]> Cypherpunks repositories - gostls13.git/log
gostls13.git
11 years agoruntime: update malloc comment for MSpan.needzero
Russ Cox [Thu, 13 Feb 2014 19:31:48 +0000 (14:31 -0500)]
runtime: update malloc comment for MSpan.needzero

Missed this suggestion in CL 57680046.

LGTM=iant
R=iant
CC=golang-codereviews
https://golang.org/cl/63390043

11 years agolib9: fix Solaris build
David du Colombier [Thu, 13 Feb 2014 19:30:42 +0000 (20:30 +0100)]
lib9: fix Solaris build

In file included from src/lib9/utf/utfecpy.c:17:0:
src/lib9/utf/utfdef.h:28:0: error: "nil" redefined [-Werror]
In file included from src/lib9/utf/utfrrune.c:17:0:
src/lib9/utf/utfdef.h:28:0: error: "nil" redefined [-Werror]

LGTM=bradfitz
R=golang-codereviews, bradfitz
CC=golang-codereviews
https://golang.org/cl/63410043

11 years agolib9: enable on Plan 9
David du Colombier [Thu, 13 Feb 2014 19:06:41 +0000 (20:06 +0100)]
lib9: enable on Plan 9

This change depends on CL 57170052.

LGTM=rsc
R=rsc, bradfitz
CC=golang-codereviews
https://golang.org/cl/60840049

11 years agolib9: fix inconsistencies and warnings on Plan 9
David du Colombier [Thu, 13 Feb 2014 19:05:55 +0000 (20:05 +0100)]
lib9: fix inconsistencies and warnings on Plan 9

warning: src/lib9/fmt/dorfmt.c:64 unreachable code RETURN
warning: src/lib9/fmt/fltfmt.c:184 set and not used: p
warning: src/lib9/utf/utflen.c:35 unreachable code RETURN
warning: src/lib9/utf/utfrrune.c:45 unreachable code RETURN
warning: src/lib9/utf/utfrune.c:44 unreachable code RETURN

LGTM=rsc
R=rsc, iant, gobot
CC=golang-codereviews
https://golang.org/cl/57170052

11 years agocmd/gc: for loop init statement misanalyzed by escape analysis
Daniel Morsing [Thu, 13 Feb 2014 19:04:43 +0000 (19:04 +0000)]
cmd/gc: for loop init statement misanalyzed by escape analysis

Logically, the init statement is in the enclosing scopes loopdepth, not inside the for loop.

Fixes #7313.

LGTM=rsc
R=golang-codereviews, gobot, rsc
CC=golang-codereviews
https://golang.org/cl/62430043

11 years agocmd/go, go/build: support .m files
Carlos Castillo [Thu, 13 Feb 2014 18:11:44 +0000 (10:11 -0800)]
cmd/go, go/build: support .m files

go/build is changed to list the .m files in a package, and match them for build constraints, adding them to a new field: Package.MFiles.

The go tool is changed to support building .m files and linking in the results during CGO and SWIG builds. This means packages that create a C interface to calls Objective-C code from go are now go-gettable without producing and distributing .syso files. This change is analogous to the one in Go 1.2 made to support C++ built code.

This change doesn't support .mm files (Objective C++).

Also added support for these MFiles to go list's -json mode.

Fixes #6536.

LGTM=iant
R=golang-codereviews, iant
CC=golang-codereviews
https://golang.org/cl/60590044

11 years agocrypto/x509: Add certificate signature request (CSR) support.
Kyle Isom [Thu, 13 Feb 2014 17:54:04 +0000 (12:54 -0500)]
crypto/x509: Add certificate signature request (CSR) support.

This change adds support for parsing and serialisation of PKCS #10,
certificate signature requests.

LGTM=agl
R=golang-codereviews, agl
CC=agl, golang-codereviews, nick
https://golang.org/cl/49830048

11 years agomath/big: add support for general encoding interfaces
Michael T. Jones [Thu, 13 Feb 2014 16:42:19 +0000 (08:42 -0800)]
math/big: add support for general encoding interfaces
TextMarshaller and TextUnmarshaller to ease transport of
unlimited precision rational numbers.

Fixes #7287.

Consists of encode and decode functions and two test
functions, one using JSON and one using XML. Each
verifies round trips for integers (rationals with
denominator == 1) and for fractional vaues.

LGTM=gri
R=gri, cookieo9, bradfitz, mtj
CC=golang-codereviews
https://golang.org/cl/61180043

11 years agocmd/gc: fix windows build
Dmitriy Vyukov [Thu, 13 Feb 2014 16:15:19 +0000 (20:15 +0400)]
cmd/gc: fix windows build
c:\src\go\pkg\obj\windows_amd64\libgc.a(lex.o): In function `catcher':
c:/src/go/src/cmd/gc/lex.c:181: undefined reference to `noted'

LGTM=0intro
R=0intro
CC=golang-codereviews
https://golang.org/cl/63270043

11 years agoruntime: introduce MSpan.needzero instead of writing to span data
Russ Cox [Thu, 13 Feb 2014 16:10:31 +0000 (11:10 -0500)]
runtime: introduce MSpan.needzero instead of writing to span data

This cleans up the code significantly, and it avoids any
possible problems with madvise zeroing out some but
not all of the data.

Fixes #6400.

LGTM=dave
R=dvyukov, dave
CC=golang-codereviews
https://golang.org/cl/57680046

11 years agoruntime: fix concurrent GC sweep
Dmitriy Vyukov [Thu, 13 Feb 2014 15:36:45 +0000 (19:36 +0400)]
runtime: fix concurrent GC sweep
The issue was that one of the MSpan_Sweep callers
was doing sweep with preemption enabled.
Additional checks are added.

LGTM=rsc
R=rsc, dave
CC=golang-codereviews
https://golang.org/cl/62990043

11 years agocmd/gc: catch notes on Plan 9
David du Colombier [Thu, 13 Feb 2014 15:35:51 +0000 (16:35 +0100)]
cmd/gc: catch notes on Plan 9

LGTM=rsc
R=rsc, jas, gobot
CC=ality, golang-codereviews
https://golang.org/cl/51650051

11 years agonet: only return unique hosts during hostname lookup on plan 9
Nicolas Owens [Thu, 13 Feb 2014 15:26:16 +0000 (10:26 -0500)]
net: only return unique hosts during hostname lookup on plan 9

TestLookupHost expects that no duplicate addresses are returned. when cs is consulted for a name, e.g net!localhost!1, it will possibly return multiple available paths, e.g. via il and tcp. this confuses the tests.

