6.03ms × (1.00,1.01)  ~ (p=0.799)
BenchmarkGoParse                    8.63ms × (0.89,1.10)    8.58ms × (0.93,1.09)  ~ (p=0.667)
BenchmarkGoParse-2                  8.20ms × (0.97,1.04)    8.37ms × (0.97,1.04)  +1.96% (p=0.001)
BenchmarkGoParse-4                  8.00ms × (0.98,1.02)    8.14ms × (0.99,1.02)  +1.75% (p=0.000)
BenchmarkRegexpMatchEasy0_32         162ns × (1.00,1.01)     164ns × (0.98,1.04)  +1.35% (p=0.011)
BenchmarkRegexpMatchEasy0_32-2       161ns × (1.00,1.01)     161ns × (1.00,1.00)  ~ (p=0.185)
BenchmarkRegexpMatchEasy0_32-4       161ns × (1.00,1.00)     161ns × (1.00,1.00)  -0.19% (p=0.001)
BenchmarkRegexpMatchEasy0_1K         540ns × (0.99,1.02)     566ns × (0.98,1.04)  +4.98% (p=0.000)
BenchmarkRegexpMatchEasy0_1K-2       540ns × (0.99,1.01)     557ns × (0.99,1.01)  +3.21% (p=0.000)
BenchmarkRegexpMatchEasy0_1K-4       541ns × (0.99,1.01)     559ns × (0.99,1.01)  +3.26% (p=0.000)
BenchmarkRegexpMatchEasy1_32         139ns × (0.98,1.04)     139ns × (0.99,1.03)  ~ (p=0.979)
BenchmarkRegexpMatchEasy1_32-2       139ns × (0.99,1.04)     139ns × (0.99,1.02)  ~ (p=0.777)
BenchmarkRegexpMatchEasy1_32-4       139ns × (0.98,1.04)     139ns × (0.99,1.04)  ~ (p=0.771)
BenchmarkRegexpMatchEasy1_1K         890ns × (0.99,1.03)     885ns × (1.00,1.01)  -0.50% (p=0.004)
BenchmarkRegexpMatchEasy1_1K-2       888ns × (0.99,1.01)     885ns × (0.99,1.01)  -0.37% (p=0.004)
BenchmarkRegexpMatchEasy1_1K-4       890ns × (0.99,1.02)     884ns × (1.00,1.00)  -0.70% (p=0.000)
BenchmarkRegexpMatchMedium_32        252ns × (0.99,1.01)     251ns × (0.99,1.01)  ~ (p=0.081)
BenchmarkRegexpMatchMedium_32-2      254ns × (0.99,1.04)     252ns × (0.99,1.01)  -0.78% (p=0.027)
BenchmarkRegexpMatchMedium_32-4      253ns × (0.99,1.04)     252ns × (0.99,1.01)  -0.70% (p=0.022)
BenchmarkRegexpMatchMedium_1K       72.9µs × (0.99,1.01)    72.7µs × (1.00,1.00)  ~ (p=0.064)
BenchmarkRegexpMatchMedium_1K-2     74.1µs × (0.98,1.05)    72.9µs × (1.00,1.01)  -1.61% (p=0.001)
BenchmarkRegexpMatchMedium_1K-4     73.6µs × (0.99,1.05)    72.8µs × (1.00,1.00)  -1.13% (p=0.007)
BenchmarkRegexpMatchHard_32         3.88µs × (0.99,1.03)    3.92µs × (0.98,1.05)  ~ (p=0.143)
BenchmarkRegexpMatchHard_32-2       3.89µs × (0.99,1.03)    3.93µs × (0.98,1.09)  ~ (p=0.278)
BenchmarkRegexpMatchHard_32-4       3.90µs × (0.99,1.05)    3.93µs × (0.98,1.05)  ~ (p=0.252)
BenchmarkRegexpMatchHard_1K          118µs × (0.99,1.01)     117µs × (0.99,1.02)  -0.54% (p=0.003)
BenchmarkRegexpMatchHard_1K-2        118µs × (0.99,1.01)     118µs × (0.99,1.03)  ~ (p=0.581)
BenchmarkRegexpMatchHard_1K-4        118µs × (0.99,1.02)     117µs × (0.99,1.01)  -0.54% (p=0.002)
BenchmarkRevcomp                     991ms × (0.95,1.10)     989ms × (0.94,1.08)  ~ (p=0.879)
BenchmarkRevcomp-2                   978ms × (0.95,1.11)     962ms × (0.96,1.08)  ~ (p=0.257)
BenchmarkRevcomp-4                   979ms × (0.96,1.07)     974ms × (0.96,1.11)  ~ (p=0.678)
BenchmarkTemplate                    141ms × (0.99,1.02)     145ms × (0.99,1.02)  +2.75% (p=0.000)
BenchmarkTemplate-2                  135ms × (0.98,1.02)     138ms × (0.99,1.02)  +2.34% (p=0.000)
BenchmarkTemplate-4                  136ms × (0.98,1.02)     140ms × (0.99,1.02)  +2.71% (p=0.000)
BenchmarkTimeParse                   640ns × (0.99,1.01)     622ns × (0.99,1.01)  -2.88% (p=0.000)
BenchmarkTimeParse-2                 640ns × (0.99,1.01)     622ns × (1.00,1.00)  -2.81% (p=0.000)
BenchmarkTimeParse-4                 640ns × (1.00,1.01)     622ns × (0.99,1.01)  -2.82% (p=0.000)
BenchmarkTimeFormat                  730ns × (0.98,1.02)     731ns × (0.98,1.03)  ~ (p=0.767)
BenchmarkTimeFormat-2                709ns × (0.99,1.02)     707ns × (0.99,1.02)  ~ (p=0.347)
BenchmarkTimeFormat-4                717ns × (0.98,1.01)     718ns × (0.98,1.02)  ~ (p=0.793)

