]> Cypherpunks repositories - gostls13.git/log
gostls13.git
9 years agocmd/go: fix processing of flags for test binaries.
Rob Pike [Tue, 22 Sep 2015 21:23:32 +0000 (14:23 -0700)]
cmd/go: fix processing of flags for test binaries.

The usage message says:

test [-c] [-i] [build and test flags] [packages] [flags for test binary]

but this was not what was implemented. Instead, after packages are named,
flag processing continues, which makes it impossible, for example, to pass
to the binary a flag with the same name as a test flag. This was triggered
by the -v flag in glog.

Consider this test:

package pkg

... imports ...

var v = flag.Int("v", 0, "v flag")

func TestFoo(t *testing.T) {
if *v != 7 { log.Fatal(*v) }
}

Attempting to run this test with go test pkg -v=7 would give a usage
message. This change allows it. In fact it allows

go test -v pkg -v=7

The solution is to implement the usage message. One compatibility
issue is that flags after the package name are no longer processed
as test flags, so this no longer works:

go test foo -cover

One must write

go test -cover foo

I do not think this is onerous but it must be called out in the
release notes.

Fixes #12177.

Change-Id: Ib9267884b47a6b0c183efa888ec78333272113aa
Reviewed-on: https://go-review.googlesource.com/14826
Reviewed-by: Ian Lance Taylor <iant@golang.org>
9 years agostrconv: slightly simplified roundShortest; better comments
Robert Griesemer [Wed, 23 Sep 2015 16:52:21 +0000 (09:52 -0700)]
strconv: slightly simplified roundShortest; better comments

Change-Id: If886f15468680f7e1c589873066b4391eb9784b5
Reviewed-on: https://go-review.googlesource.com/14856
Reviewed-by: Alan Donovan <adonovan@google.com>
9 years agomath/big: factored out an internal accessor method (cleanup), added benchmark
Robert Griesemer [Wed, 23 Sep 2015 17:05:44 +0000 (10:05 -0700)]
math/big: factored out an internal accessor method (cleanup), added benchmark

Current result of DecimalConversion benchmark (for future reference):

BenchmarkDecimalConversion-8    10000     204770 ns/op

Measured on Mac Mini (late 2012) running OS X 10.10.5,
2.3 GHz Intel Core i7, 8 GB 1333 MHz DDR3.

Also: Removed comment suggesting to implement decimal by representing
digits as numbers 0..9 rather than ASCII chars '0'..'9' to avoid
repeated +/-'0' operations. Tried and it appears (per above benchmark)
that the +/-'0' operations are neglibile but the addition conversion
passes around it are not and that it makes things significantly slower.

Change-Id: I6ee033b1172043248093cc5d02abff5fc54c2e7a
Reviewed-on: https://go-review.googlesource.com/14857
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Run-TryBot: Robert Griesemer <gri@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Alan Donovan <adonovan@google.com>
9 years agoos: document that behavior of Seek on O_APPEND files is not specified
Shenghou Ma [Wed, 23 Sep 2015 15:49:44 +0000 (11:49 -0400)]
os: document that behavior of Seek on O_APPEND files is not specified

The actual behavior varies across platforms, and due to the inherent
race, we can't do anything better (other than to always return 0).

Fixes #12710.

Change-Id: Icb52f0f1f0a267e0f9f70767cae427f3f0239965
Reviewed-on: https://go-review.googlesource.com/14881
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
9 years agoruntime: fix incomplete sentence in comment
Shenghou Ma [Tue, 22 Sep 2015 06:18:57 +0000 (02:18 -0400)]
runtime: fix incomplete sentence in comment

Fixes #12709.

Change-Id: If5a2536458fcd26d6f003dde1bfc02f86b09fa94
Reviewed-on: https://go-review.googlesource.com/14793
Reviewed-by: Andrew Gerrand <adg@golang.org>
9 years agomath/big: implement negative precision for Float.Append/Text
Robert Griesemer [Wed, 23 Sep 2015 00:32:54 +0000 (17:32 -0700)]
math/big: implement negative precision for Float.Append/Text

Enabled all but a handful of disabled Float formatting test cases.

Fixes #10991.

Change-Id: Id18e160e857be2743429a377000e996978015a1a
Reviewed-on: https://go-review.googlesource.com/14850
Reviewed-by: Alan Donovan <adonovan@google.com>
9 years agoruntime: test that timeBeginPeriod succeeds
Alex Brainman [Thu, 23 Jul 2015 01:42:51 +0000 (11:42 +1000)]
runtime: test that timeBeginPeriod succeeds

Change-Id: I5183f767dadb6d24a34d2460d02e97ddbaab129a
Reviewed-on: https://go-review.googlesource.com/12546
Run-TryBot: Minux Ma <minux@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
9 years agocompress/gzip: detect truncated streams
Joe Tsai [Tue, 22 Sep 2015 09:14:28 +0000 (02:14 -0700)]
compress/gzip: detect truncated streams

Reader fails to detect truncated streams since calls to io.ReadFull
do not check if the error is io.EOF.

Change-Id: I052cd03161e43fec17e3d328106c40e17923e52b
Reviewed-on: https://go-review.googlesource.com/14832
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agocompress/flate: detect truncated streams
Joe Tsai [Tue, 22 Sep 2015 08:28:09 +0000 (01:28 -0700)]
compress/flate: detect truncated streams

Reader failed to detect truncated streams since calls to
io.ReadFull did not check if the error is io.EOF.

Change-Id: I0634e0d8de1ab04e8f93242c27a9f89e57743e87
Reviewed-on: https://go-review.googlesource.com/14833
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agoos: add check for ERROR_BAD_NETPATH in windows IsNotExist
Alex Brainman [Tue, 15 Sep 2015 06:59:20 +0000 (16:59 +1000)]
os: add check for ERROR_BAD_NETPATH in windows IsNotExist

Otherwise IsNotExist does not account for not existent servers and shares.

Fixes #12374

Change-Id: I37f6850198f91dcb02a4a917b793339d7e30e934
Reviewed-on: https://go-review.googlesource.com/14579
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
9 years agodoc: document go1.4.3
Chris Broadfoot [Wed, 23 Sep 2015 03:45:16 +0000 (20:45 -0700)]
doc: document go1.4.3

Change-Id: Ib1bfe4038e2b125a31acd9ff7772e462b0a6358f
Reviewed-on: https://go-review.googlesource.com/14852
Reviewed-by: Andrew Gerrand <adg@golang.org>
9 years agoregexp: add runnable example to regex.Split
Seth Hoenig [Fri, 18 Sep 2015 04:26:39 +0000 (23:26 -0500)]
regexp: add runnable example to regex.Split

The existing comment for regex.Split contains a plain text example,
while many of the other regex functions have runnable examples. This
change provides a runnable example for Split.

Change-Id: I5373f57f532fe843d7d0adcf4b513061ec797047
Reviewed-on: https://go-review.googlesource.com/14737
Reviewed-by: Andrew Gerrand <adg@golang.org>
Run-TryBot: Andrew Gerrand <adg@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agoencoding/json: spell "marshaling" and "unmarshaling" consistently
Andrew Gerrand [Tue, 1 Sep 2015 07:51:39 +0000 (17:51 +1000)]
encoding/json: spell "marshaling" and "unmarshaling" consistently

Fixes #12431

Change-Id: I67c42bf2cd9285f471387248fd9c22a16b158349
Reviewed-on: https://go-review.googlesource.com/14150
Reviewed-by: Dmitri Shuralyov <shurcool@gmail.com>
Reviewed-by: Rob Pike <r@golang.org>
9 years agotest/fixedbugs: update overly restrictive test case
Robert Griesemer [Tue, 22 Sep 2015 16:37:56 +0000 (09:37 -0700)]
test/fixedbugs: update overly restrictive test case

See discussion in https://go-review.googlesource.com/14830 .

