This reduces the overhead necessary to work with OS-specific
file details, hides the implementation of FileStat, and
preserves the implementation-specific nature of Sys.
Expressions such as:
stat.(*os.FileInfo).Sys.(*syscall.Stat_t).Uid
fi1.(*os.FileStat).SameFile(fi2.(*os.FileStat))
Are now spelled as::
stat.Sys().(*syscall.Stat_t).Uid
os.SameFile(fi1, fi2)
R=cw, bradfitz, rsc
CC=golang-dev
https://golang.org/cl/5448079
This patch adds a function to get the current cpu ticks. This is
deemed to be 'sufficiently random' to use to seed fastrand to mitigate
the algorithmic complexity attacks on the hash table implementation.
On AMD64 we use the RDTSC instruction. For 386, this instruction,
while valid, is not recognized by 8a so I've inserted the opcode by
hand. For ARM, this routine is currently stubbed to return a constant
0 value.
Future work: update 8a to recognize RDTSC.
Fixes#2630.
R=rsc
CC=golang-dev
https://golang.org/cl/5606048
- Unified bounary conditions for NFC and NFD and removed some indirections.
This enforces boundaries at the character level, which is typically what
the user expects. (NFD allows a boundary between 'a' and '`', for example,
which may give unexpected results for collation. The current implementation
is already stricter than the standard, so nothing much changes. This change
just formalizes it.
- Moved methods of qcflags to runeInfo.
- Swapped YesC and YesMaybe bits in qcFlags. This is to aid future changes.
- runeInfo return values use named fields in preperation for struct change.
- Replaced some left-over uint32s with rune.
R=r
CC=golang-dev
https://golang.org/cl/5607050
Without this change, fd3 can be collected by the garbage
collector and finalized, which causes the file descriptor to
be closed, which causes the call to os.Open to return 3 rather
than the expected descriptor number.
R=golang-dev, gri, bradfitz, bradfitz, iant
CC=golang-dev
https://golang.org/cl/5607056
The rule is that build directives can be preceded only
by blank lines and other line comments, not /* */ comments.
R=golang-dev, adg, bradfitz
CC=golang-dev
https://golang.org/cl/5619045
This is a minimal API extension, it makes it possible
to implement missing Int functionality externally w/o
compromising efficiency. It is the hope that this will
reduce the number of feature requests going directly
into the big package.
Also: Fixed some naming inconsistencies: The receiver
is only called z when it is also the result.
R=golang-dev, agl
CC=golang-dev
https://golang.org/cl/5607055
This can drastically reduce the number of system
calls made by programs that repeatedly query the
environment.
R=golang-dev, rsc
CC=golang-dev
https://golang.org/cl/5599054
Ignore result of setting SO_BROADCAST.
Disable TestSimpleListenMulticastUDP as
setIPv4MulticastInterface is not implemented.
R=golang-dev, bradfitz, mikioh.mikioh
CC=golang-dev
https://golang.org/cl/5610044
Black box test is too time-consuming, as the bug
does not appear until Scan has processed 2 GB of
input in total across multiple calls, so no test.
Thanks to Frederick Mayle for the diagnosis and fix.
Fixes#2809.
R=golang-dev, r
CC=golang-dev
https://golang.org/cl/5611043
Both are unused and undocumented.
InvalidConnError is also non-idiomatic: a FooError type can
typically describe something, else it would be an ErrFoo
variable.
R=golang-dev, alex.brainman
CC=golang-dev
https://golang.org/cl/5609045
The type being unavailable on Windows was the only API
signature difference in the net package.
R=golang-dev, rsc
CC=golang-dev
https://golang.org/cl/5608043
- enable AllMethods flag (default: not set)
- fix logic determining which methods to show
- added respective test case in testdata/e.go for AllMethods = false
- added test case set for AllMethods = true
The critical changes/files to look at are:
- testdata/e{0,1,2}.golden: T4.M should only show up as method of T5 in e2.golden
- reader.go: always include top-level methods, and negate former logic for embedded methods
(rewrote as a switch for better comprehensability)
Fixes#2791.