Change-Id: Ie779c47e912bf80eb918bafa13638bd8dfd6c2d9
Reviewed-on: https://go-review.googlesource.com/9406
Reviewed-by: Rick Hudson <rlh@golang.org>
10 years agocmd/internal/obj: clean up Biobuf
Dave Cheney [Fri, 1 May 2015 01:51:47 +0000 (11:51 +1000)]
cmd/internal/obj: clean up Biobuf

This is a follow up to rev 443a32e707d2 which reduces some of the
duplication between methods and functions that operate on obj.Biobuf.

obj.Biobuf has Flush and Write methods as well as helpers which duplicate
those methods, consolidate on the former and remove the latter.

Also, address a final comment from CL 9525.

Change-Id: I67deaf3a163bb489a9bb21bb39524785d7a2f6c5
Reviewed-on: https://go-review.googlesource.com/9527
Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
Reviewed-by: Ian Lance Taylor <iant@golang.org>
10 years agocmd/internal/gc: Toughen escape analysis against some bugs.
David Chase [Fri, 1 May 2015 15:16:35 +0000 (11:16 -0400)]
cmd/internal/gc: Toughen escape analysis against some bugs.

Ensures that parameter flow bits are not set for tags EscScope, EscHeap, EscNever;
crash the compiler earl to expose faulty logic, rather than flake out silently downstream.

Change-Id: I1428129980ae047d02975f033d56cbbd04f49579
Reviewed-on: https://go-review.googlesource.com/9601
Reviewed-by: Russ Cox <rsc@golang.org>
10 years agoRevert "runtime/pprof: write heap statistics to heap profile always"
Josh Bleecher Snyder [Fri, 1 May 2015 15:10:24 +0000 (15:10 +0000)]
Revert "runtime/pprof: write heap statistics to heap profile always"

This reverts commit c26fc88d56ee4f93c98fc8923fe256121e6199cf.

This broke pprof. See the comments at 9491.

Change-Id: Ic99ce026e86040c050a9bf0ea3024a1a42274ad1
Reviewed-on: https://go-review.googlesource.com/9565
Reviewed-by: Daniel Morsing <daniel.morsing@gmail.com>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
10 years agocmd/dist, runtime: Make stack guard larger for non-optimized builds
Keith Randall [Thu, 30 Apr 2015 23:57:23 +0000 (16:57 -0700)]
cmd/dist, runtime: Make stack guard larger for non-optimized builds

Kind of a hack, but makes the non-optimized builds pass.

Fixes #10079

Change-Id: I26f41c546867f8f3f16d953dc043e784768f2aff
Reviewed-on: https://go-review.googlesource.com/9552
Reviewed-by: Russ Cox <rsc@golang.org>
10 years agocmd/internal/gc: improve flow of input params to output params
David Chase [Thu, 26 Mar 2015 20:36:15 +0000 (16:36 -0400)]
cmd/internal/gc: improve flow of input params to output params

This includes the following information in the per-function summary:

outK = paramJ   encoded in outK bits for paramJ
outK = *paramJ  encoded in outK bits for paramJ
heap = paramJ   EscHeap
heap = *paramJ  EscContentEscapes

Note that (currently) if the address of a parameter is taken and
returned, necessarily a heap allocation occurred to contain that
reference, and the heap can never refer to stack, therefore the
parameter and everything downstream from it escapes to the heap.