Change-Id: I94f25f92b8cdaa509d2c335865a645228425804d
Reviewed-on: https://go-review.googlesource.com/14837
Reviewed-by: Ian Lance Taylor <iant@golang.org>
9 years agoruntime: fix offset in invalidptr panic message
Austin Clements [Mon, 21 Sep 2015 15:49:52 +0000 (11:49 -0400)]
runtime: fix offset in invalidptr panic message

Change-Id: I00e1eebbf5e1a01c8fad5ca5324aa8eec1e4d731
Reviewed-on: https://go-review.googlesource.com/14792
Reviewed-by: Rick Hudson <rlh@golang.org>
9 years agoruntime: optimize duffcopy on amd64
Ilya Tocar [Tue, 22 Sep 2015 11:32:05 +0000 (14:32 +0300)]
runtime: optimize duffcopy on amd64

Use movups to copy 16 bytes at a time.
Results (haswell):

name            old time/op  new time/op  delta
CopyFat8-48     0.62ns ± 3%  0.63ns ± 3%     ~     (p=0.535 n=20+20)
CopyFat12-48    0.92ns ± 2%  0.93ns ± 3%     ~     (p=0.594 n=17+18)
CopyFat16-48    1.23ns ± 2%  1.23ns ± 2%     ~     (p=0.839 n=20+19)
CopyFat24-48    1.85ns ± 2%  1.84ns ± 0%   -0.48%  (p=0.014 n=19+20)
CopyFat32-48    2.45ns ± 0%  2.45ns ± 1%     ~     (p=1.000 n=16+16)
CopyFat64-48    3.30ns ± 2%  2.14ns ± 1%  -35.00%  (p=0.000 n=20+18)
CopyFat128-48   6.05ns ± 0%  3.98ns ± 0%  -34.22%  (p=0.000 n=18+17)
CopyFat256-48   11.9ns ± 3%   7.7ns ± 0%  -35.87%  (p=0.000 n=20+17)
CopyFat512-48   23.0ns ± 2%  15.1ns ± 2%  -34.52%  (p=0.000 n=20+18)
CopyFat1024-48  44.8ns ± 1%  29.8ns ± 2%  -33.48%  (p=0.000 n=17+19)

Change-Id: I8a78773c656d400726a020894461e00c59f896bf
Reviewed-on: https://go-review.googlesource.com/14836
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Keith Randall <khr@golang.org>
9 years agomath/big: add test cases for min/max exponent values
Robert Griesemer [Mon, 21 Sep 2015 22:25:52 +0000 (15:25 -0700)]
math/big: add test cases for min/max exponent values

Change-Id: I2e74e39628285e2fecaab712be6cff230619a6c2
Reviewed-on: https://go-review.googlesource.com/14778
Reviewed-by: Alan Donovan <adonovan@google.com>
9 years agomath/big: optimize Float.Parse by reducing powers of 10 to powers of 2 and 5
Robert Griesemer [Sun, 20 Sep 2015 01:24:16 +0000 (18:24 -0700)]
math/big: optimize Float.Parse by reducing powers of 10 to powers of 2 and 5

Instead of computing the final adjustment factor as a power of 10,
it's more efficient to split 10**e into 2**e * 5**e . Powers of 2
are trivially added to the Float exponent, and powers of 5 are
smaller and thus faster to compute.

Also, use a table of uint64 values rather than float64 values for
initial power value. uint64 values appear to be faster to convert
to Floats (useful for small exponents).

Added two small benchmarks to confirm that there's no regresssion.

benchmark                         old ns/op     new ns/op     delta
BenchmarkParseFloatSmallExp-8     17543         16220         -7.54%
BenchmarkParseFloatLargeExp-8     60865         59996         -1.43%

Change-Id: I3efd7556b023316f86f334137a67fe0c6d52f8ef
Reviewed-on: https://go-review.googlesource.com/14782
Reviewed-by: Alan Donovan <adonovan@google.com>
9 years agosrc/cmd/compile/internal/gc: remove now unnecessary restriction on float exponent...
Robert Griesemer [Mon, 21 Sep 2015 22:58:19 +0000 (15:58 -0700)]
src/cmd/compile/internal/gc: remove now unnecessary restriction on float exponent parsing

https://go-review.googlesource.com/#/c/13778/ fixed this issue in math/big.
Remove restriction in compiler.

Fixes #11326.

Change-Id: I1429d0dd0d79431706c65616413373fff58f081e
Reviewed-on: https://go-review.googlesource.com/14830
Reviewed-by: Rob Pike <r@golang.org>
9 years agoimage/png: integer underflow when decoding
Nathan Otterness [Sat, 19 Sep 2015 00:08:47 +0000 (20:08 -0400)]
image/png: integer underflow when decoding

This change addresses an integer underflow appearing only on systems
using a 32-bit int type. The patch addresses the problem by limiting the
length of unknown chunks to 0x7fffffff. This value appears to already be
checked for when parsing other chunk types, so the bug shouldn't appear
elsewhere in the package. The PNG spec recommends the maximum size for
any chunk to remain under 2^31, so this shouldn't cause errors with
valid images.

Fixes #12687

Change-Id: I17f0e1683515532c661cf2b0b2bc65309d1b7bb7
Reviewed-on: https://go-review.googlesource.com/14766
Reviewed-by: Nigel Tao <nigeltao@golang.org>
9 years agocmd/asm: fix crash triggered by nested #define
Rob Pike [Mon, 21 Sep 2015 21:18:13 +0000 (14:18 -0700)]
cmd/asm: fix crash triggered by nested #define

A panic was in place for an impossible condition that turned
out to be possible if one used a macro to define a macro.

Another go-fuzz "win".

Fixes #12654.

Change-Id: I0a7bb0f0eabb260c986bf7a2288860c78d8db1af
Reviewed-on: https://go-review.googlesource.com/14777
Reviewed-by: Andrew Gerrand <adg@golang.org>
9 years agocmd/compile/internal/gc: use sort.Interface for reflect methods
Håvard Haugen [Mon, 14 Sep 2015 22:16:35 +0000 (00:16 +0200)]
cmd/compile/internal/gc: use sort.Interface for reflect methods

Generate slices of method *Sig(nature)s instead of linked lists.
Remove custom lsort function in favor of sort.Interface.

Eliminates another use of stringsCompare.

Passes go build -a -toolexec 'toolstash -cmp' std cmd.

Change-Id: I9ed1664b7f55be9e967dd7196e396a76f6ea3422
Reviewed-on: https://go-review.googlesource.com/14559
Reviewed-by: Dave Cheney <dave@cheney.net>
9 years agocmd/vet: build the binary only once in the test
Rob Pike [Mon, 21 Sep 2015 18:06:43 +0000 (11:06 -0700)]
cmd/vet: build the binary only once in the test

Recent changes caused vet to build the binary for each Test function.
This is wasteful and will become only more so as more tests are added.
Use testing.Main to build only once.

Verified that compilation errors still appear if the binary cannot be
built.

Before:
real 0m11.169s
user 0m18.328s
sys 0m2.152s

After:
real 0m5.132s
user 0m9.404s
sys 0m1.168s

Of course if the compiler were fast we might not notice, but vet is
a big program and growing bigger all the time, as are the tests.

Change-Id: I209a8fdcace94bc5cec946f5dd365d7191f44c02
Reviewed-on: https://go-review.googlesource.com/14822
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
9 years agogo/printer: avoid stomping on err before checking it
Tarmigan Casebolt [Mon, 24 Aug 2015 02:51:26 +0000 (19:51 -0700)]
go/printer: avoid stomping on err before checking it

Change-Id: I97ba31e758d3396842ad99a08af696e49a5f1a7d
Reviewed-on: https://go-review.googlesource.com/13954
Run-TryBot: Robert Griesemer <gri@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Robert Griesemer <gri@golang.org>
9 years agocmd/vet: copy changes from golang.org/x/tools to cmd/vet
Rob Pike [Mon, 21 Sep 2015 17:30:04 +0000 (10:30 -0700)]
cmd/vet: copy changes from golang.org/x/tools to cmd/vet

This means bringing over the examples flag and sorting doc.go.

Subsequent changes will generalize the examples flag to a general
test naming flag, but let's start with the original code.

No more changes to golang.org/x/tools please. This should not have
happened (and letting it happen was partly my fault).