R=rsc, rsc
CC=golang-dev
https://golang.org/cl/5576057
This fixes some test noise in TestStressSurpriseServerCloses when
ulimit -n something low, like 256 on a Mac.
Previously, when the server closed on us and we were expecting more
responses (like we are in that test), we'd read an "Unexpected EOF"
and just forget about the client's net.Conn. Now it's closed,
rather than waiting on the finalizer to release the fd.
R=golang-dev, rsc
CC=golang-dev
https://golang.org/cl/5602043
I missed an overflow in contract because I suspected that the prime
elimination would take care of it. It didn't, and I forgot to get back
to the overflow. Because of this, p224Contract may have produced a
non-minimal representation, causing flakey failures ~0.02% of the
time.
R=golang-dev, rsc
CC=golang-dev
https://golang.org/cl/5592045
This CL introduces new function ListenMulticastUDP to fix
multicast UDP listening across multiple listeners issue,
to replace old multicast methods JoinGroup and LeaveGroup
on UDPConn.
This CL also enables multicast testing by default.
Fixes#2730.
R=rsc, paul.a.lalonde, fullung, devon.odell
CC=golang-dev
https://golang.org/cl/5562048
We support SSLv3 as a server but not as a client (and we don't want to
support it as a client). This change fixes the error message when
connecting to an SSLv3 server since SSLv3 support on the server side
made mutualVersion accept SSLv3.
R=golang-dev, rsc
CC=golang-dev
https://golang.org/cl/5545073
Ensure that file descriptors have not already been leaked into our
environment - close any that are open at the start of the
TestExtraFiles test.
Also use the appropriate command for listing open files.
R=golang-dev, bradfitz
CC=golang-dev
https://golang.org/cl/5574062
This patch adds a hash seed to the Hmap struct. Each seed is
initialized by runtime.fastrand1(). This is the first step of a
solution to issue 2630. Fastrand1 still needs to be updated to provide
us with actually random bits.
R=golang-dev, rsc
CC=golang-dev
https://golang.org/cl/5599046
Consequently, remove many package Makefiles,
and shorten the few that remain.
gomake becomes 'go tool make'.
Turn off test phases of run.bash that do not work,
flagged with $BROKEN. Future CLs will restore these,
but this seemed like a big enough CL already.
R=golang-dev, r
CC=golang-dev
https://golang.org/cl/5601057
We weren't properly deleting the various header
files (that were temporarily renamed) if a $CC
for the current $GOARCH didn't exist. And since
the compiler checks the current directory for
headers before any -I arguments, this had the
unfortunate side effect of including the last
generated headers instead of the correct ones.
R=r, rsc
CC=golang-dev
https://golang.org/cl/5581055
Also delete gotest, since it's messy to fix and slated for deletion anyway.
A couple of things outside src can't be tested any more. "go test" will be
fixed and these tests will be re-enabled. They're noisy for now.
Fixes#284.
R=rsc
CC=golang-dev
https://golang.org/cl/5598049
Cookies recieved in a response to a POST request are stored
in the client's jar like they are for GET requests.
R=golang-dev, rsc
CC=bradfitz, golang-dev
https://golang.org/cl/5576065
fix, vet
yacc is also fixed (it was wrong before)
All that's left is the commands used during compilation
This looks like a huge CL, but it's almost all file renames.
The action is in cmd/go/pkg.go, the Makefiles, and .../doc.go.
R=golang-dev, rsc
CC=golang-dev
https://golang.org/cl/5595044
1) create go-tool dir in make.bash
2) clean up stale binaries in make.bash
3) add 'tool' command to go
4) convert goyacc->yacc as a first test tool
Since goyacc stands alone, it's a safe trial.
R=rsc
CC=golang-dev
https://golang.org/cl/5576061
They're not portable, and pkg os is supposed to be portable.
Fixes#2562
R=golang-dev, mikioh.mikioh, r, n13m3y3r, rsc
CC=golang-dev
https://golang.org/cl/5574078
Added flag AllMethods: if not set (future default), embedded
methods of exported (and thus visible) embedded fields are not
shown in the final package documentation
The actual change for AllMethods is just in sortedFuncs. All
other changes are simplifications of the existing logic (mostly
deletion of code): Because method conflicts due to embedding
must always be detected, remove any premature elimination of
types and methods. Instead collect all named types and all
methods and do the filtering at the end.