The per-function summary information now has a tuneable number of bits
(2 is probably noticeably better than 1, 3 is likely overkill, but it
is now easy to check and the -m debugging output includes information
that allows you to figure out if more would be better.)

A new test was  added to check pointer flow through struct-typed and
*struct-typed parameters and returns; some of these are sensitive to
the number of summary bits, and ought to yield better results with a
more competent escape analysis algorithm.  Another new test checks
(some) correctness with array parameters, results, and operations.

The old analysis inferred a piece of plan9 runtime was non-escaping by
counteracting overconservative analysis with buggy analysis; with the
bug fixed, the result was too conservative (and it's not easy to fix
in this framework) so the source code was tweaked to get the desired
result.  A test was added against the discovered bug.

The escape analysis was further improved splitting the "level" into
3 parts, one tracking the conventional "level" and the other two
computing the highest-level-suffix-from-copy, which is used to
generally model the cancelling effect of indirection applied to
address-of.

With the improved escape analysis enabled, it was necessary to
modify one of the runtime tests because it now attempts to allocate
too much on the (small, fixed-size) G0 (system) stack and this
failed the test.

Compiling src/std after touching src/runtime/*.go with -m logging
turned on shows 420 fewer heap allocation sites (10538 vs 10968).

Profiling allocations in src/html/template with
for i in {1..5} ;
  do go tool 6g -memprofile=mastx.${i}.prof  -memprofilerate=1 *.go;
  go tool pprof -alloc_objects -text  mastx.${i}.prof ;
done

showed a 15% reduction in allocations performed by the compiler.

Update #3753
Update #4720
Fixes #10466

Change-Id: I0fd97d5f5ac527b45f49e2218d158a6e89951432
Reviewed-on: https://go-review.googlesource.com/8202
Run-TryBot: David Chase <drchase@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Russ Cox <rsc@golang.org>
10 years agoruntime/cgo, cmd/dist: turn off exc_bad_access handler by default
David Crawshaw [Thu, 30 Apr 2015 20:38:10 +0000 (16:38 -0400)]
runtime/cgo, cmd/dist: turn off exc_bad_access handler by default

App Store policy requires programs do not reference the exc_server
symbol. (Some public forum threads show that Unity ran into this
several years ago and it is a hard policy rule.) While some research
suggests that I could write my own version of exc_server, the
expedient course is to disable the exception handler by default.

Go programs only need it when running under lldb, which is primarily
used by tests. So enable the exception handler in cmd/dist when we
are running the tests.

Fixes #10646

Change-Id: I853905254894b5367edb8abd381d45585a78ee8b
Reviewed-on: https://go-review.googlesource.com/9549
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
10 years agoruntime: adjust traceTickDiv for non-x86 architectures
Shenghou Ma [Thu, 23 Apr 2015 06:43:10 +0000 (02:43 -0400)]
runtime: adjust traceTickDiv for non-x86 architectures

Fixes #10554.
Fixes #10623.

Change-Id: I90fbaa34e3d55c8758178f8d2e7fa41ff1194a1b
Signed-off-by: Shenghou Ma <minux@golang.org>
Reviewed-on: https://go-review.googlesource.com/9247
Reviewed-by: Dmitry Vyukov <dvyukov@google.com>
Reviewed-by: Dave Cheney <dave@cheney.net>
10 years agocmd/cover: fix build
Dave Cheney [Fri, 1 May 2015 02:38:11 +0000 (12:38 +1000)]
cmd/cover: fix build

Fix the various builds which don't have a real filesystem or don't support forking.

Change-Id: I3075c662fe6191ecbe70ba359b73d9a88bb06f35
Reviewed-on: https://go-review.googlesource.com/9528
Reviewed-by: Rob Pike <r@golang.org>
Reviewed-by: Dave Cheney <dave@cheney.net>
10 years agocmd/cover: try once again to fix the build
Rob Pike [Fri, 1 May 2015 02:14:45 +0000 (19:14 -0700)]
cmd/cover: try once again to fix the build

Forgot to update the references to the old cover package. No excuse.