Change-Id: Ia879ea1d15d82372df14853f919263125dfb7b96
Reviewed-on: https://go-review.googlesource.com/14821
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
9 years agocmd/doc: don't drop const/var block if first entry is unexported
Rob Pike [Fri, 18 Sep 2015 21:53:33 +0000 (14:53 -0700)]
cmd/doc: don't drop const/var block if first entry is unexported

The code assumed that if the first entry was unexported, all the
entries were. The fix is simple: delete a bunch of code.

Fixes #12286.

Change-Id: Icb09274e99ce97df4d8bddbe59d17a5c0622e4c6
Reviewed-on: https://go-review.googlesource.com/14780
Reviewed-by: Andrew Gerrand <adg@golang.org>
9 years agoencoding/json: scanner: use byte, more consistent
Marvin Stenger [Sun, 20 Sep 2015 22:45:21 +0000 (00:45 +0200)]
encoding/json: scanner: use byte, more consistent

The fields step and redoState of struct scanner are now defined as
`func(s *scanner, c byte) int` instead of
`func(s *scanner, c int) int`, since bytes are sufficient.
Further changes improve the consistency in the scanner.go file.

Change-Id: Ifb85f2130d728d2b936d79914d87a1f0b5c6ee7d
Reviewed-on: https://go-review.googlesource.com/14801
Reviewed-by: Andrew Gerrand <adg@golang.org>
Run-TryBot: Andrew Gerrand <adg@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agocmd/compile/internal/gc: eliminate stringsCompare for stackvar sorting
Håvard Haugen [Sat, 19 Sep 2015 21:55:27 +0000 (23:55 +0200)]
cmd/compile/internal/gc: eliminate stringsCompare for stackvar sorting

Passes go build -a -toolexec 'toolstash -cmp' std cmd.

Change-Id: I2a87d31da74affdf3d0f358d0efdb3f1c646d917
Reviewed-on: https://go-review.googlesource.com/14759
Reviewed-by: Dave Cheney <dave@cheney.net>
9 years agobufio: fix overflow calculation in Scan
Rob Pike [Sat, 19 Sep 2015 18:39:22 +0000 (11:39 -0700)]
bufio: fix overflow calculation in Scan

I was being too clever, as usual. Write the obvious code to make sure
that when we grow the buffer we don't overflow.

Change-Id: I1641831177b0bb8a89ab6e9bcabccf6c2fcfe1d2
Reviewed-on: https://go-review.googlesource.com/14781
Reviewed-by: Minux Ma <minux@golang.org>
9 years agonet: check 'ok' return in dnsmsg when initially packing rr.Header()
Tarmigan Casebolt [Mon, 24 Aug 2015 02:51:31 +0000 (19:51 -0700)]
net: check 'ok' return in dnsmsg when initially packing rr.Header()

In the present code, there is no way for ok to ever return false, but
it still a good idea to check it.

Change-Id: I8f360018b33a5d85dabbbbec0f89ffc81f77ecbb
Reviewed-on: https://go-review.googlesource.com/13956
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agoruntime: race instrument read of convT2E/I arg
Dmitry Vyukov [Fri, 18 Sep 2015 09:40:36 +0000 (11:40 +0200)]
runtime: race instrument read of convT2E/I arg

Sometimes this read is instrumented by compiler when it creates
a temp to take address, but sometimes it is not (e.g. for global vars
compiler takes address of the global directly).

Instrument convT2E/I similarly to chansend and mapaccess.

Fixes #12664

Change-Id: Ia7807f15d735483996426c5f3aed60a33b279579
Reviewed-on: https://go-review.googlesource.com/14752
Reviewed-by: Keith Randall <khr@golang.org>
Run-TryBot: Dmitry Vyukov <dvyukov@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agocmd/dist: skip spectralnorm shootout test on the linux-arm-arm5 builder
Brad Fitzpatrick [Sat, 19 Sep 2015 05:39:03 +0000 (05:39 +0000)]
cmd/dist: skip spectralnorm shootout test on the linux-arm-arm5 builder

Temporary fix to get the arm5 builder happy again.

Without hardware floating point, this test takes over 20 minutes to
run.

A proper solution would probably be to run all the benchmark tests,
but with a much lower iteration count, just to exercise the code.

Updates golang/go#12688

Change-Id: Ie56c93d3bf2a5a693a33217ba1b1df3c6c856442
Reviewed-on: https://go-review.googlesource.com/14775
Reviewed-by: Dave Cheney <dave@cheney.net>
Reviewed-by: Minux Ma <minux@golang.org>
9 years agocmd/compile/internal/gc: move intLiteral to gc.Node
Dave Cheney [Fri, 18 Sep 2015 08:56:47 +0000 (18:56 +1000)]
cmd/compile/internal/gc: move intLiteral to gc.Node

intLiteral is used by the gins wrappers in arm64, ppc64 and
mips64. Refactor the function to a method on gc.Node and update
the callers to use the common copy.

Change-Id: I2db90d801a9cb18f8526eb921e13daa75ca1cf6f
Reviewed-on: https://go-review.googlesource.com/14744
Reviewed-by: Aram Hăvărneanu <aram@mgk.ro>
Reviewed-by: Dave Cheney <dave@cheney.net>
Run-TryBot: Dave Cheney <dave@cheney.net>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agoruntime: remove flaky TestInvalidptrCrash to fix build
Austin Clements [Sat, 19 Sep 2015 01:17:07 +0000 (21:17 -0400)]
runtime: remove flaky TestInvalidptrCrash to fix build

This test fails on arm64 and some amd64 OSs and fails on Linux/amd64
if you remove the first runtime.GC(), which should be unnecessary, and
run it in all.bash (but not if you run it in isolation). I don't
understand any of these failures, so for now just remove this test.

TBR=rlh

Change-Id: Ibed00671126000ed7dc5b5d4af1f86fe4a1e30e1
Reviewed-on: https://go-review.googlesource.com/14767
Reviewed-by: Austin Clements <austin@google.com>
Run-TryBot: Austin Clements <austin@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agoruntime: avoid debug prints of huge objects
Austin Clements [Fri, 18 Sep 2015 16:06:24 +0000 (12:06 -0400)]
runtime: avoid debug prints of huge objects

Currently when the GC prints an object for debugging (e.g., for a
failed invalidptr or checkmark check), it dumps the entire object. To
avoid inundating the user with output for really large objects, limit
this to printing just the first 128 words (which are most likely to be
useful in identifying the type of an object) and the 32 words around
the problematic field.

Change-Id: Id94a5c9d8162f8bd9b2a63bf0b1bfb0adde83c68
Reviewed-on: https://go-review.googlesource.com/14764
Reviewed-by: Rick Hudson <rlh@golang.org>
9 years agoruntime: improve invalid pointer error message
Austin Clements [Fri, 18 Sep 2015 15:55:31 +0000 (11:55 -0400)]
runtime: improve invalid pointer error message

By default, the runtime panics if it detects a pointer to an
unallocated span. At this point, this usually catches bad uses of
unsafe or cgo in user code (though it could also catch runtime bugs).
Unfortunately, the rather cryptic error misleads users, offers users
little help with debugging their own problem, and offers the Go
developers little help with root-causing.

Improve the error message in various ways. First, the wording is
improved to make it clearer what condition was detected and to suggest
that this may be the result of incorrect use of unsafe or cgo. Second,
we add a dump of the object containing the bad pointer so that there's
at least some hope of figuring out why a bad pointer was stored in the
Go heap.

Change-Id: I57b91b12bc3cb04476399d7706679e096ce594b9
Reviewed-on: https://go-review.googlesource.com/14763
Reviewed-by: Rick Hudson <rlh@golang.org>
9 years agocrypto/x509: return err if marshalPublicKey fails to marshal an rsa public key
Tarmigan Casebolt [Mon, 24 Aug 2015 02:51:16 +0000 (19:51 -0700)]
crypto/x509: return err if marshalPublicKey fails to marshal an rsa public key

Change-Id: I9bd5c1b66fd90f0b54bd1a8f3e57b6830d2b7733
Reviewed-on: https://go-review.googlesource.com/13846
Reviewed-by: Adam Langley <agl@golang.org>
Run-TryBot: Adam Langley <agl@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agodoc: go1.6.txt: template.IsTrue, bufio.Scanner.Buffer
Rob Pike [Fri, 18 Sep 2015 18:58:35 +0000 (11:58 -0700)]
doc: go1.6.txt: template.IsTrue, bufio.Scanner.Buffer

Change-Id: Iaa01d34caf09c12c017dc0379d4fe1d2cffd5340
Reviewed-on: https://go-review.googlesource.com/14728
Reviewed-by: Rob Pike <r@golang.org>
9 years agobufio: allow Scanner to accept a user-provided buffer
Rob Pike [Tue, 15 Sep 2015 21:14:44 +0000 (14:14 -0700)]
bufio: allow Scanner to accept a user-provided buffer

Add Scanner.Buffer, which lets the user give a buffer to
the scanner and set the maximum token size.