Miscellaneous:
- renamed baseType -> namedType
- streamline logic for recording embedded types
- record embedded types via a map (simpler data structures)
AllMethods is set by default; so the output is unchanged and
the tests pass. The next CL will enable the AllMethods flag
and have adjusted tests (and fix issue 2791).
R=rsc
CC=golang-dev
https://golang.org/cl/5572076
+eliminates a possibility of sending a call to Done several times.
+fixes memory leak in case of temporal Write errors.
+fixes data race on Client.shutdown.
+fixes data race on Client.closing.
+fixes comments.
Fixes#2780.
R=r, rsc
CC=golang-dev, mpimenov
https://golang.org/cl/5571063
The new url.URL's parsing can be too canonicalizing for
certain applications. By keeping the original request URI
around, we give applications a gross escape hatch while
keeping the URL package clean and simple for normal uses.
(From a discussion with Gary Burd, Gustavo Niemeyer,
and Russ Cox.)
Fixes#2782
R=golang-dev, rsc, dsymonds
CC=golang-dev
https://golang.org/cl/5580044
It means serious user error that can lead to
hard to debug issues under load, log entry
will not harm.
R=golang-dev, rsc
CC=golang-dev
https://golang.org/cl/5574075
Separating Method from Func made the code only more complicated
without adding much to the useability/readability of the API.
Reverted to where it was, but leaving the new method-specific
fields Orig and Level.
Former clients (godoc) of doc.Method only used the Func fields;
and because Func was embedded, no changes are needed with respect
to the removal of Method.
Changed type of Func.Recv from ast.Expr to string. This was a
long-standing TODO. Also implemented Func.Orig field (another TODO).
No further go/doc API changes are expected for Go 1.
R=rsc, r, r
CC=golang-dev
https://golang.org/cl/5577043
Added a cache to compensate for extra call overhead.
go test -bench=Print marginally faster (in the noise).
R=r
CC=golang-dev
https://golang.org/cl/5574061
The implementation is divided into 4 phases:
1) export filtering of an incoming AST if necessary (exports.go)
2) reading of a possibly filtered AST (reader.go: type reader)
3) method set computation (reader.go)
4) sorting and creation of final documentation (reader.go)
In contrast to the old implementation, the presentation data
(Names, Docs, Decls, etc.) are created immediately upon reading
the respective AST node. Also, all types are collected (embedded
or not) in a uniform way.
Once the entire AST has been processed, all methods and types
have been collected and the method sets for each type can be
computed (phase 3).
To produce the final documentation, the method sets and value
maps are sorted.
There are no API changes. Passes the existing test suite unchanged.
R=rsc, rogpeppe
CC=golang-dev
https://golang.org/cl/5554044
Reimplement the test based on code from adg@golang.org.
The previous version has a race since the file is closed via defer
rather than in the go routine. This meant that the file could be
closed before the go routine has actually received io.EOF. It then
receives EBADF and continues to do zero-byte writes to the pipe.
This addresses an issue seen on FreeBSD and OpenBSD, where the test
passes but exits with a SIGPIPE, resulting in a failure.
R=golang-dev, adg
CC=golang-dev
https://golang.org/cl/5554083
work in progress, and we are not ready to freeze its API for Go 1.
Package html still exists, containing just two functions: EscapeString
and UnescapeString.
Both the packages at exp/html and html are "package html". The former
is a superset of the latter.
At some point in the future, the exp/html code will move back into
html, once we have finalized the parser API.
R=rsc, dsymonds
CC=golang-dev
https://golang.org/cl/5571059
Marshaler has a number of open areas that need
further thought (e.g. it doesn't handle attributes,
it's supposed to handle tag names internally but has
no information to do so, etc).
We're removing it now and will bring it back with an
interface that covers these aspects, after Go 1.
Related to issue 2771, but doesn't fix it.