LGTM=aram
R=jas, 0intro, aram
CC=golang-codereviews
https://golang.org/cl/58120045

11 years agoarchive/tar: support extended attributes
Alexander Larsson [Thu, 13 Feb 2014 09:08:30 +0000 (20:08 +1100)]
archive/tar: support extended attributes

This adds support for archives with the SCHILY.xattr field in the
pax header. This is what gnu tar and star generate.
Fixes #7154.

LGTM=dsymonds
R=golang-codereviews, gobot, dsymonds
CC=golang-codereviews
https://golang.org/cl/54570043

11 years agocmd/6c, cmd/8c, cmd/8g: fix print of pc (which is vlong).
Shenghou Ma [Thu, 13 Feb 2014 08:09:03 +0000 (03:09 -0500)]
cmd/6c, cmd/8c, cmd/8g: fix print of pc (which is vlong).
While we're at it, fix a wrong for statement in cmd/8g.

LGTM=rsc
R=rsc, golang-codereviews
CC=golang-codereviews
https://golang.org/cl/62700044

11 years agocmd/gc: fix warnings on Plan 9
David du Colombier [Thu, 13 Feb 2014 06:28:22 +0000 (07:28 +0100)]
cmd/gc: fix warnings on Plan 9

warning: src/cmd/gc/popt.c:700 format mismatch d VLONG, arg 4
warning: src/cmd/gc/popt.c:700 format mismatch d VLONG, arg 5

LGTM=rsc
R=rsc
CC=golang-codereviews
https://golang.org/cl/62910043

11 years agoruntime/pprof: fix arm build after CL 61270043
Russ Cox [Thu, 13 Feb 2014 06:16:20 +0000 (01:16 -0500)]
runtime/pprof: fix arm build after CL 61270043

TBR=dvyukov
CC=golang-codereviews
https://golang.org/cl/62960043

11 years agocodereview: warn about 'hg upload' not uploading
Russ Cox [Thu, 13 Feb 2014 05:06:46 +0000 (05:06 +0000)]
codereview: warn about 'hg upload' not uploading

This seems to be what causes the 'chunk mismatch' errors.
I don't know why the server tells us there's nothing to upload,
but at the least we can warn that it did, so that the user
will try again.

LGTM=bradfitz
R=golang-codereviews, bradfitz
CC=golang-codereviews
https://golang.org/cl/62840043

11 years agoA+C: Mikhail Panchenko (individual CLA)
Andrew Gerrand [Thu, 13 Feb 2014 04:34:30 +0000 (15:34 +1100)]
A+C: Mikhail Panchenko (individual CLA)

Generated by addca.

R=gobot
CC=golang-codereviews
https://golang.org/cl/62890043

11 years agoA+C: Michael Pearson (individual CLA)
Andrew Gerrand [Thu, 13 Feb 2014 04:30:15 +0000 (15:30 +1100)]
A+C: Michael Pearson (individual CLA)

Generated by addca.

R=gobot
CC=golang-codereviews
https://golang.org/cl/62880043

11 years agoA+C: Marc Weistroff (individual CLA)
Andrew Gerrand [Thu, 13 Feb 2014 04:29:10 +0000 (15:29 +1100)]
A+C: Marc Weistroff (individual CLA)

Generated by addca.

R=gobot
CC=golang-codereviews
https://golang.org/cl/62870043

11 years agoA+C: Alexey Palazhchenko (individual CLA)
Andrew Gerrand [Thu, 13 Feb 2014 04:27:36 +0000 (15:27 +1100)]
A+C: Alexey Palazhchenko (individual CLA)

Generated by addca.

R=gobot
CC=golang-codereviews
https://golang.org/cl/62860043

11 years agocmd/5g: fix regopt bug in copyprop
Russ Cox [Thu, 13 Feb 2014 03:54:55 +0000 (03:54 +0000)]
cmd/5g: fix regopt bug in copyprop

copyau1 was assuming that it could deduce the type of the
middle register p->reg from the type of the left or right
argument: in CMPF F1, F2, the p->reg==2 must be a D_FREG
because p->from is F1, and in CMP R1, R2, the p->reg==2 must
be a D_REG because p->from is R1.

This heuristic fails for CMP $0, R2, which was causing copyau1
not to recognize p->reg==2 as a reference to R2, which was
keeping it from properly renaming the register use when
substituting registers.

cmd/5c has the right approach: look at the opcode p->as to
decide the kind of register. It is unclear where 5g's copyau1
came from; perhaps it was an attempt to avoid expanding 5c's
a2type to include new instructions used only by 5g.

Copy a2type from cmd/5c, expand to include additional instructions,
and make it crash the compiler if asked about an instruction
it does not understand (avoid silent bugs in the future if new
instructions are added).

Should fix current arm build breakage.

While we're here, fix the print statements dumping the pred and
succ info in the asm listing to pass an int arg to %.4ud
(Prog.pc is a vlong now, due to the liblink merge).

TBR=ken2
CC=golang-codereviews
https://golang.org/cl/62730043

11 years agocmd/5g: fix print format in peephole debugging
Anthony Martin [Thu, 13 Feb 2014 01:03:21 +0000 (17:03 -0800)]
cmd/5g: fix print format in peephole debugging

Fixes #7294.

LGTM=minux.ma, dave, bradfitz
R=golang-codereviews, minux.ma, dave, bradfitz
CC=golang-codereviews
https://golang.org/cl/61370043

11 years agodebug/macho: Add support for opening fat/universal binaries.
Robert Sesek [Thu, 13 Feb 2014 00:04:13 +0000 (11:04 +1100)]
debug/macho: Add support for opening fat/universal binaries.

New testdata was created from existing using:
$ lipo gcc-386-darwin-exec gcc-amd64-darwin-exec -create -output fat-gcc-386-amd64-darwin-exec

Fixes #7250.

LGTM=dave
R=golang-codereviews, dave, josharian, bradfitz
CC=golang-codereviews
https://golang.org/cl/60190043

11 years agodebug/macho: add test file for CL 60190043
Dave Cheney [Wed, 12 Feb 2014 23:58:23 +0000 (10:58 +1100)]
debug/macho: add test file for CL 60190043

From the description of CL 60190043

debug/macho: Add support for opening fat/universal binaries.

New testdata was created from existing using:
$ lipo gcc-386-darwin-exec gcc-amd64-darwin-exec -create -output
fat-gcc-386-amd64-darwin-exec

Update #7250

LGTM=iant
R=golang-codereviews, gobot, dsymonds, iant
CC=golang-codereviews
https://golang.org/cl/61720044

11 years agocrypto/md5: always test the portable block function too
Brad Fitzpatrick [Wed, 12 Feb 2014 21:31:05 +0000 (13:31 -0800)]
crypto/md5: always test the portable block function too

So it doesn't bitrot.