We call it Buffer not SetBuffer for consistency with Split, which
perhaps should have been called SetSplit; too late regardless.

Both Buffer and Split panic if they are called after Scan. The
panic in Split is new, but the comment on the method already
said it needed to be called first, so we might as well add the
verification while we're doing it for Buffer.

This method allows precise user control of storage.

Fixes #11702.

Change-Id: I80e3d0e3830562fdabd4f7b08f322e1378248c39
Reviewed-on: https://go-review.googlesource.com/14599
Reviewed-by: Andrew Gerrand <adg@golang.org>
Reviewed-by: roger peppe <rogpeppe@gmail.com>
9 years agocmd/dist: shard shootout test units
Brad Fitzpatrick [Fri, 18 Sep 2015 16:39:35 +0000 (16:39 +0000)]
cmd/dist: shard shootout test units

Instead of a 10 second test unit, make it 13 sub-second ones. This
takes advantage of multiple builders better.

Fixes #12623

Change-Id: I3fb2eb02f899f25749e34b546b9d41b742a746cd
Reviewed-on: https://go-review.googlesource.com/14738
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agoasm: handle EOF better
Rob Pike [Thu, 17 Sep 2015 20:46:46 +0000 (13:46 -0700)]
asm: handle EOF better

Add some error catches to prevent looping at EOF.
Also give better diagnostics.
Also add tests for these cases.

Fixes #12656.

Change-Id: I1355fc149b71c868e740bfa53de29c25d160777d
Reviewed-on: https://go-review.googlesource.com/14710
Reviewed-by: Andrew Gerrand <adg@golang.org>
9 years agoasm: give error message for unadorned symbol reference
Rob Pike [Thu, 17 Sep 2015 17:37:07 +0000 (10:37 -0700)]
asm: give error message for unadorned symbol reference

On amd64, the program

TEXT    foo0(SB),7,$-8
    ADDQ R520, R1
    RET

used to trigger this error because R520 was being passed through to obj:

asm: doasm: notfound ft=23 tt=23 00000 (x.s:2) ADDQ 0, 0 23 23

Now it gets this one, as it is indeed a parse error:

x.s:2: illegal addressing mode for symbol R520

This couldn't be fixed until #12632 had been fixed for arm64.

Fixes #12470.

Change-Id: I19830c4ae9337887b93f85d9a239e2b89dbb2219
Reviewed-on: https://go-review.googlesource.com/14691
Reviewed-by: Aram Hăvărneanu <aram@mgk.ro>
9 years agogo/types: handle import "C" more like cmd/compile
Robert Griesemer [Thu, 17 Sep 2015 23:08:57 +0000 (16:08 -0700)]
go/types: handle import "C" more like cmd/compile

Fixes #12667.

Change-Id: I68e73e26da9938606304163ae2637e3c6bacd6f6
Reviewed-on: https://go-review.googlesource.com/14722
Reviewed-by: Alan Donovan <adonovan@google.com>
9 years agogo/types: unified handling of assignment errors
Robert Griesemer [Thu, 17 Sep 2015 21:13:56 +0000 (14:13 -0700)]
go/types: unified handling of assignment errors

- simpler code
- closer to gc error messages
- more context information in some cases

Change-Id: Iad155a887b838a4fc1edf719eed18269670b5ede
Reviewed-on: https://go-review.googlesource.com/14720
Reviewed-by: Alan Donovan <adonovan@google.com>
9 years agogo/types: cleanup handling of multi-valued expressions
Robert Griesemer [Thu, 17 Sep 2015 18:06:27 +0000 (11:06 -0700)]
go/types: cleanup handling of multi-valued expressions

- more uniform error messages
- removed unused code

Change-Id: I625d5c2e51a543450ad091f97cec538023ddb1dd
Reviewed-on: https://go-review.googlesource.com/14692
Reviewed-by: Alan Donovan <adonovan@google.com>
9 years agogo/types: better error message when using multi-valued expressions in single-value...
Robert Griesemer [Thu, 17 Sep 2015 00:52:12 +0000 (17:52 -0700)]
go/types: better error message when using multi-valued expressions in single-value context

Also: Added initial set of (missing and/or spread out) tests for binary operations.

Fixes #11896.

Change-Id: I037436d8318c18f9758b435eca2d45b3bdd17ef8
Reviewed-on: https://go-review.googlesource.com/14660
Reviewed-by: Alan Donovan <adonovan@google.com>
9 years agogo/types: enable disabled test
Robert Griesemer [Wed, 16 Sep 2015 23:53:57 +0000 (16:53 -0700)]
go/types: enable disabled test

Change-Id: I101e9bc722f5ba4fa9941cb544d9c276fd4c0b8c
Reviewed-on: https://go-review.googlesource.com/14651
Reviewed-by: Alan Donovan <adonovan@google.com>
9 years agogo/types: report detailed reason in error messages for invalid assignments
Robert Griesemer [Wed, 16 Sep 2015 18:15:54 +0000 (11:15 -0700)]
go/types: report detailed reason in error messages for invalid assignments

Fixes #10260.

Change-Id: I52d059144608912e6f7f9516e4961a75e9463355
Reviewed-on: https://go-review.googlesource.com/14644
Reviewed-by: Alan Donovan <adonovan@google.com>
9 years agocmd/compile: convert externdecl to []*Node
Dave Cheney [Thu, 10 Sep 2015 05:57:39 +0000 (15:57 +1000)]
cmd/compile: convert externdecl to []*Node

This one of a set of changes to make the transition away from NodeList
easier by removing cases in which NodeList doesn't act semi-trivially like a
[]*Node.

This CL was originally prepared by Josh Bleecher Snyder <josharian@gmail.com>.

This change passes go build -toolexec 'toolstash -cmp' -a std.

Change-Id: Ifd73501e06e8ea5efd028b6d473b3e5d1b07a5ac
Reviewed-on: https://go-review.googlesource.com/14570
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
9 years agocmd/go: provide full path as os.Args[0] when invoking tools
Shenghou Ma [Fri, 11 Sep 2015 04:26:50 +0000 (00:26 -0400)]
cmd/go: provide full path as os.Args[0] when invoking tools

cmd/dist needs to re-exec or open itself to detect GOARM (CL 3973) and
detect host machine endianness (CL 14460).

Change-Id: If6438831ab0715ba8e236d64bb2c7c1bde1470aa
Reviewed-on: https://go-review.googlesource.com/14476
Reviewed-by: Ian Lance Taylor <iant@golang.org>
9 years agocmd/compile/internal/gc: avoid stringsCompare for string literals
Håvard Haugen [Thu, 17 Sep 2015 19:01:29 +0000 (21:01 +0200)]
cmd/compile/internal/gc: avoid stringsCompare for string literals

Passes go build -a -toolexec 'toolstash -cmp' std cmp.

Change-Id: I7567355d405c976c5d91a0cd4e9486ebeb348dbb
Reviewed-on: https://go-review.googlesource.com/14682
Reviewed-by: Dave Cheney <dave@cheney.net>
Run-TryBot: Dave Cheney <dave@cheney.net>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
9 years agocmd/compile/internal/gc: add unit test for cmpstackvar
Håvard Haugen [Tue, 15 Sep 2015 19:43:53 +0000 (21:43 +0200)]
cmd/compile/internal/gc: add unit test for cmpstackvar

A followup CL will rewrite listsort to use the new cmpstackvarlt and
change cmpstackvar to avoid stringsCompare.