R=golang-dev, rsc
CC=golang-dev
https://golang.org/cl/5574057
In order to allow buildscript.sh to generate buildscripts for all
$GOOS/$GOARCH combinations, we have to generate dummy files for cmd/go.
Fixes#2586.
R=rsc, golang-dev
CC=golang-dev
https://golang.org/cl/5557050
Also, add an explicit error type when the right hand side is an unexported
function.
R=golang-dev, gri, rogpeppe, agl, rsc
CC=golang-dev
https://golang.org/cl/5564048
Includes gofix module. The only case not covered should be
xml.Unmarshal, since it remains with a similar interface, and
would require introspecting the type of its first argument
better.
Fixes#2626.
R=golang-dev, rsc, gustavo
CC=golang-dev
https://golang.org/cl/5574053
CL 5572043 removed the last uses of this field.
The information is readily available from Type.Decl.
R=rsc, golang-dev
CC=golang-dev
https://golang.org/cl/5570049
The bitLen function currently shifts out blocks of 8 bits at a time.
This change replaces this sorta-linear algorithm with a log(N)
one (shift out 16 bits, then 8, then 4, then 2, then 1).
I left the start of it linear at 16 bits at a time so that
the function continues to work with 32 or 64 bit values
without any funkiness.
The algorithm is similar to several of the nlz ("number of
leading zeros") algorithms from "Hacker's Delight" or the
"bit twiddling hacks" pages.
Doesn't make a big difference to the existing benchmarks, but
I'm using the code in a different context that calls bitLen
much more often, so it seemed worthwhile making the existing
codebase faster so that it's a better building block.
Microbenchmark results on a 64-bit Macbook Pro using 6g from weekly.2012-01-20:
benchmark old ns/op new ns/op delta
big.BenchmarkBitLen0 4 6 +50.12%
big.BenchmarkBitLen1 4 6 +33.91%
big.BenchmarkBitLen2 6 6 +3.05%
big.BenchmarkBitLen3 7 6 -19.05%
big.BenchmarkBitLen4 9 6 -30.19%
big.BenchmarkBitLen5 11 6 -42.23%
big.BenchmarkBitLen8 16 6 -61.78%
big.BenchmarkBitLen9 5 6 +18.29%
big.BenchmarkBitLen16 18 7 -60.99%
big.BenchmarkBitLen17 7 6 -4.64%
big.BenchmarkBitLen31 19 7 -62.49%
On an ARM machine (with the previous weekly):
benchmark old ns/op new ns/op delta
big.BenchmarkBitLen0 37 50 +36.56%
big.BenchmarkBitLen1 59 51 -13.69%
big.BenchmarkBitLen2 74 59 -20.40%
big.BenchmarkBitLen3 92 60 -34.89%
big.BenchmarkBitLen4 110 59 -46.09%
big.BenchmarkBitLen5 127 60 -52.68%
big.BenchmarkBitLen8 181 59 -67.24%
big.BenchmarkBitLen9 78 60 -23.05%
big.BenchmarkBitLen16 199 69 -65.13%
big.BenchmarkBitLen17 91 70 -23.17%
big.BenchmarkBitLen31 210 95 -54.43%
R=golang-dev, dave, edsrzf, gri
CC=golang-dev
https://golang.org/cl/5570044
This lets the client of go/build specify additional tags that
can be recognized in a // +build directive. For example,
a build for a custom environment like App Engine might
include "appengine" in the BuildTags list, so that packages
can be written with some files saying
// +build appengine (build only on app engine)
or
// +build !appengine (build only when NOT on app engine)
App Engine here is just a hypothetical context. I plan to use
this in the cmd/go sources to distinguish the bootstrap version
of cmd/go (which will not use networking) from the full version
using a custom tag. It might also be useful in App Engine.
Also, delete Build and Script, which we did not end up using for
cmd/go and which never got turned on for real in goinstall.
R=r, adg
CC=golang-dev
https://golang.org/cl/5554079
Golden files have extension .d.golden where d is the mode value (0 or 1 for now)
(i.e., testdata/file.out is now testdata/file.0.golden, and there is a new file
testdata/file.1.golden for each testcase)
R=rsc
CC=golang-dev
https://golang.org/cl/5573046
Marshalling of []byte in attributes and the general
marshalling of named []byte types was fixed.