Like the sha1 version (https://golang.org/cl/62270043)

LGTM=agl
R=agl
CC=golang-codereviews
https://golang.org/cl/62420043

11 years agoruntime: fix non-concurrent sweep
Russ Cox [Wed, 12 Feb 2014 20:54:21 +0000 (15:54 -0500)]
runtime: fix non-concurrent sweep

State of the world:

CL 46430043 introduced a new concurrent sweep but is broken.

CL 62360043 made the new sweep non-concurrent
to try to fix the world while we understand what's wrong with
the concurrent version.

This CL fixes the non-concurrent form to run finalizers.
This CL is just a band-aid to get the build green again.

Dmitriy is working on understanding and then fixing what's
wrong with the concurrent sweep.

TBR=dvyukov
CC=golang-codereviews
https://golang.org/cl/62370043

11 years agoA+C: Kyle Isom (individual CLA)
Adam Langley [Wed, 12 Feb 2014 20:40:45 +0000 (15:40 -0500)]
A+C: Kyle Isom (individual CLA)

Generated by addca.

R=gobot
CC=golang-codereviews
https://golang.org/cl/57680047

11 years agoruntime: temporary disable concurrent GC sweep
Dmitriy Vyukov [Wed, 12 Feb 2014 20:03:27 +0000 (00:03 +0400)]
runtime: temporary disable concurrent GC sweep
We see failures on builders, e.g.:
http://build.golang.org/log/70bb28cd6bcf8c4f49810a011bb4337a61977bf4

LGTM=rsc, dave
R=rsc, dave
CC=golang-codereviews
https://golang.org/cl/62360043

11 years agocmd/cc, cmd/gc, cmd/ld: consolidate print format routines
Anthony Martin [Wed, 12 Feb 2014 19:29:11 +0000 (14:29 -0500)]
cmd/cc, cmd/gc, cmd/ld: consolidate print format routines

We now use the %A, %D, %P, and %R routines from liblink
across the board.

Fixes #7178.
Fixes #7055.

LGTM=iant
R=golang-codereviews, gobot, rsc, dave, iant, remyoudompheng
CC=golang-codereviews
https://golang.org/cl/49170043

11 years agocrypto/sha1: always test the portable block function too
Brad Fitzpatrick [Wed, 12 Feb 2014 19:27:36 +0000 (11:27 -0800)]
crypto/sha1: always test the portable block function too

So it doesn't bitrot.

LGTM=agl
R=golang-codereviews, agl
CC=golang-codereviews
https://golang.org/cl/62270043

11 years agoruntime: more precise mprof sampling
Dmitriy Vyukov [Wed, 12 Feb 2014 18:36:45 +0000 (22:36 +0400)]
runtime: more precise mprof sampling
Better sampling of objects that are close in size to sampling rate.
See the comment for details.

LGTM=rsc
R=golang-codereviews, rsc
CC=golang-codereviews
https://golang.org/cl/43830043

11 years agoruntime: improve cpu profiles for GC/syscalls/cgo
Dmitriy Vyukov [Wed, 12 Feb 2014 18:31:36 +0000 (22:31 +0400)]
runtime: improve cpu profiles for GC/syscalls/cgo
Current "System->etext" is not very informative.
Add parent "GC" frame.
Replace un-unwindable syscall/cgo frames with Go stack that leads to the call.

LGTM=rsc
R=rsc, alex.brainman, ality
CC=golang-codereviews
https://golang.org/cl/61270043

11 years agoAn ARM version of sha1block.go with a big improvement in throughput
Nick Craig-Wood [Wed, 12 Feb 2014 18:24:52 +0000 (13:24 -0500)]
An ARM version of sha1block.go with a big improvement in throughput
(up to 2.8x).

This is a partially unrolled version which performs better for small
hashes and only sacrifices a small amount of ultimate speed to a fully
unrolled version which uses 5k of code.

Code size

  Before 1636 bytes
  After 1880 bytes
  15% larger

Benchmarks on Samsung Exynos 5 ARMv7 Chromebook

benchmark              old ns/op    new ns/op    delta
BenchmarkHash8Bytes         1907         1136  -40.43%
BenchmarkHash1K            20280         7547  -62.79%
BenchmarkHash8K           148469        52576  -64.59%

benchmark               old MB/s     new MB/s  speedup
BenchmarkHash8Bytes         4.19         7.04    1.68x
BenchmarkHash1K            50.49       135.68    2.69x
BenchmarkHash8K            55.18       155.81    2.82x

LGTM=dave, agl
R=dave, bradfitz, agl, adg, nick
CC=golang-codereviews
https://golang.org/cl/56990044

11 years agoruntime: refactor level-triggered IO support
Dmitriy Vyukov [Wed, 12 Feb 2014 18:24:29 +0000 (22:24 +0400)]
runtime: refactor level-triggered IO support
Remove GOOS_solaris ifdef from netpoll code,
instead introduce runtime edge/level triggered IO flag.
Replace armread/armwrite with a single arm(mode) function,
that's how all other interfaces look like and these functions
will need to do roughly the same thing anyway.

LGTM=rsc
R=golang-codereviews, dave, rsc
CC=golang-codereviews
https://golang.org/cl/55500044

11 years agoruntime: refactor chan code
Dmitriy Vyukov [Wed, 12 Feb 2014 18:21:38 +0000 (22:21 +0400)]
runtime: refactor chan code
1. Make internal chan functions static.
2. Move selgen local variable instead of a member of G struct.
3. Change "bool *pres/selected" parameter of chansend/chanrecv to "bool block",
   which is simpler, faster and less code.
-37 lines total.

LGTM=rsc
R=golang-codereviews, dave, gobot, rsc
CC=bradfitz, golang-codereviews, iant, khr
https://golang.org/cl/58610043

11 years agoruntime: concurrent GC sweep
Dmitriy Vyukov [Wed, 12 Feb 2014 18:16:42 +0000 (22:16 +0400)]
runtime: concurrent GC sweep
Moves sweep phase out of stoptheworld by adding
background sweeper goroutine and lazy on-demand sweeping.

It turned out to be somewhat trickier than I expected,
because there is no point in time when we know size of live heap
nor consistent number of mallocs and frees.
So everything related to next_gc, mprof, memstats, etc becomes trickier.

At the end of GC next_gc is conservatively set to heap_alloc*GOGC,
which is much larger than real value. But after every sweep
next_gc is decremented by freed*GOGC. So when everything is swept
next_gc becomes what it should be.