Change-Id: Idf0857a3bd67f9e2243ba82aa0bff510612927c3
Reviewed-on: https://go-review.googlesource.com/14611
Reviewed-by: Dave Cheney <dave@cheney.net>
9 years agocmd/compile/internal/gc: cleaning; use range when appropriate
Marvin Stenger [Tue, 8 Sep 2015 20:22:44 +0000 (22:22 +0200)]
cmd/compile/internal/gc: cleaning; use range when appropriate

Made use of range statement in for loops.
Cleaning along the way:
-remove unnecessary variable declarations
-rename variables
-remove dead code

This change passes go build -toolexec 'toolstash -cmp' -a std.

Change-Id: Ife8c2a98482a81ba91f5bbb65142d9f3dc46d6ee
Reviewed-on: https://go-review.googlesource.com/14379
Run-TryBot: Dave Cheney <dave@cheney.net>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Dave Cheney <dave@cheney.net>
9 years agotext/template: export isTrue
Rob Pike [Tue, 15 Sep 2015 16:27:22 +0000 (09:27 -0700)]
text/template: export isTrue

The definition of 'truth' used by if etc. is not trivial to compute, so publish
the implementation to allow custom template functions to have the
same definition as the template language itself.

Fixes #12033.

Change-Id: Icdfd6039722d7d3f984ba0905105eb3253e14831
Reviewed-on: https://go-review.googlesource.com/14593
Reviewed-by: Andrew Gerrand <adg@golang.org>
9 years agoencoding/gob: document allocation/merge behavior
Rob Pike [Thu, 17 Sep 2015 18:17:02 +0000 (11:17 -0700)]
encoding/gob: document allocation/merge behavior

This is understood, obvious (to me), and well known but has not been clearly documented.

Fixes #11117.

Change-Id: Ib2b1e318924748d1eac0d735ad6286533be7fd39
Reviewed-on: https://go-review.googlesource.com/14693
Reviewed-by: Ian Lance Taylor <iant@golang.org>
9 years agomisc/ios: skip revoked certificates
David Crawshaw [Thu, 17 Sep 2015 16:13:53 +0000 (12:13 -0400)]
misc/ios: skip revoked certificates

Change-Id: If65e5e55b359a61740d2ef185147bb6df90e0b0c
Reviewed-on: https://go-review.googlesource.com/14654
Reviewed-by: Hyang-Ah Hana Kim <hyangah@gmail.com>
9 years agocmd/asm/internal/asm: add aliases for ARM64 condition codes
Aram Hăvărneanu [Thu, 17 Sep 2015 09:47:16 +0000 (11:47 +0200)]
cmd/asm/internal/asm: add aliases for ARM64 condition codes

Add CS as an alias for HS, and CC as an alias for LO, otherwise

CSINV CS, R1, R2, R3

was interpreted as

CSINV 0, R1, R2, R3

Also fix the corresponding faulty test.

Fixes #12632
Updates #12470

Change-Id: I974cfc7e5ced682d4754ba09b0b102cb08a46567
Reviewed-on: https://go-review.googlesource.com/14680
Reviewed-by: Rob Pike <r@golang.org>
9 years agoos: touch up the EINTR retry loop in OpenFile
Aaron Jacobs [Thu, 17 Sep 2015 06:28:15 +0000 (16:28 +1000)]
os: touch up the EINTR retry loop in OpenFile

In particular, don't use goto and do restrict the behavior to darwin.
This addresses comments from http://golang.org/cl/14484.

Change-Id: I5b99e1762d1c5b27fdd12b72a5c6d981f6a92f0f
Reviewed-on: https://go-review.googlesource.com/14673
Reviewed-by: Rob Pike <r@golang.org>
9 years agoruntime/trace: fix tracing of blocking system calls
Shawn Walker-Salas [Tue, 18 Aug 2015 21:13:38 +0000 (14:13 -0700)]
runtime/trace: fix tracing of blocking system calls

The placement and invocation of traceGoSysCall when using
entersyscallblock() instead of entersyscall() differs enough that the
TestTraceSymbolize test can fail on some platforms.

This change moves the invocation of traceGoSysCall for entersyscall() so
that the same number of "frames to skip" are present in the trace as when
entersyscallblock() is used ensuring system call traces remain identical
regardless of internal implementation choices.

Fixes golang/go#12056

Change-Id: I8361e91aa3708f5053f98263dfe9feb8c5d1d969
Reviewed-on: https://go-review.googlesource.com/13861
Run-TryBot: Dmitry Vyukov <dvyukov@google.com>
Reviewed-by: Dmitry Vyukov <dvyukov@google.com>
9 years agoruntime: print errno and byte count before crashing in mem_windows.go
Alex Brainman [Wed, 16 Sep 2015 04:18:12 +0000 (14:18 +1000)]
runtime: print errno and byte count before crashing in mem_windows.go

As per iant suggestion during issue #12587 crash investigation.

Also adjust incorrect throw message in sysUsed while we are here.

Change-Id: Ice07904fdd6e0980308cb445965a696d26a1b92e
Reviewed-on: https://go-review.googlesource.com/14633
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agoos: handle EINTR from open(2).
Aaron Jacobs [Wed, 16 Sep 2015 22:32:09 +0000 (08:32 +1000)]
os: handle EINTR from open(2).

The man page for sigaction(2) on OS X doesn't guarantee that SA_RESTART
will work for open(2) on regular files:

    The affected system calls include open(2), read(2), write(2),
    sendto(2), recvfrom(2), sendmsg(2) and recvmsg(2) on a
    communications channel or a slow device (such as a terminal, but not
    a regular file) and during a wait(2) or ioctl(2).

I've never observed EINTR from open(2) for a traditional file system
such as HFS+, but it's easy to observe with a fuse file system that is
slightly slow (cf. https://goo.gl/UxsVgB). After this change, the
problem can no longer be reproduced when calling os.OpenFile.

Fixes #11180.

Change-Id: I967247430e20a7d29a285b3d76bf3498dc4773db
Reviewed-on: https://go-review.googlesource.com/14484
Reviewed-by: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agoruntime/debug: more explicit Stack docs
David Crawshaw [Wed, 16 Sep 2015 14:29:53 +0000 (10:29 -0400)]
runtime/debug: more explicit Stack docs

Change-Id: I81a7f22be827519b5290b4acbcba357680cad3c4
Reviewed-on: https://go-review.googlesource.com/14605
Reviewed-by: Rob Pike <r@golang.org>
9 years agocmd/link: fix addition of -Wl,-z,relro
Michael Hudson-Doyle [Wed, 16 Sep 2015 10:06:16 +0000 (22:06 +1200)]
cmd/link: fix addition of -Wl,-z,relro

Not sure how I managed to do this, or get it past review.

Change-Id: I141b97ef8e09dcc9c910c45493a584a3dced2b28
Reviewed-on: https://go-review.googlesource.com/14634
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agogo/parser: better error message for incorrect type switch header
Robert Griesemer [Mon, 14 Sep 2015 23:59:15 +0000 (16:59 -0700)]
go/parser: better error message for incorrect type switch header

Fixes 11829.

Change-Id: I2e39f61e12953147b0cd6a11d29179c500c94964
Reviewed-on: https://go-review.googlesource.com/14566
Reviewed-by: Chris Manghane <cmang@golang.org>
9 years agogo/parser: comma is not permitted at the end of a struct field list
Robert Griesemer [Mon, 14 Sep 2015 23:30:37 +0000 (16:30 -0700)]
go/parser: comma is not permitted at the end of a struct field list

Fixes #11611.

Change-Id: I63d35cf15c3be759c899e3e561e631330dcc0bbb
Reviewed-on: https://go-review.googlesource.com/14565
Run-TryBot: Robert Griesemer <gri@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Chris Manghane <cmang@golang.org>
9 years agodoc: go1.6.txt: add math/rand.Read
Rob Pike [Wed, 16 Sep 2015 17:55:25 +0000 (10:55 -0700)]
doc: go1.6.txt: add math/rand.Read

Change-Id: I94af55bb894409b77bd87df36be9471dcc544fda
Reviewed-on: https://go-review.googlesource.com/14627
Reviewed-by: Rob Pike <r@golang.org>
9 years agomath/rand: make Rand fulfill the Reader interface
Spencer Nelson [Fri, 11 Sep 2015 17:41:45 +0000 (13:41 -0400)]
math/rand: make Rand fulfill the Reader interface

Add a Read function to Rand which reads random bytes into a buffer.