A []byte field also won't be nil if an XML element
was mapped to it, even if the element is empty.
Tests were introduced to make sure that *struct{}
fields works correctly for element presence testing.
No changes to the logic made in that regard.
R=rsc
CC=golang-dev
https://golang.org/cl/5539070
It was 2^31, but that could cause overflow and trouble.
Reduce it to 2^30 and add a TODO.
R=golang-dev, iant
CC=golang-dev
https://golang.org/cl/5562049
It takes too much memory to be reliable and causes
trouble on 32-bit machines.
Sigh.
Fixes#2756.
R=golang-dev, bradfitz
CC=golang-dev
https://golang.org/cl/5567043
Make the panic detectable, and use that in ioutil.ReadFile to
give an error if the file is too big.
R=golang-dev, minux.ma, bradfitz
CC=golang-dev
https://golang.org/cl/5563045
- use proper Win64 gcc calling convention when
calling initcgo on amd64
- increase g0 stack size to 64K on amd64 to make
it the same as 386
- implement C.sleep
- do not use C.stat, since it is renamed to C._stat by mingw
- use fopen to implement TestErrno, since C.strtol
always succeeds on windows
- skip TestSetEnv on windows, because os.Setenv
sets windows process environment, while C.getenv
inspects internal C runtime variable instead
R=golang-dev, vcc.163, rsc
CC=golang-dev
https://golang.org/cl/5500094
instead of the origin.
This makes YCbCr match the other image types (e.g. RGBA, Gray) in
that an image's bounds is not restricted to the positive quadrant.
Also optimize the YCbCr draw code by hoisting some computation
outside of the loop.
benchmark old ns/op new ns/op delta
draw.BenchmarkYCbCr 2544418 2373558 -6.72%
Like https://golang.org/cl/4681044/ I don't think a gofix is
feasible. People will have to make manual changes. On the other hand,
directly manipulating YCbCr images is relatively rare, compared to
RGBA images, and if other code just uses the jpeg and draw packages
instead of messing directly with a YCbCr's []byte representations,
then things should just continue to work.
R=r
CC=golang-dev
https://golang.org/cl/5558048
Remove NewMD5, NewSHA1 and NewSHA256 in favor of using New and
explicitly importing the used hash-function. This way when using, for
example, HMAC with RIPEMD there's no md5, sha1 and sha256 linked in
through the hmac package.
A gofix rule is included, and applied to the standard library (3 files
altered).
This change is the result of a discussion at
https://golang.org/cl/5550043/ to pull the discussion about
deprecating these functions out of that issue.
R=golang-dev, agl
CC=golang-dev, r, rsc
https://golang.org/cl/5556058
Fix originally from rogpeppe in 5414048 but was rolled
back due to test breakage.
This CL makes the test more robust to order of operations.
Fixes#2480 again.
R=golang-dev, gri
CC=golang-dev
https://golang.org/cl/5536072
The previous version of all the node.String methods printed the parse
tree and was useful for developing the parse tree code. Now that that's done,
we might as well print the nodes using the standard template syntax.
It's much easier to read and makes error reporting look more natural.
Helps issue 2644.
R=rsc, n13m3y3r
CC=golang-dev
https://golang.org/cl/5553066
So as to give out stack trace for panic in examples.
This behavior also matches the tests'.
Fixes#2691.
R=golang-dev
CC=golang-dev
https://golang.org/cl/5554061
(Sending to r because of the API change.)
Over time we might want to add support for other key types.
While I was in the code, I also made the use of RawSubject the same
between Subject and Issuer when creating certificates.
R=r, rsc
CC=golang-dev
https://golang.org/cl/5554049
(Sending to r because of the API change.)
This change alters the API for crypto/elliptic to permit different
implementations in the future. This will allow us to add faster,
constant-time implementations of the standard curves without any more
API changes.
As a demonstration, it also adds a constant-time implementation of
P224. Since it's only 32-bit, it's actually only about 40% the speed
of the generic code on a 64-bit system.