For mprof I had to introduce 3-generation scheme (allocs, revent_allocs, prev_allocs),
because by the end of GC we know number of frees for the *previous* GC.

Significant caution is required to not cross yet-unknown real value of next_gc.
This is achieved by 2 means:
1. Whenever I allocate a span from MCentral, I sweep a span in that MCentral.
2. Whenever I allocate N pages from MHeap, I sweep until at least N pages are
returned to heap.
This provides quite strong guarantees that heap does not grow when it should now.

http-1
allocated                    7036         7033      -0.04%
allocs                         60           60      +0.00%
cputime                     51050        46700      -8.52%
gc-pause-one             34060569      1777993     -94.78%
gc-pause-total               2554          133     -94.79%
latency-50                 178448       170926      -4.22%
latency-95                 284350       198294     -30.26%
latency-99                 345191       220652     -36.08%
rss                     101564416    101007360      -0.55%
sys-gc                    6606832      6541296      -0.99%
sys-heap                 88801280     87752704      -1.18%
sys-other                 7334208      7405928      +0.98%
sys-stack                  524288       524288      +0.00%
sys-total               103266608    102224216      -1.01%
time                        50339        46533      -7.56%
virtual-mem             292990976    293728256      +0.25%

garbage-1
allocated                 2983818      2990889      +0.24%
allocs                      62880        62902      +0.03%
cputime                  16480000     16190000      -1.76%
gc-pause-one            828462467    487875135     -41.11%
gc-pause-total            4142312      2439375     -41.11%
rss                    1151709184   1153712128      +0.17%
sys-gc                   66068352     66068352      +0.00%
sys-heap               1039728640   1039728640      +0.00%
sys-other                37776064     40770176      +7.93%
sys-stack                 8781824      8781824      +0.00%
sys-total              1152354880   1155348992      +0.26%
time                     16496998     16199876      -1.80%
virtual-mem            1409564672   1402281984      -0.52%

LGTM=rsc
R=golang-codereviews, sameer, rsc, iant, jeremyjackins, gobot
CC=golang-codereviews, khr
https://golang.org/cl/46430043

11 years agoencoding/json: fix test failure
Dmitriy Vyukov [Wed, 12 Feb 2014 17:50:58 +0000 (21:50 +0400)]
encoding/json: fix test failure
$ go test -cpu=1,1,1,1,1,1,1,1,1 encoding/json
--- FAIL: TestIndentBig (0.00 seconds)
        scanner_test.go:131: Indent(jsonBig) did not get bigger
On 4-th run initBig generates an empty array.

LGTM=bradfitz
R=golang-codereviews, bradfitz
CC=golang-codereviews
https://golang.org/cl/49930051

11 years agocrypto/subtle: panic if slices of different lengths are passed to ConstantTimeCompare.
Adam Langley [Wed, 12 Feb 2014 16:58:48 +0000 (11:58 -0500)]
crypto/subtle: panic if slices of different lengths are passed to ConstantTimeCompare.

ConstantTimeCompare has always been documented to take equal length
slices but perhaps this is too subtle, even for 'subtle'.

Fixes #7304.

LGTM=hanwen, bradfitz
R=golang-codereviews, hanwen, bradfitz
CC=golang-codereviews
https://golang.org/cl/62190043

11 years agocrypto/tls: better error messages.
Adam Langley [Wed, 12 Feb 2014 16:20:01 +0000 (11:20 -0500)]
crypto/tls: better error messages.

LGTM=bradfitz
R=golang-codereviews, bradfitz
CC=golang-codereviews
https://golang.org/cl/60580046

11 years agonet/http: make responseAndError satisfy the net.Error interface
Rick Arnold [Wed, 12 Feb 2014 15:59:58 +0000 (07:59 -0800)]
net/http: make responseAndError satisfy the net.Error interface

Allow clients to check for timeouts without relying on error substring
matching.

Fixes #6185.

LGTM=bradfitz
R=golang-codereviews, bradfitz
CC=golang-codereviews
https://golang.org/cl/55470048

11 years agodebug/pe: delete unnecessary type conversions
Robert Dinu [Wed, 12 Feb 2014 15:35:54 +0000 (07:35 -0800)]
debug/pe: delete unnecessary type conversions

Fixes #7104.

LGTM=iant
R=golang-dev, iant
CC=golang-codereviews
https://golang.org/cl/61480049

11 years agoA+C: Alexander Larsson (individual CLA)
David Symonds [Wed, 12 Feb 2014 01:34:04 +0000 (12:34 +1100)]
A+C: Alexander Larsson (individual CLA)

Generated by addca.

R=gobot
CC=golang-codereviews
https://golang.org/cl/53410046

11 years agogo/parser: better error messages for if/switch/for conditions/expressions
Robert Griesemer [Wed, 12 Feb 2014 00:45:31 +0000 (16:45 -0800)]
go/parser: better error messages for if/switch/for conditions/expressions

Fixes #7102.

LGTM=adonovan
R=adonovan
CC=golang-codereviews
https://golang.org/cl/56770045

11 years agoapi: update next.txt
Shenghou Ma [Wed, 12 Feb 2014 00:37:36 +0000 (19:37 -0500)]
api: update next.txt

LGTM=bradfitz
R=golang-codereviews, bradfitz
CC=golang-codereviews
https://golang.org/cl/54830045

11 years agocmd/ld, cmd/6l: part 2 of solaris/amd64 linker changes.
Shenghou Ma [Tue, 11 Feb 2014 23:43:05 +0000 (18:43 -0500)]
cmd/ld, cmd/6l: part 2 of solaris/amd64 linker changes.
Second part of the solaris/amd64 linker changes: relocation and symbol table.

LGTM=iant
R=golang-codereviews, iant
CC=golang-codereviews
https://golang.org/cl/61330043

11 years agogo/parser: check presence of 2nd and 3rd index in 3-index slice
Robert Griesemer [Tue, 11 Feb 2014 21:40:37 +0000 (13:40 -0800)]
go/parser: check presence of 2nd and 3rd index in 3-index slice

Fixes #7305.

LGTM=adonovan
R=bradfitz, adonovan
CC=golang-codereviews
https://golang.org/cl/58950045

11 years agocmd/6g, cmd/8g, cmd/5g: make the undefined instruction have no successors
Daniel Morsing [Tue, 11 Feb 2014 20:25:40 +0000 (20:25 +0000)]
cmd/6g, cmd/8g, cmd/5g: make the undefined instruction have no successors

The UNDEF instruction was listed in the instruction data as having the next instruction in the stream as its successor. This confused the optimizer into adding a load where it wasn't needed, in turn confusing the liveness analysis pass for GC bitmaps into thinking that the variable was live.