Fixes #8330

Change-Id: I85b90277b8be9287c6697def8dbefe0029b6ee06
Reviewed-on: https://go-review.googlesource.com/14522
Reviewed-by: Rob Pike <r@golang.org>
9 years agoruntime: optimize duffzero for amd64.
Ilya Tocar [Wed, 9 Sep 2015 11:10:12 +0000 (14:10 +0300)]
runtime: optimize duffzero for amd64.

Use MOVUPS to zero 16 bytes at a time.

results (haswell):

name             old time/op  new time/op  delta
ClearFat8-48     0.62ns ± 2%  0.62ns ± 1%     ~     (p=0.085 n=20+15)
ClearFat12-48    0.93ns ± 2%  0.93ns ± 2%     ~     (p=0.757 n=19+19)
ClearFat16-48    1.23ns ± 1%  1.23ns ± 1%     ~     (p=0.896 n=19+17)
ClearFat24-48    1.85ns ± 2%  1.84ns ± 0%   -0.51%  (p=0.023 n=20+15)
ClearFat32-48    2.45ns ± 0%  2.46ns ± 2%     ~     (p=0.053 n=17+18)
ClearFat40-48    1.99ns ± 0%  0.92ns ± 2%  -53.54%  (p=0.000 n=19+20)
ClearFat48-48    2.15ns ± 1%  0.92ns ± 2%  -56.93%  (p=0.000 n=19+20)
ClearFat56-48    2.46ns ± 1%  1.23ns ± 0%  -49.98%  (p=0.000 n=19+14)
ClearFat64-48    2.76ns ± 0%  2.14ns ± 1%  -22.21%  (p=0.000 n=17+17)
ClearFat128-48   5.21ns ± 0%  3.99ns ± 0%  -23.46%  (p=0.000 n=17+19)
ClearFat256-48   10.3ns ± 4%   7.7ns ± 0%  -25.37%  (p=0.000 n=20+17)
ClearFat512-48   20.2ns ± 4%  15.0ns ± 1%  -25.58%  (p=0.000 n=20+17)
ClearFat1024-48  39.7ns ± 2%  29.7ns ± 0%  -25.05%  (p=0.000 n=19+19)

Change-Id: I200401eec971b2dd2450c0651c51e378bd982405
Reviewed-on: https://go-review.googlesource.com/14408
Reviewed-by: Keith Randall <khr@golang.org>
Run-TryBot: Keith Randall <khr@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agohash/crc32: add AMD64 optimized IEEE CRC calculation
Klaus Post [Sun, 30 Aug 2015 20:29:00 +0000 (22:29 +0200)]
hash/crc32: add AMD64 optimized IEEE CRC calculation

IEEE is the most commonly used CRC-32 polynomial, used by zip, gzip and others.

Based on http://www.intel.com/content/dam/www/public/us/en/documents/white-papers/fast-crc-computation-generic-polynomials-pclmulqdq-paper.pdf

benchmark                       old ns/op     new ns/op     delta
BenchmarkIEEECrc1KB-8           3193          352           -88.98%
BenchmarkIEEECrc4KB-8           5025          1307          -73.99%
BenchmarkCastagnoliCrc1KB-8     126           126           +0.00%

benchmark                       old MB/s     new MB/s     speedup
BenchmarkIEEECrc1KB-8           320.68       2901.92      9.05x
BenchmarkIEEECrc4KB-8           815.08       3131.80      3.84x
BenchmarkCastagnoliCrc1KB-8     8100.80      8109.78      1.00x

Change-Id: I99c9a48365f631827f516e44f97e86155f03cb90
Reviewed-on: https://go-review.googlesource.com/14080
Reviewed-by: Keith Randall <khr@golang.org>
9 years agocmd/go: indent first test binary flag description for `go test -h`
Tormod Erevik Lea [Wed, 16 Sep 2015 08:54:31 +0000 (10:54 +0200)]
cmd/go: indent first test binary flag description for `go test -h`

Fixes #12642

Change-Id: I0b94437055b7d444f5caf7ea310e85357c467bdf
Reviewed-on: https://go-review.googlesource.com/14612
Reviewed-by: Ian Lance Taylor <iant@golang.org>
9 years agoruntime/debug: implement Stack using runtime.Stack
David Crawshaw [Tue, 15 Sep 2015 21:16:38 +0000 (17:16 -0400)]
runtime/debug: implement Stack using runtime.Stack

Fixes #12363

Change-Id: I1a025ab6a1cbd5a58f5c2bce5416788387495428
Reviewed-on: https://go-review.googlesource.com/14604
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: David Crawshaw <crawshaw@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agoruntime: preserve R11 in darwin/arm entrypoint
David Crawshaw [Tue, 15 Sep 2015 17:40:24 +0000 (13:40 -0400)]
runtime: preserve R11 in darwin/arm entrypoint

The _rt0_arm_darwin_lib entrypoint has to conform to the darwin ARMv7
calling convention, which requires functions to preserve the value of
R11. Go uses R11 as the liblink REGTMP register, so save it manually.

Also avoid using R4, which is also callee-save.

Fixes #12590

Change-Id: I9c3b374e330f81ff8fc9c01fa20505a33ddcf39a
Reviewed-on: https://go-review.googlesource.com/14603
Reviewed-by: Ian Lance Taylor <iant@golang.org>
9 years agotest: move allocation before munmap in recover4
Michael Hudson-Doyle [Wed, 16 Sep 2015 04:20:31 +0000 (16:20 +1200)]
test: move allocation before munmap in recover4

recover4 allocates 16 pages of memory via mmap, makes a 4 page hole in it with
munmap, allocates another 16 pages of memory via normal allocation and then
tries to copy from one to the other. For some reason on arm64 (but no other
platform I have tested) the second allocation sometimes causes the runtime to
ask the kernel for 4 additional pages of memory -- which the kernel satisfies
by remapping the pages that were just unmapped!

Moving the second allocation before the munmap fixes this behaviour, I can run
recover4 tens of thousands of times without failure with this fix vs a failure
rate of ~0.5% before.

Fixes #12549

Change-Id: I490b895b606897e4f7f25b1b51f5d485a366fffb
Reviewed-on: https://go-review.googlesource.com/14632
Reviewed-by: Dave Cheney <dave@cheney.net>
9 years agocmd/compile/internal/gc: remove dead code
Håvard Haugen [Sun, 13 Sep 2015 21:57:28 +0000 (23:57 +0200)]
cmd/compile/internal/gc: remove dead code

Found with https://github.com/remyoudompheng/go-misc/deadcode:

deadcode: walk.go:2228:1: applywritebarrier_bv is unused
deadcode: subr.go:355:1: gethunk is unused
deadcode: subr.go:1991:1: localexpr is unused
deadcode: dcl.go:82:1: poptodcl is unused
deadcode: swt.go:810:1: dumpcase is unused
deadcode: esc.go:251:1: satAdd8 is unused
deadcode: esc.go:387:1: outputsPerTag is unused
deadcode: obj.go:190:1: duint64 is unused
deadcode: obj.go:287:1: dstringptr is unused
deadcode: plive.go:95:1: xmalloc is unused
deadcode: plive.go:119:1: freeblock is unused

followed by

deadcode: go.go:633:1: hunk is unused
deadcode: go.go:635:1: nhunk is unused
deadcode: go.go:637:1: thunk is unused

after 'gethunk' was removed.

Some dead code in bv.go, mparith3.go, and dcl.go was left as is.

Passes go build -a -toolexec 'toolstash -cmp' std cmd.

Change-Id: Ia63519adedc8650d7095572ddd454fd923d3204d
Reviewed-on: https://go-review.googlesource.com/14610
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

9 years agocmd/compile/internal/gc: remove unnecessary stringsCompare
Håvard Haugen [Mon, 14 Sep 2015 18:53:44 +0000 (20:53 +0200)]
cmd/compile/internal/gc: remove unnecessary stringsCompare

Remove several uses of stringsCompare.

Passes go build -a -toolexec 'toolstash -cmp' std cmd.

