Escape analysis was incorrectly assuming that
functions without bodies don't leak their
parameters. This meant that sync/atomic's
TestAddInt64 was allocating x on its stack,
and then x was not properly aligned for use
with the atomic 64-bit instructions. Obviously
we should figure out the alignment story on 5g
too, but this fix is correct and should restore the
build to 'ok'.
TBR=lvd
CC=golang-dev
https://golang.org/cl/4964047
Also standardize on 'utf8' as encoding name.
Apparently either is acceptable.
The user, because it is a StringProperty,
must be of type unicode in order to handle
Unicode correctly. It must *not* have type string.
The desc, because it is a BlobProperty, must
be of type string in order to handle Unicode correctly.
It must *not* have type unicode.
Yay encoding type pedantry without static typing.
R=adg, mattn.jp
CC=golang-dev
https://golang.org/cl/4973045
cdecl calbacks have been implemented in C/ASM code, just Go function is missing
R=golang-dev, alex.brainman
CC=golang-dev
https://golang.org/cl/4969047
-s now means *disable* escape analysis.
Fix escape leaks for struct/slice/map literals.
Add ... tracking.
Rewrite new(T) and slice literal into stack allocation when safe.
Add annotations to reflect.
Reflect is too chummy with the compiler,
so changes like these affect it more than they should.
R=lvd, dave, gustavo
CC=golang-dev
https://golang.org/cl/4954043
8g/cgen.c:
8g/gobj.c
. dropped unnecessary assignments;
8g/gg.h
. added varargckk pragmas;
8g/ggen.c
. dropped duplicate assignment;
8g/gsubr.c
. adjusted format in print statement;
. dropped unnecessary assignment;
. replaced GCC's _builtin_return_address(0) with Plan 9's
getcallerpc(&n) which is defined as a macro in <u.h>;
8g/list.c
. adjusted format in snprint statement;
8g/opt.h
. added varargck pragma (Adr*) that is specific for the invoking
modules;
8g/peep.c
. dropped unnecessary incrementation;
R=rsc
CC=golang-dev
https://golang.org/cl/4974044
The fix is to add ' ' after ';' so that we match
what we used to generate.
Packages like http look for the string with
the space in it, and I don't see a reason to
be so terse.
Also s/buffer/b/
TBR=bradfitz
CC=golang-dev
https://golang.org/cl/4959044
Doing a socket/listen on an unspecified address with an unspecified
address family is likely to result in an AF_INET6 socket on an IPv6
capable system, which under OpenBSD means IPv6 only - not IPv4 *and*
IPv6. In this case trying to connect to this socket from an IPv4
loopback address is not going to end well.
R=golang-dev, rsc
CC=golang-dev
https://golang.org/cl/4807057
When the encoder was updated to respect the ",omitempty"
struct tag options, the decoder half was never updated to know
about the new struct tag format. (the format is now an optional
name, followed by zero or more ",option" strings)
This only affected people who used ",omitempty" along with
a field name. In that case, the serialized JSON wouldn't
decode to the original value.
R=golang-dev, dvyukov
CC=golang-dev
https://golang.org/cl/4965049
This introduces new APIs.
- DialConfig can open client connection using Config, so user can specify protocol version, tls.Config if necessary.
- Message can be used to send/receive text/binary data in a frame.
- JSON can be used to send/receive JSON data in a frame.
R=golang-dev, adg, rsc, m, tarmigan, raul.san, yohcop
CC=golang-dev
https://golang.org/cl/4635084
Write to image.*.Pix directly in the case of RGB, RGBA and NRGBA
images. For the latter two, the file format matches the memory layout
so a simple copy can be used.
RGB image before/after:
tiff.BenchmarkDecoder 748137 ns/op (62.39 MB/s) 251256 ns/op (185.76 MB/s) x3.0
NRGBA image before/after:
tiff.BenchmarkDecoder 775540 ns/op (80.12 MB/s) 116721 ns/op (532.34 MB/s) x6.6
R=nigeltao
CC=golang-dev
https://golang.org/cl/4929046
- implemented stand-alone Throttle mechanism
- added new flag -index_throttle to godoc
- index throttling enables index creation when running
godoc on app engine as it keeps godoc responsive
R=rsc, dsymonds, adg
CC=golang-dev
https://golang.org/cl/4963043
This fixes goinstall so it doesn't try to install unneeded
packages or get confused with non-existent loops.
R=golang-dev, adg, gustavo
CC=golang-dev
https://golang.org/cl/4958046
This makes decimal a good test
case for the escape analysis.
With escape analysis:
benchmark old ns/op new ns/op delta
BenchmarkAtof64Decimal 1954 243 -87.56%
BenchmarkAtof64Float 2008 293 -85.41%
BenchmarkAtof64FloatExp 10106 8814 -12.78%
BenchmarkAtof64Big 5113 3486 -31.82%
R=golang-dev, gri
CC=golang-dev
https://golang.org/cl/4861042
#include "go.h" (or "gg.h")
becomes
#include <u.h>
#include <libc.h>
#include "go.h"
so that go.y can #include <stdio.h>
after <u.h> but before "go.h".
This is necessary on Plan 9.
R=ken2
CC=golang-dev
https://golang.org/cl/4971041
The following testcases now pass:
`<a href=x` tests that we do not error on partial unquoted attrs.
`<a href=x ` tests that spaces do end unquoted attrs on spaces.
`<a href=''` tests that we recognize the end of single quoted attrs.
`<a href=""` tests that we recognize the end of double quoted attrs.
R=golang-dev, nigeltao
CC=golang-dev
https://golang.org/cl/4932051