R=r, rsc
CC=golang-dev
https://golang.org/cl/5528088
TestNonStandardNormalValues runs 1.5s,
the change reduces it to 0.2s in short mode.
The problem is with slow machines, emulators and dynamic tools.
R=golang-dev, r
CC=golang-dev
https://golang.org/cl/5540065
Also, disable server-side generation of TOC for commands as they would
only ever show Overview. The JS does a better job (for now).
Fixes#2732.
R=gri, dsymonds
CC=golang-dev
https://golang.org/cl/5558046
It works with NewTicker too, but is potentially a bit less efficient,
and reads wrong.
This is what happens when you TBR Windows changes, I guess.
R=golang-dev, gri, iant
CC=golang-dev
https://golang.org/cl/5536060
Previously, a timeout (in int64 nanoseconds) applied to a granularity
even smaller than one operation: a 100 byte read with a 1 second timeout
could take 100 seconds, if the bytes all arrived on the network 1 second
apart. This was confusing.
Rather than making the timeout granularity be per-Read/Write,
this CL makes callers set an absolute deadline (in time.Time)
after which operations will fail. This makes it possible to
set deadlines at higher levels, without knowing exactly how
many read/write operations will happen in e.g. reading an HTTP
request.
Fixes#2723
R=r, rsc, dave
CC=golang-dev
https://golang.org/cl/5555048
This matches the other color models. It seems that this was missed
during the review of 5544073.
R=nigeltao, david.crawshaw
CC=golang-dev
https://golang.org/cl/5536057
Packages to test are kept in ./testdata together
with the corresponding golden (packagename.out)
file.
To update the golden files, run: go test -update
R=rsc
CC=golang-dev
https://golang.org/cl/5543054
- Use a faster closure-based iterator rather than a channel-based one.
- Otherwise: better code organization, but no other API changes.
R=r, r
CC=golang-dev
https://golang.org/cl/5557051
First was, apart from adding tests, a single line of code (to add the
constructor function). Adding SHA512-based hashing to crypto/hmac
required minor rework of the package because of a previously hardcoded
block-size in it's implementation. Instead of using a hash.Hash
generator function the constructor function now uses a crypto.Hash
type, which was extended to expose information about block size.
The only standard library package impacted by the change is
crypto/tls, for which the fix is included in this patch. It might be
useful to extend gofix to include this API change too.
R=agl, r, rsc, r
CC=golang-dev
https://golang.org/cl/5550043
We need to avoid allocating an extra word for the interface value
passing the floating-point value as an interface{}. It's easy.
Fixes#2722.
R=golang-dev, gri
CC=golang-dev
https://golang.org/cl/5553044
The package documentation did not mention them.
They were described only in godoc for gotest, and that's going away.
R=golang-dev, rsc, adg
CC=golang-dev
https://golang.org/cl/5539079
On 32-bit machines, %g takes an extra malloc. I don't know why yet,
but this makes the test pass again, and enables it even for -short.
Fixes#2653.
R=golang-dev, bradfitz, r
CC=golang-dev
https://golang.org/cl/5542055
Duplicated fields from URL were dropped so that its behavior
is simple and expected when being stringified and when being
operated by packages like http. Most of the preserved fields
are in unencoded form, except for RawQuery which continues to
exist and be more easily handled via url.Query().
The RawUserinfo field was also replaced since it wasn't practical
to use and had limitations when operating with empty usernames
and passwords which are allowed by the RFC. In its place the
Userinfo type was introduced and made accessible through the
url.User and url.UserPassword functions.
What was previous built as:
url.URL{RawUserinfo: url.EncodeUserinfo("user", ""), ...}
Is now built as:
url.URL{User: url.User("user"), ...}
R=rsc, bradfitz, gustavo
CC=golang-dev
https://golang.org/cl/5498076
The terminal code in exp/terminal was forked from the code in exp/ssh.
This change removes the duplicated code from exp/ssh in favour of
using exp/terminal.