Change-Id: I3f2323df2ad8c03bad77e0a91d6e2e714803705b
Reviewed-on: https://go-review.googlesource.com/14556
Reviewed-by: Dave Cheney <dave@cheney.net>
9 years agocmd/go: don't run TestIssue7573 if cgo not supported
Ian Lance Taylor [Tue, 15 Sep 2015 21:04:35 +0000 (14:04 -0700)]
cmd/go: don't run TestIssue7573 if cgo not supported

Fixes #12629.

Change-Id: Iee96dc4f806a38f3cd8e065b8d0d5f682bb7e29b
Reviewed-on: https://go-review.googlesource.com/14597
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
9 years agointernal/obj: protect against nil addr.Sym
Rob Pike [Tue, 15 Sep 2015 17:59:27 +0000 (10:59 -0700)]
internal/obj: protect against nil addr.Sym

This has been the root cause of a number of crashes caused by
fuzz throwing modem noise at the assembler, which in turn attempts
to print diagnostics but instead just gets crashes.

Fixes #12627.

Change-Id: I72c2da79d8eb240e1a37aa6140454c552b05e0f1
Reviewed-on: https://go-review.googlesource.com/14595
Reviewed-by: Ian Lance Taylor <iant@golang.org>
9 years agosrc/cmd/compile/internal/gc: fix type assertion in overflow check
Robert Griesemer [Tue, 15 Sep 2015 18:16:58 +0000 (11:16 -0700)]
src/cmd/compile/internal/gc: fix type assertion in overflow check

Fixes #11600.

Change-Id: I8871d4e525168fed35115855483a237bbd6e5445
Reviewed-on: https://go-review.googlesource.com/14596
Reviewed-by: Josh Bleecher Snyder <josharian@gmail.com>
9 years agodebug/dwarf: add test for split DWARF
Austin Clements [Sun, 13 Sep 2015 16:49:05 +0000 (12:49 -0400)]
debug/dwarf: add test for split DWARF

This adds a test that debug/dwarf can read the skeleton DWARF data
from a split DWARF image (though it doesn't currently support piecing
the external DWARF data back together). This should work because
there's nothing particularly different about skeleton DWARF data, but
previously failed because of poor handling of unrecognized attributes.

Updates #12592.

Change-Id: I2fc5f4679883b05ebd7ec9f0b5c398a758181a32
Reviewed-on: https://go-review.googlesource.com/14542
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Reviewed-by: jcd . <jcd@golang.org>
9 years agodebug/dwarf: return ClassUnknown if attribute class cannot be determined
Austin Clements [Sun, 13 Sep 2015 16:12:13 +0000 (12:12 -0400)]
debug/dwarf: return ClassUnknown if attribute class cannot be determined

Currently, if the .debug_abbrev section of an ELF file contains
attributes that aren't known to the dwarf package and that have form
formSecOffset, the dwarf package will fail to open the DWARF data with
an error like "decoding dwarf section abbrev at offset 0x17: cannot
determine class of unknown attribute with formSecOffset". For the most
part, the class is implied by the form encoded in the abbrev section,
but formSecOffset can imply many different DWARF classes. Hence,
debug/dwarf disambiguates these using a table of known attributes.
However, it will reject the entire image if it encounters an attribute
it can't determine the class of. This is particularly unfortunate
because the caller may never even uses the offending attribute.

Fix this by introducing a ClassUnknown attribute class to use as a
fallback in these cases. This allows the dwarf package to load the
DWARF data and isolates the problem to just the affected attributes.

Fixes #12592.

Change-Id: I766227b136e9757f8b89c0b3ab8e9ddea899d94f
Reviewed-on: https://go-review.googlesource.com/14541
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Reviewed-by: jcd . <jcd@golang.org>
9 years agoruntime: on unexpected netpoll error, throw instead of looping
Ian Lance Taylor [Fri, 10 Jul 2015 22:28:01 +0000 (15:28 -0700)]
runtime: on unexpected netpoll error, throw instead of looping

The current code prints an error message and then tries to carry on.
This is not helpful for Go users: they see a message that means
nothing and that they can do nothing about.  In the only known case of
this message, in issue 11498, the best guess is that the netpoll code
went into an infinite loop.  Instead of doing that, crash the program.

Fixes #11498.

Change-Id: Idda3456c5b708f0df6a6b56c5bb4e796bbc39d7c
Reviewed-on: https://go-review.googlesource.com/12047
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Reviewed-by: Dmitry Vyukov <dvyukov@google.com>
9 years agoruntime: fix aeshash of empty string
Keith Randall [Tue, 1 Sep 2015 19:53:15 +0000 (12:53 -0700)]
runtime: fix aeshash of empty string

Aeshash currently computes the hash of the empty string as
hash("", seed) = seed.  This is bad because the hash of a compound
object with empty strings in it doesn't include information about
where those empty strings were.  For instance [2]string{"", "foo"}
and [2]string{"foo", ""} might get the same hash.

Fix this by returning a scrambled seed instead of the seed itself.
With this fix, we can remove the scrambling done by the generated
array hash routines.

The test also rejects hash("", seed) = 0, if we ever thought
it would be a good idea to try that.

The fallback hash is already OK in this regard.

Change-Id: Iaedbaa5be8d6a246dc7e9383d795000e0f562037
Reviewed-on: https://go-review.googlesource.com/14129
Reviewed-by: jcd . <jcd@golang.org>
9 years agomath/big: Add small complete example of big.Rat usage
Alberto Donizetti [Sun, 6 Sep 2015 17:59:12 +0000 (19:59 +0200)]
math/big: Add small complete example of big.Rat usage

Updates #11241

Change-Id: If71f651f3b8aca432c91314358b93f195217d9ec
Reviewed-on: https://go-review.googlesource.com/14317
Reviewed-by: Robert Griesemer <gri@golang.org>
9 years agodoc: go1.6.txt: template.Funcs checks names
Rob Pike [Tue, 15 Sep 2015 16:12:31 +0000 (09:12 -0700)]
doc: go1.6.txt: template.Funcs checks names

Change-Id: I7de85034d499a9f859ab37d56463073f5cb29b35
Reviewed-on: https://go-review.googlesource.com/14592
Reviewed-by: Rob Pike <r@golang.org>
9 years agoasm: more early returns on errors
Rob Pike [Mon, 14 Sep 2015 18:48:19 +0000 (11:48 -0700)]
asm: more early returns on errors

More protection against random input bytes.

Fixes #12614.

Change-Id: Ie9f817de1376a10bb80b22ecee3bae4f1d26cc6c
Reviewed-on: https://go-review.googlesource.com/14563
Reviewed-by: Andrew Gerrand <adg@golang.org>
9 years agotext/template: verify that names in FuncMap are valid identifiers
Rob Pike [Mon, 14 Sep 2015 18:27:20 +0000 (11:27 -0700)]
text/template: verify that names in FuncMap are valid identifiers

There was no verification in Funcs that the map had valid names,
which meant that the error could only be caught when parsing
the template that tried to use them. Fix this by validating the names
in Funcs and panicking before parsing if there is a bad name.

This is arguably an API change, since it didn't trigger a panic
before, but Funcs did already panic if the function itself was no
good, so I argue it's an acceptable change to add more sanity
checks.

Fixes #9685.

Change-Id: Iabf1d0602c49d830f3ed71ca1ccc7eb9a5521ff5
Reviewed-on: https://go-review.googlesource.com/14562
Reviewed-by: Andrew Gerrand <adg@golang.org>
9 years agocmd/compile/internal/gc: cleaning lex.go
Marvin Stenger [Thu, 10 Sep 2015 22:03:19 +0000 (00:03 +0200)]
cmd/compile/internal/gc: cleaning lex.go

Cleaning along the way:
-convert variable types from int to bool
-remove unnecessary functions
-remove unnecessary type conversion
-remove unnecessary variable declarations
-transform struct{string,string} with lookup to map[string]string

This change passes go build -toolexec 'toolstash -cmp' -a std.