Change-Id: If17b7521f0bf70bc0c8da9c5adf246d90f644637
Reviewed-on: https://go-review.googlesource.com/9564
Reviewed-by: Rob Pike <r@golang.org>
10 years agocmd/cover: fix build
Rob Pike [Fri, 1 May 2015 02:10:59 +0000 (19:10 -0700)]
cmd/cover: fix build

TBR=rsc

Change-Id: I6ec69013027213c5e7adedd2edb89dea6af876d9
Reviewed-on: https://go-review.googlesource.com/9563
Reviewed-by: Rob Pike <r@golang.org>
10 years agocmd/8g: don't call gc.Fatal during initalisation
Dave Cheney [Fri, 1 May 2015 00:15:33 +0000 (10:15 +1000)]
cmd/8g: don't call gc.Fatal during initalisation

Fixes #10592

Calling gc.Fatal before gc.Main has been called ends up flushing gc.bstdout before
it is properly set up. Ideally obj.Bflush would handle this case, but that type
and its callers are rather convoluted, so take the simpler route and avoid calling
gc.Fatal altogether.

Change-Id: I338b469e86edba558b6bedff35bb904bfc3d6990
Reviewed-on: https://go-review.googlesource.com/9525
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Reviewed-by: Dave Cheney <dave@cheney.net>
Run-TryBot: Dave Cheney <dave@cheney.net>
TryBot-Result: Gobot Gobot <gobot@golang.org>

10 years agotext/template: allow newlines in raw quotes
Rob Pike [Thu, 30 Apr 2015 19:11:35 +0000 (12:11 -0700)]
text/template: allow newlines in raw quotes

This was disallowed for error-checking reasons but people ask for
it, it's easy, and it's clear what it all means.

Fixes #7323.

Change-Id: I26542f5ac6519e45b335ad789713a4d9e356279b
Reviewed-on: https://go-review.googlesource.com/9537
Reviewed-by: Russ Cox <rsc@golang.org>
10 years agodoc/go1.5.txt: cover has moved
Rob Pike [Fri, 1 May 2015 00:42:16 +0000 (17:42 -0700)]
doc/go1.5.txt: cover has moved

Change-Id: Ie4b59d72e2b704559e075494e79fdc7b0bca6556
Reviewed-on: https://go-review.googlesource.com/9562
Reviewed-by: Rob Pike <r@golang.org>
10 years agocmd/cover: copy to standard repository from golang.org/x/tools/cmd/cover
Rob Pike [Thu, 30 Apr 2015 21:37:43 +0000 (14:37 -0700)]
cmd/cover: copy to standard repository from golang.org/x/tools/cmd/cover

This required dealing with the ill-advised split of the profile code
into a separate package. I just copied it over unchanged. The package
does not deserve to be in the standard repository. We can cope
with the duplication.

Also update the go command to know about the new location.

Fixes #10528.

Change-Id: I05170ef3663326d57b9c18888d01163acd9256b6
Reviewed-on: https://go-review.googlesource.com/9560
Reviewed-by: Russ Cox <rsc@golang.org>
10 years agodoc/progs: remove flaky timing-sensitive test
Matthew Dempsky [Thu, 30 Apr 2015 10:01:22 +0000 (03:01 -0700)]
doc/progs: remove flaky timing-sensitive test

Package time already has enough inherently flaky tests covering its
behavior.  No need for more of them.

Fixes #10632.

Change-Id: I1229e9fcc2e28ba2c9b0b79f73638e35dbbe8bbf
Reviewed-on: https://go-review.googlesource.com/9517
Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
10 years agocmd/go: support -buildmode=c-shared for gccgo
Ian Lance Taylor [Thu, 30 Apr 2015 19:08:46 +0000 (12:08 -0700)]
cmd/go: support -buildmode=c-shared for gccgo

Change-Id: I4cdfd5a59e0468e9e5400aa06334b21cc80913cd
Reviewed-on: https://go-review.googlesource.com/9550
Reviewed-by: David Crawshaw <crawshaw@golang.org>
10 years agonet/http: document ServeFile and FileServer index.html redirect behavior
Brad Fitzpatrick [Thu, 30 Apr 2015 20:24:58 +0000 (13:24 -0700)]
net/http: document ServeFile and FileServer index.html redirect behavior

Fixes #9876

Change-Id: I97a354fde827dfccc9e373fadea2e37d094439b0
Reviewed-on: https://go-review.googlesource.com/9538
Reviewed-by: Rob Pike <r@golang.org>