R=rsc
CC=golang-dev
https://golang.org/cl/5375064
This CL revises existing platform-dependent default socket
options to make it possible to accomodate multiple multicast
datagram listeners on a single service port.
Also removes unnecessary SO_REUSEADDR, SO_REUSEPORT socket
options from unicast datagram sockets by default.
Fixes#1692.
R=devon.odell, alex.brainman, rsc
CC=golang-dev
https://golang.org/cl/5538052
(This was missing in the last change because I uploaded it from the
wrong machine.)
Large pastes previously misbehaved because the code tried reading from
the terminal before checking whether an line was already buffered.
Large pastes can cause multiples lines to be read at once from the
terminal.
R=bradfitz
CC=golang-dev
https://golang.org/cl/5542049
exp/proxy provides client support for tunneling connections through
various proxies.
This is an initial, incomplete sketch of the code to lay down an
API.
R=golang-dev, r, r, bradfitz, rsc
CC=golang-dev
https://golang.org/cl/5490062
Also verified in external test suite that this fixes MySQL
resource exhaustion problems, and also exposed a double-free
bug in the gosqlite3 driver (where gosqlite3 either got lucky
before, or was working around this bug)
R=golang-dev, rsc
CC=golang-dev
https://golang.org/cl/5544057
The implementation is similar to the one from the double-conversion
library used in the Chrome V8 engine.
old ns/op new ns/op speedup
BenchmarkAppendFloatDecimal 591 480 1.2x
BenchmarkAppendFloat 2956 486 6.1x
BenchmarkAppendFloatExp 10622 503 21.1x
BenchmarkAppendFloatNegExp 40343 483 83.5x
BenchmarkAppendFloatBig 2798 664 4.2x
See F. Loitsch, ``Printing Floating-Point Numbers Quickly and
Accurately with Integers'', Proceedings of the ACM, 2010.
R=rsc
CC=golang-dev, remy
https://golang.org/cl/5502079
Merge package files in the go/ast MergePackageFiles
function always in the same order (sorted by filename)
instead of map iteration order to obtain the same
package file each time. This functionality is used
by godoc when displaying packages in ?m=src mode.
Also: minor cleanup in godoc.go.
R=rsc
CC=golang-dev
https://golang.org/cl/5540054
This CL improves the xml package in the following ways:
- makes its interface match established conventions
- brings Marshal and Unmarshal closer together
- fixes a large number of bugs and adds tests
- improves speed significantly
- organizes and simplifies the code
Fixes#2426.
Fixes#2406.
Fixes#1989.
What follows is a detailed list of those changes.
- All matching is case sensitive without special processing
to the field name or xml tag in an attempt to match them.
Customize the field tag as desired to match the correct XML
elements.
- Flags are ",flag" rather than "flag". The names "attr",
"chardata", etc, may be used to name actual XML elements.
- Overriding of attribute names is possible with "name,attr".
- Attribute fields are marshalled properly if they have
non-string types. Previously they were unmarshalled, but were
ignored at marshalling time.
- Comment fields tagged with ",comment" are marshalled properly,
rather than being marshalled as normal fields.
- The handling of the Any field has been replaced by the ",any"
flag to avoid unexpected results when using the field name for
other purposes, and has also been fixed to interact properly
with name paths. Previously the feature would not function
if any field in the type had a name path in its tag.
- Embedded struct support fixed and cleaned so it works when
marshalling and also when using field paths deeper than one level.
- Conflict reporting on field names have been expanded to cover
all fields. Previously it'd catch only conflicts of paths
deeper than one level. Also interacts correctly with embedded
structs now.
- A trailing '>' is disallowed in xml tags. It used to be
supported for removing the ambiguity between "attr" and "attr>",
but the marshalling support for that was broken, and it's now
unnecessary. Use "name" instead of "name>".
- Fixed docs to point out that a XMLName doesn't have to be
an xml.Name (e.g. a struct{} is a good fit too). The code was
already working like that.
- Fixed asymmetry in the precedence of XML element names between
marshalling and unmarshalling. Marshal would consider the XMLName
of the field type before the field tag, while unmarshalling would
do the opposite. Now both respect the tag of the XMLName field
first, and a nice error message is provided in case an attempt
is made to name a field with its tag in a way that would
conflict with the underlying type's XMLName field.