Change-Id: I259728fe4afd7f23b67f08fab856ce0abee57b21
Reviewed-on: https://go-review.googlesource.com/14435
Reviewed-by: Ian Lance Taylor <iant@golang.org>
9 years agonet: remove named parameters in Listener.Accept doc signature
Brad Fitzpatrick [Tue, 15 Sep 2015 03:58:21 +0000 (20:58 -0700)]
net: remove named parameters in Listener.Accept doc signature

They added no value.

Change-Id: I9e690379d2dfd983266de0ea5231f2b57c8b1517
Reviewed-on: https://go-review.googlesource.com/14568
Reviewed-by: Ian Lance Taylor <iant@golang.org>
9 years agocmd/go: skip external tests on linux/arm
Dave Cheney [Fri, 7 Aug 2015 01:15:54 +0000 (11:15 +1000)]
cmd/go: skip external tests on linux/arm

CL 13166 skipped external tests on freebsd/arm with the rationale
that the cmd/go tests are not architecture dependent.

This CL does the same for linux/arm to help linux/arm users who are
building Go on platforms like the Raspberry Pi where ./all.bash
frequently times out due to a lack of resources.

Change-Id: Iae1a25b63b74200da3f1b5637da0fa5c2dceeb83
Reviewed-on: https://go-review.googlesource.com/13342
Reviewed-by: Ian Lance Taylor <iant@golang.org>
9 years agoruntime: provide room for first 4 syscall parameters in windows usleep2
Alex Brainman [Thu, 3 Sep 2015 07:48:21 +0000 (17:48 +1000)]
runtime: provide room for first 4 syscall parameters in windows usleep2

Windows amd64 requires all syscall callers to provide room for first
4 parameters on stack. We do that for all our syscalls, except inside
of usleep2. In https://codereview.appspot.com/7563043#msg3 rsc says:

"We don't need the stack alignment and first 4 parameters on amd64
because it's just a system call, not an ordinary function call."

He seems to be wrong on both counts. But alignment is already fixed.
Fix parameter space now too.

Fixes #12444

Change-Id: I66a2a18d2f2c3846e3aa556cc3acc8ec6240bea0
Reviewed-on: https://go-review.googlesource.com/14282
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
9 years agogo/parser: better error message for unexpected ',' in struct type
Robert Griesemer [Mon, 14 Sep 2015 19:01:04 +0000 (12:01 -0700)]
go/parser: better error message for unexpected ',' in struct type

Fixes #12437.

Change-Id: I5463970a6259527003eb0e12903a338cc78e0683
Reviewed-on: https://go-review.googlesource.com/14564
Reviewed-by: Chris Manghane <cmang@golang.org>
9 years agoruntime: unblock special glibc signals on each thread
Ian Lance Taylor [Fri, 4 Sep 2015 17:58:42 +0000 (10:58 -0700)]
runtime: unblock special glibc signals on each thread

Glibc uses some special signals for special thread operations.  These
signals will be used in programs that use cgo and invoke certain glibc
functions, such as setgid.  In order for this to work, these signals
need to not be masked by any thread.  Before this change, they were
being masked by programs that used os/signal.Notify, because it
carefully masks all non-thread-specific signals in all threads so that a
dedicated thread will collect and report those signals (see ensureSigM
in signal1_unix.go).

This change adds the two glibc special signals to the set of signals
that are unmasked in each thread.

Fixes #12498.

Change-Id: I797d71a099a2169c186f024185d44a2e1972d4ad
Reviewed-on: https://go-review.googlesource.com/14297
Reviewed-by: David Crawshaw <crawshaw@golang.org>
9 years agoruntime: minor clarifications of markroot
Austin Clements [Mon, 14 Sep 2015 18:28:09 +0000 (14:28 -0400)]
runtime: minor clarifications of markroot

This puts the _Root* indexes in a more friendly order and tweaks
markrootSpans to use a for-range loop instead of its own indexing.

Change-Id: I2c18d55c9a673ea396b6424d51ef4997a1a74825
Reviewed-on: https://go-review.googlesource.com/14548
Reviewed-by: Rick Hudson <rlh@golang.org>
9 years agoruntime: remove unused g.readyg field
Austin Clements [Mon, 3 Aug 2015 16:13:05 +0000 (12:13 -0400)]
runtime: remove unused g.readyg field

Commit 0e6a6c5 removed readyExecute a long time ago, but left behind
the g.readyg field that was used by readyExecute. Remove this now
unused field.

Change-Id: I41b87ad2b427974d256ec7a7f6d4bdc2ce8a13bb
Reviewed-on: https://go-review.googlesource.com/13111
Reviewed-by: Rick Hudson <rlh@golang.org>
9 years agoruntime: simplify mSpan_Sweep
Austin Clements [Tue, 21 Jul 2015 15:31:44 +0000 (11:31 -0400)]
runtime: simplify mSpan_Sweep

This is a cleanup following cc8f544, which was a minimal change to fix
issue #11617. This consolidates the two places in mSpan_Sweep that
update sweepgen. Previously this was necessary because sweepgen must
be updated before freeing the span, but we freed large spans early.
Now we free large spans later, so there's no need to duplicate the
sweepgen update. This also means large spans can take advantage of the
sweepgen sanity checking performed for other spans.

Change-Id: I23b79dbd9ec81d08575cd307cdc0fa6b20831768
Reviewed-on: https://go-review.googlesource.com/12451
Reviewed-by: Rick Hudson <rlh@golang.org>
9 years agoruntime: split marking of span roots into 128 subtasks
Austin Clements [Tue, 4 Aug 2015 14:45:29 +0000 (10:45 -0400)]
runtime: split marking of span roots into 128 subtasks

Marking of span roots can represent a significant fraction of the time
spent in mark termination. Simply traversing the span list takes about
1ms per GB of heap and if there are a large number of finalizers (for
example, for network connections), it may take much longer.

Improve the situation by splitting the span scan into 128 subtasks
that can be executed in parallel and load balanced by the markroots
parallel for. This lets the GC balance this job across the Ps.

A better solution is to do this during concurrent mark, or to improve
it algorithmically, but this is a simple change with a lot of bang for
the buck.

This was suggested by Rhys Hiltner.

Updates #11485.

Change-Id: I8b281adf0ba827064e154a1b6cc32d4d8031c03c
Reviewed-on: https://go-review.googlesource.com/13112
Reviewed-by: Keith Randall <khr@golang.org>
9 years agoruntime: fix hashing of trace stacks
Austin Clements [Thu, 13 Aug 2015 03:39:10 +0000 (23:39 -0400)]
runtime: fix hashing of trace stacks

The call to hash the trace stack reversed the "seed" and "size"
arguments to memhash and, hence, always called memhash with a 0 size,
which dutifully returned a hash value that depended only on the number
of PCs in the stack and not their values. As a result, all stacks were
put in to a very subset of the 8,192 buckets.

Fix this by passing these arguments in the correct order.

Change-Id: I67cd29312f5615c7ffa23e205008dd72c6b8af62
Reviewed-on: https://go-review.googlesource.com/13613
Reviewed-by: Dmitry Vyukov <dvyukov@google.com>
9 years agocmd/compiler/internal/gc: make Type.Copyto a []*Node
Håvard Haugen [Sun, 13 Sep 2015 20:58:12 +0000 (22:58 +0200)]
cmd/compiler/internal/gc: make Type.Copyto a []*Node

Passes go build -a -toolexec 'toolstash -cmp' std cmd

Change-Id: Ief4613cfb341172a85e3a894f44fb2bb308c7b55
Reviewed-on: https://go-review.googlesource.com/14554
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Dave Cheney <dave@cheney.net>
9 years agodatabase/sql: fix typo
Andrew Gerrand [Mon, 14 Sep 2015 00:18:22 +0000 (00:18 +0000)]
database/sql: fix typo

Fixes #12606

Change-Id: Ib68cb20108ad35c3dd96e606649c4c8f9c0f085c
Reviewed-on: https://go-review.googlesource.com/14571
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
9 years agocmd/compile/internal/gc: clean up errcmp
Håvard Haugen [Wed, 2 Sep 2015 18:25:46 +0000 (20:25 +0200)]
cmd/compile/internal/gc: clean up errcmp

Change-Id: Id07811a25bf4aa3ff834e7254a3dfb04522b2926
Reviewed-on: https://go-review.googlesource.com/14174
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Dave Cheney <dave@cheney.net>