Fixes #7229.

LGTM=iant, rsc
R=golang-codereviews, bradfitz, iant, dave, rsc
CC=golang-codereviews
https://golang.org/cl/56910045

11 years agoarchive/zip: re-use flate.Writers when writing compressed files
Brad Fitzpatrick [Tue, 11 Feb 2014 19:41:25 +0000 (11:41 -0800)]
archive/zip: re-use flate.Writers when writing compressed files

Prevents a ton of garbage. (Noticed this when writing large
Camlistore zip archives to Amazon Glacier)

Note that the Closer part of the io.WriteCloser is never given
to users. It's an internal detail of the package.

benchmark                         old ns/op     new ns/op     delta
BenchmarkCompressedZipGarbage     42884123      40732373      -5.02%

benchmark                         old allocs     new allocs     delta
BenchmarkCompressedZipGarbage     204            149            -26.96%

benchmark                         old bytes     new bytes     delta
BenchmarkCompressedZipGarbage     4397576       66744         -98.48%

LGTM=adg, rsc
R=adg, rsc
CC=golang-codereviews
https://golang.org/cl/54300053

11 years agoruntime: do not profile blocked netpoll on windows
Dmitriy Vyukov [Tue, 11 Feb 2014 09:41:46 +0000 (13:41 +0400)]
runtime: do not profile blocked netpoll on windows
There is frequently a thread hanging on GQCS,
currently it skews profiles towards netpoll,
but it is not bad and is not consuming any resources.

R=alex.brainman
CC=golang-codereviews
https://golang.org/cl/61560043

11 years agoruntime: homogenize panic strings on Plan 9
David du Colombier [Tue, 11 Feb 2014 08:34:43 +0000 (09:34 +0100)]
runtime: homogenize panic strings on Plan 9

LGTM=rsc
R=rsc
CC=golang-codereviews
https://golang.org/cl/61410046

11 years agoA+C: Péter Szilágyi (individual CLA)
Andrew Gerrand [Tue, 11 Feb 2014 06:29:40 +0000 (17:29 +1100)]
A+C: Péter Szilágyi (individual CLA)

Generated by addca.

R=gobot
CC=golang-codereviews
https://golang.org/cl/61410050

11 years agoarchive/zip: use correct test, fix 32-bit build
Andrew Gerrand [Tue, 11 Feb 2014 05:27:14 +0000 (16:27 +1100)]
archive/zip: use correct test, fix 32-bit build

LGTM=dsymonds
R=dsymonds
CC=golang-codereviews
https://golang.org/cl/61070047

11 years agoarchive/zip: actually test uncompressed size
Andrew Gerrand [Tue, 11 Feb 2014 05:09:42 +0000 (16:09 +1100)]
archive/zip: actually test uncompressed size

Fixes #7292.

LGTM=dsymonds
R=dsymonds
CC=golang-codereviews
https://golang.org/cl/61650046

11 years agodoc: remove "Code that grows with grace" video from front page
Andrew Gerrand [Tue, 11 Feb 2014 05:00:44 +0000 (16:00 +1100)]
doc: remove "Code that grows with grace" video from front page

It was deleted from Vimeo. I will attempt to get a new version online.

Fixes #7297.

LGTM=r
R=r
CC=golang-codereviews
https://golang.org/cl/61740043

11 years agocmd/gc: change compile error to "use of package %S without selector"
Brendan Daniel Tracey [Tue, 11 Feb 2014 01:27:31 +0000 (20:27 -0500)]
cmd/gc: change compile error to "use of package %S without selector"

At present, when a package identifier is used outside of a selector expression, gc gives the error "use of package %S outside selector". However, in the selector expression x.f, the spec defines f as the selector. This change makes the error clearer.

Fixes #7133.

LGTM=rsc
R=golang-codereviews, rsc
CC=golang-codereviews
https://golang.org/cl/50060047

11 years agoA+C: Brendan Daniel Tracey (individual CLA)
Russ Cox [Tue, 11 Feb 2014 01:27:21 +0000 (20:27 -0500)]
A+C: Brendan Daniel Tracey (individual CLA)

Generated by addca.

R=gobot
CC=golang-codereviews
https://golang.org/cl/55870046

11 years agodoc: restore race detector manual
Andrew Gerrand [Mon, 10 Feb 2014 22:26:34 +0000 (09:26 +1100)]
doc: restore race detector manual

This got deleted mistakenly during the content re-org.

Fixes #7275.

LGTM=dvyukov
R=golang-codereviews, minux.ma, dvyukov
CC=golang-codereviews
https://golang.org/cl/60850043

11 years agocontainer/heap: avoid and/or ambiguity in documentation
Robert Griesemer [Mon, 10 Feb 2014 20:48:56 +0000 (12:48 -0800)]
container/heap: avoid and/or ambiguity in documentation

(per suggestion by Doug McIlroy)

LGTM=r
R=r
CC=golang-codereviews
https://golang.org/cl/50580046

11 years agoruntime: handle "sys: trap: divide error" note on Plan 9
David du Colombier [Mon, 10 Feb 2014 20:47:52 +0000 (21:47 +0100)]
runtime: handle "sys: trap: divide error" note on Plan 9

Fixes #7286.

LGTM=rsc
R=rsc
CC=golang-codereviews
https://golang.org/cl/61410044

11 years agolib/codereview: allow addca CLs without LGTM
Russ Cox [Mon, 10 Feb 2014 19:40:19 +0000 (14:40 -0500)]
lib/codereview: allow addca CLs without LGTM

LGTM=gri
R=golang-codereviews, gri
CC=golang-codereviews
https://golang.org/cl/61610043

11 years agocmd/go: fix comment
Dmitriy Vyukov [Mon, 10 Feb 2014 17:48:43 +0000 (21:48 +0400)]
cmd/go: fix comment

R=r
CC=golang-codereviews
https://golang.org/cl/61650044

11 years agoruntime: fix crash during cpu profiling
Dmitriy Vyukov [Mon, 10 Feb 2014 16:24:47 +0000 (20:24 +0400)]
runtime: fix crash during cpu profiling
mp->mcache can be concurrently modified by runtime·helpgc.
In such case sigprof can remember mcache=nil, then helpgc sets it to non-nil,
then sigprof restores it back to nil, GC crashes with nil mcache.

R=rsc
CC=golang-codereviews
https://golang.org/cl/58860044

11 years agocmd/go: better advice for heap profiling
Dmitriy Vyukov [Mon, 10 Feb 2014 14:59:07 +0000 (18:59 +0400)]
cmd/go: better advice for heap profiling

R=golang-codereviews, bradfitz
CC=golang-codereviews
https://golang.org/cl/60760048

11 years agoruntime: do not cpu profile idle threads on windows
Dmitriy Vyukov [Mon, 10 Feb 2014 11:40:55 +0000 (15:40 +0400)]
runtime: do not cpu profile idle threads on windows
Currently this leads to a significant skew towards 'etext' entry,
since all idle threads are profiled every tick.
Before:
Total: 66608 samples
   63188  94.9%  94.9%    63188  94.9% etext
     278   0.4%  95.3%      278   0.4% sweepspan
     216   0.3%  95.6%      448   0.7% runtime.mallocgc
     122   0.2%  95.8%      122   0.2% scanblock
     113   0.2%  96.0%      113   0.2% net/textproto.canonicalMIMEHeaderKey
After:
Total: 8008 samples
    3949  49.3%  49.3%     3949  49.3% etext
     231   2.9%  52.2%      231   2.9% scanblock
     211   2.6%  54.8%      211   2.6% runtime.cas64
     182   2.3%  57.1%      408   5.1% runtime.mallocgc
     178   2.2%  59.3%      178   2.2% runtime.atomicload64

LGTM=alex.brainman
R=golang-codereviews, alex.brainman
CC=golang-codereviews
https://golang.org/cl/61250043

11 years agomisc/pprof: support web profiles on windows
Dmitriy Vyukov [Mon, 10 Feb 2014 09:52:58 +0000 (13:52 +0400)]
misc/pprof: support web profiles on windows

LGTM=bradfitz
R=golang-codereviews, bradfitz, alex.brainman
CC=golang-codereviews
https://golang.org/cl/61260044

11 years agocmd/go: skip $ORIGIN test on darwin systems
Dave Cheney [Mon, 10 Feb 2014 02:35:39 +0000 (13:35 +1100)]
cmd/go: skip $ORIGIN test on darwin systems

Fixes #7293.

Update #7261

The bsd ld(1) does not understand $ORIGIN and has restrictions on using -rpath when using clang(1), the default compiler on darwin.

LGTM=iant
R=iant
CC=golang-codereviews
https://golang.org/cl/58480045

11 years agocmd/go: fix test.bash
Dave Cheney [Sun, 9 Feb 2014 23:41:47 +0000 (10:41 +1100)]
cmd/go: fix test.bash

Fixes #7260.

Fix three broken tests in test.bash

The test for issue 4568 was confused by go $ACTION . producing a package root of "", avoiding this mode fixes the test but weakens the test.

The test for issue 4773 was broken on linux because math/Rand would fail to resolve as a package causing the test for duplicates to be skipped.

Finally, the last breakage was a small change in the error message.

Also, add test for foldDup.

LGTM=iant
R=iant, rsc
CC=golang-codereviews
https://golang.org/cl/61070044

11 years agoarchive/zip: add flate writing benchmark
Brad Fitzpatrick [Sun, 9 Feb 2014 21:56:47 +0000 (13:56 -0800)]
archive/zip: add flate writing benchmark

LGTM=adg
R=adg
CC=golang-codereviews
https://golang.org/cl/60530049

11 years agoinclude, linlink, cmd/6l, cmd/ld: part 1 of solaris/amd64 linker changes.
Shenghou Ma [Sun, 9 Feb 2014 21:45:38 +0000 (16:45 -0500)]
include, linlink, cmd/6l, cmd/ld: part 1 of solaris/amd64 linker changes.
rsc suggested that we split the whole linker changes into three parts.
This is the first one, mostly dealing with adding Hsolaris.

LGTM=iant
R=golang-codereviews, iant, dave
CC=golang-codereviews
https://golang.org/cl/54210050

11 years agoC: add Robert Sesek (Google CLA)
Andrew Gerrand [Sun, 9 Feb 2014 21:25:47 +0000 (08:25 +1100)]
C: add Robert Sesek (Google CLA)

LGTM=bradfitz
R=dave, bradfitz
CC=golang-codereviews
https://golang.org/cl/61310043

11 years agosyscall: add missing include statement to bootstrap error code and signal generator
Mikio Hara [Sun, 9 Feb 2014 08:20:59 +0000 (17:20 +0900)]
syscall: add missing include statement to bootstrap error code and signal generator

LGTM=iant
R=golang-codereviews, iant
CC=golang-codereviews
https://golang.org/cl/54300054

11 years agocmd/6g: faster memmove/memset-like code using unaligned load/stores.
Rémy Oudompheng [Fri, 7 Feb 2014 22:58:21 +0000 (23:58 +0100)]
cmd/6g: faster memmove/memset-like code using unaligned load/stores.

This changes makes sgen and clearfat use unaligned instructions for
the trailing bytes, like the runtime memmove does, resulting in faster
code when manipulating types whose size is not a multiple of 8.

LGTM=khr
R=khr, iant, rsc
CC=golang-codereviews
https://golang.org/cl/51740044

11 years agocmd/gc: revert CL 60740044
David du Colombier [Fri, 7 Feb 2014 16:05:16 +0000 (17:05 +0100)]
cmd/gc: revert CL 60740044

LGTM=iant
R=iant
CC=golang-codereviews
https://golang.org/cl/60590047

11 years agocmd/gc: fix nil pointer dereference
David du Colombier [Fri, 7 Feb 2014 14:43:40 +0000 (15:43 +0100)]
cmd/gc: fix nil pointer dereference

LGTM=iant
R=golang-codereviews, dave, iant
CC=golang-codereviews
https://golang.org/cl/60740044

11 years agonet: remove superfluous type conversion
Alex Brainman [Fri, 7 Feb 2014 05:58:45 +0000 (16:58 +1100)]
net: remove superfluous type conversion

LGTM=minux.ma, dvyukov
R=golang-codereviews, minux.ma, dvyukov
CC=golang-codereviews
https://golang.org/cl/60900043

11 years agoruntime: faster memclr on x86.
Keith Randall [Fri, 7 Feb 2014 01:43:22 +0000 (17:43 -0800)]
runtime: faster memclr on x86.

Use explicit SSE writes instead of REP STOSQ.

benchmark               old ns/op    new ns/op    delta
BenchmarkMemclr5               22            5  -73.62%
BenchmarkMemclr16              27            5  -78.49%
BenchmarkMemclr64              28            6  -76.43%
BenchmarkMemclr256             34            8  -74.94%
BenchmarkMemclr4096           112           84  -24.73%
BenchmarkMemclr65536         1902         1920   +0.95%

LGTM=dvyukov
R=golang-codereviews, dvyukov
CC=golang-codereviews
https://golang.org/cl/60090044

11 years agoos/exec: use filepath.Base in Command
Alex Brainman [Fri, 7 Feb 2014 01:30:30 +0000 (12:30 +1100)]
os/exec: use filepath.Base in Command

filepath.Base covers all scenarios
(for example paths like d:hello.txt)
on windows

LGTM=iant, bradfitz
R=golang-codereviews, iant, bradfitz
CC=golang-codereviews
https://golang.org/cl/59740050

11 years agosyscall: make use of signed char explicit in generating z-files on freebsd/arm
Mikio Hara [Fri, 7 Feb 2014 01:23:53 +0000 (10:23 +0900)]
syscall: make use of signed char explicit in generating z-files on freebsd/arm

This CL is in preparation to make cgo work on freebsd/arm.

The signedness of C char might be a problem when we make bare syscall
APIs, Go structures, using built-in bootstrap scripts with cgo because
they do translate C stuff to Go stuff internally. For now almost all
the C compilers assume that the type of char will be unsigned on arm
by default but it makes a different view from amd64, 386.

This CL just passes -fsigned-char, let the type of char be signed,
option which is supported on both gcc and clang to the underlying C
compilers through cgo for avoiding such inconsistency on syscall API.

LGTM=iant
R=iant
CC=golang-codereviews
https://golang.org/cl/59740051

11 years agosyscall: regenerate EABI call convention compliant syscalls for freebsd/arm
Mikio Hara [Fri, 7 Feb 2014 01:23:26 +0000 (10:23 +0900)]
syscall: regenerate EABI call convention compliant syscalls for freebsd/arm

This CL is in preparation to make cgo work on freebsd/arm.

LGTM=iant
R=iant
CC=golang-codereviews
https://golang.org/cl/59490051

11 years agosyscall: fix build on freebsd/arm
Mikio Hara [Fri, 7 Feb 2014 01:23:02 +0000 (10:23 +0900)]
syscall: fix build on freebsd/arm

This CL is in preparation to make cgo work on freebsd/arm.

It's just for fixing build fails on freebsd/arm, we still need to
update z-files later for fixing several package test fails.

How to generate z-files on freebsd/arm in the bootstrapping phase:
1. run freebsd on appropriate arm-eabi platforms
2. both syscall z-files and runtime def-files in the current tree are
   broken about EABI padding, fix them by hand
3. run make.bash again to build $GOTOOLDIR/cgo
4. use $GOTOOLDIR/cgo directly

LGTM=iant
R=iant, dave
CC=golang-codereviews
https://golang.org/cl/59490052

11 years agoruntime/cgo: fix build on freebsd/arm
Mikio Hara [Fri, 7 Feb 2014 01:22:34 +0000 (10:22 +0900)]
runtime/cgo: fix build on freebsd/arm

This CL is in preparation to make cgo work on freebsd/arm.

LGTM=iant
R=iant
CC=golang-codereviews
https://golang.org/cl/60500044

11 years agoruntime: fix build on freebsd/arm
Mikio Hara [Fri, 7 Feb 2014 01:22:13 +0000 (10:22 +0900)]
runtime: fix build on freebsd/arm

This CL is in preparation to make cgo work on freebsd/arm.

How to generate defs-files on freebsd/arm in the bootstrapping phase:
1. run freebsd on appropriate arm-eabi platforms
2. both syscall z-files and runtime def-files in the current tree are
   broken about EABI padding, fix them by hand
3. run make.bash again to build $GOTOOLDIR/cgo
4. use $GOTOOLDIR/cgo directly

LGTM=minux.ma, iant
R=iant, minux.ma, dave
CC=golang-codereviews
https://golang.org/cl/59580045

11 years agonet/mail: correctly handle whitespaces when formatting an email address
Jakub Ryszard Czarnowicz [Thu, 6 Feb 2014 23:49:10 +0000 (10:49 +1100)]
net/mail: correctly handle whitespaces when formatting an email address

Whitespace characters are allowed in quoted-string according to RFC 5322 without
being "Q"-encoding. Address.String() already always formats the name portion in
quoted string, so whitespace characters should be allowed in there.

Fixes #6641.

LGTM=dave, dsymonds
R=golang-codereviews, gobot, dsymonds, dave
CC=golang-codereviews
https://golang.org/cl/55770043

11 years agocmd/go: do not create world-writable files.
Shenghou Ma [Thu, 6 Feb 2014 22:54:33 +0000 (17:54 -0500)]
cmd/go: do not create world-writable files.
Fixes #7240.

LGTM=tommi.virtanen, bradfitz
R=golang-codereviews, tommi.virtanen, bradfitz
CC=golang-codereviews
https://golang.org/cl/60480045

11 years agomath/big: replace goto with for loop
Richard Musiol [Thu, 6 Feb 2014 22:44:30 +0000 (14:44 -0800)]
math/big: replace goto with for loop
I just added support for goto statements to my GopherJS project and now I am trying to get rid of my patches. These occurrences of goto however are a bit problematic:
GopherJS has to emulate gotos, so there is some performance drawback when doing so. In this case the drawback is major, since this is a core function of math/big which is called quite often. Additionally I can't see any reason here why the implementation with gotos should be preferred over my proposal.
That's why I would kindly ask to include this patch, even though it is functional equivalent to the existing code.

LGTM=bradfitz
R=golang-codereviews, bradfitz
CC=golang-codereviews
https://golang.org/cl/55470046

11 years agodoc/install.html: get rid of explicit version number in docs.
Shenghou Ma [Thu, 6 Feb 2014 21:09:01 +0000 (16:09 -0500)]
doc/install.html: get rid of explicit version number in docs.
So that we don't need to update the file each time a new version is released.

LGTM=dave, r, adg
R=golang-codereviews, dave, josharian, gobot, adg, r
CC=golang-codereviews
https://golang.org/cl/48730043

11 years agocmd/go, cmd/cgo, make.bash: cross compiling with cgo enabled
Elias Naur [Thu, 6 Feb 2014 17:11:00 +0000 (09:11 -0800)]
cmd/go, cmd/cgo, make.bash: cross compiling with cgo enabled

Introduce two new environment variables, CC_FOR_TARGET and CXX_FOR_TARGET.
CC_FOR_TARGET defaults to CC and is used when compiling for GOARCH, while
CC remains for compiling for GOHOSTARCH.
CXX_FOR_TARGET defaults to CXX and is used when compiling C++ code for
GOARCH.

CGO_ENABLED defaults to disabled when cross compiling and has to be
explicitly enabled.

Update #4714

LGTM=minux.ma, iant
R=golang-codereviews, minux.ma, iant, rsc, dominik.honnef
CC=golang-codereviews
https://golang.org/cl/57100043

11 years agoA+C: Denis Brandolini (individual CLA)
Brad Fitzpatrick [Wed, 5 Feb 2014 15:32:16 +0000 (07:32 -0800)]
A+C: Denis Brandolini (individual CLA)

Generated by addca.

TBR=iant
R=gobot
CC=golang-codereviews
https://golang.org/cl/52350044

11 years agoencoding/json: mention escaping of '&'
Shenghou Ma [Wed, 5 Feb 2014 06:24:51 +0000 (01:24 -0500)]
encoding/json: mention escaping of '&'
Fixes #7034.

LGTM=iant
R=golang-codereviews, iant
CC=golang-codereviews
https://golang.org/cl/57140043

11 years agocmd/go, doc/go1.3.txt: misc/benchcmp has been replaced by go tool benchcmp
Josh Bleecher Snyder [Tue, 4 Feb 2014 19:53:13 +0000 (11:53 -0800)]
cmd/go, doc/go1.3.txt: misc/benchcmp has been replaced by go tool benchcmp

Fixes #7016.

LGTM=r
R=r
CC=adg, bradfitz, dave, golang-codereviews
https://golang.org/cl/60100043

11 years agotext/tabwriter: improve panic backtraces
Josh Bleecher Snyder [Tue, 4 Feb 2014 18:19:02 +0000 (10:19 -0800)]
text/tabwriter: improve panic backtraces

Fixes #7117.

LGTM=gri
R=golang-codereviews, gobot, gri
CC=golang-codereviews
https://golang.org/cl/53310044

11 years agocrypto/tls: do not send the current time in hello messages
Anthony Martin [Tue, 4 Feb 2014 15:51:37 +0000 (10:51 -0500)]
crypto/tls: do not send the current time in hello messages

This reduces the ability to fingerprint TLS connections.

The impeteus for this change was a recent change to OpenSSL
by Nick Mathewson:

http://git.openssl.org/gitweb/?p=openssl.git;a=commit;h=2016265dfb

LGTM=agl
R=agl
CC=golang-codereviews
https://golang.org/cl/57230043

11 years agoruntime: add more chan benchmarks
Dmitriy Vyukov [Tue, 4 Feb 2014 05:41:48 +0000 (09:41 +0400)]
runtime: add more chan benchmarks
Add benchmarks for:
1. non-blocking failing receive (polling of "stop" chan)
2. channel-based semaphore (gate pattern)
3. select-based producer/consumer (pass data through a channel, but also wait on "stop" and "timeout" channels)

LGTM=r
R=golang-codereviews, r
CC=bradfitz, golang-codereviews, iant, khr
https://golang.org/cl/59040043

11 years agoliblink, cmd/5l: restore flag_shared
Elias Naur [Mon, 3 Feb 2014 22:49:57 +0000 (14:49 -0800)]
liblink, cmd/5l: restore flag_shared

CL 56120043 fixed and cleaned up TLS on ARM after introducing liblink, but
left flag_shared broken. This CL restores the (unsupported) flag_shared
behaviour by simply rewriting access to $runtime.tlsgm(SB) with
runtime.tlsgm(SB), to compensate for the extra indirection when going from
the R_ARM_TLS_LE32 relocation to the R_ARM_TLS_IE32 relocation.

Also, remove unnecessary symbol lookup left after 56120043.

LGTM=iant
R=iant, rsc
CC=golang-codereviews
https://golang.org/cl/57000043

11 years agoliblink, cmd/5a, cmd/5l: restore cgo on older ARM processors
Elias Naur [Mon, 3 Feb 2014 22:07:54 +0000 (14:07 -0800)]
liblink, cmd/5a, cmd/5l: restore cgo on older ARM processors

CL 56120043 fixed TLS handling on ARM after the introduction of
liblink but left older ARM processors broken.

Before liblink, the MRC instruction was replaced with a fallback
on older ARMs. CL 56120043 removed that, because the rewrite matched
bit patterns on the AWORD pseudo-instruction and could therefore change
unrelated AWORDs that happened to match.

This CL adds an AMRC instruction to encode both MRC and MCR previously
encoded as AWORDs. Then, in liblink, the AMRC instructions are either
rewritten to AWORD, or, on goarm < 7, replaced with a branch to the
fallback.

./all.bash completes successfully on an ARMv7 with either GOARM=7 or
GOARM=5. I have verified that the fallback is indeed present in both
runtime.save_gm and runtime.load_gm when GOARM=5 but not when GOARM=7.

If all goes well, this should fix the armv5 builders.

LGTM=iant
R=iant, rsc
CC=golang-codereviews
https://golang.org/cl/55540044

11 years agoos/exec: fix Command with relative paths
Brad Fitzpatrick [Mon, 3 Feb 2014 21:32:13 +0000 (16:32 -0500)]
os/exec: fix Command with relative paths

Command was (and is) documented like:
"If name contains no path separators, Command uses LookPath to
resolve the path to a complete name if possible. Otherwise it
uses name directly."

But that wasn't true. It always did LookPath, and then
set a sticky error that the user couldn't unset.
And then if cmd.Dir was changed, Start would still fail
due to the earlier sticky error being set.

This keeps LookPath in the same place as before (so no user
visible changes in cmd.Path after Command), but only does
it when the documentation says it will happen.

Also, clarify the docs about a relative Dir path.

No change in any existing behavior, except using Command
is now possible with relative paths. Previously it only
worked if you built the *Cmd by hand.

Fixes #7228

LGTM=iant
R=iant
CC=adg, golang-codereviews
https://golang.org/cl/59580044

11 years agonet/http: make a test fail harder, earlier
Brad Fitzpatrick [Mon, 3 Feb 2014 21:01:58 +0000 (16:01 -0500)]
net/http: make a test fail harder, earlier

LGTM=dave
R==r, r, dave
CC=golang-codereviews
https://golang.org/cl/59810043

11 years agoA+C: Gordon Klaus (individual CLA)
Robert Griesemer [Mon, 3 Feb 2014 19:05:51 +0000 (11:05 -0800)]
A+C: Gordon Klaus (individual CLA)

Generated by addca.

LGTM=iant
R=gobot, iant
CC=golang-codereviews
https://golang.org/cl/59770043