- Do not marshal broken "<???>" tags when in doubt. Use the type
name, and error out if that's not possible.
- Do not break down unmarshalling if there's an interface{} field
in a struct.
- Significant speed boost due to caching of type metadata and
overall allocation clean ups. The following timings reflect
processing of the the atom test data:
Old:
BenchmarkMarshal 50000 48798 ns/op
BenchmarkUnmarshal 5000 357174 ns/op
New:
BenchmarkMarshal 100000 19799 ns/op
BenchmarkUnmarshal 10000 128525 ns/op
R=cw, gustavo, kevlar, adg, rogpeppe, fullung, christoph, rsc
CC=golang-dev
https://golang.org/cl/5503078
- the main changes are removing the Doc suffix
from the exported types, so instead of
doc.TypeDoc one will have doc.Type, etc.
- All exported types now have a Name (or Names) field.
For Values, the Names field lists all declared variables
or constants.
- Methods have additional information about where they are
coming from.
- There's a mode field instead of a bool to
control the package's operation, which makes
it easier to extend w/o API changes.
Except for the partially implemented new Method type,
this is based on existing code. A clean rewrite is in
progress based on this new API.
R=rsc, kevlar
CC=golang-dev
https://golang.org/cl/5528060
Only ParseFile, ParseDir, and ParseExpr are used in the tree.
If partial parsing of code is required, it is fairly simple
to wrap the relevant piece of code into a dummy package for
parsing (see parser.ParseExpr).
Also: minor cleanups.
R=rsc
CC=golang-dev
https://golang.org/cl/5535055
These functions are mostly of interest for debugging; the
number of bytes written is uninteresting.
R=r, bradfitz
CC=golang-dev
https://golang.org/cl/5540046
In a test that does
func TestFoo(t *testing.T) {
defer cleanup()
t.Fatal("oops")
}
it can be important that cleanup run as the test fails.
The old code did this in Fatal:
t.signal <- t
runtime.Goexit()
The runtime.Goexit would run the deferred cleanup
but the send on t.signal would cause the main test loop
to move on and possibly even exit the program before
the runtime.Goexit got a chance to run.
This CL changes tRunner (the top stack frame of a test
goroutine) to send on t.signal as part of a function
deferred by the top stack frame. This delays the send
on t.signal until after runtime.Goexit has run functions
deferred by the test itself.
For the above TestFoo, this CL guarantees that cleanup
will run before the test binary exits.
This is particularly important when cleanup is doing
externally visible work, like removing temporary files
or unmounting file systems.
R=golang-dev, r
CC=golang-dev
https://golang.org/cl/5532078
Tighter octal parsing broke some tests and were disabled in
https://golang.org/cl/5530051
Those tests were broken. The CSS decoder was supposed to see CSS
hex escape sequences of the form '\' <hex>+, but those escape
sequences were instead being consumed by the template parser.
This change properly escapes those escape sequences, and uses
proper escaping for NULs.
R=golang-dev, rsc, nigeltao
CC=golang-dev
https://golang.org/cl/5529073
- Changed the Scan API semantics slightly:
The token literal string is only returned
if the token is a literal, comment, semicolon,
or illegal character. In all other cases, the
token literal value is determined by the token
value.
Clients that care about the token literal value
when not present can always use the following
piece of code:
pos, tok, lit := scanner.Scan()
if lit == "" {
lit = tok.String()
}
- Changed token.Lookup API to use a string instead
of a []byte argument.
- Both these changes were long-standing TODOs.
- Added BenchmarkScan.
This change permits a faster implementation of Scan
with much fewer string creations:
benchmark old ns/op new ns/op delta
scanner.BenchmarkScan 74404 61457 -17.40%
R=golang-dev, rsc
CC=golang-dev
https://golang.org/cl/5532076
This is a relic from the times when we switched
to automatic semicolon insertion. It's still use-
ful to have a non-exported switch for testing.
R=golang-dev, r, rsc
CC=golang-dev
https://golang.org/cl/5528077
1) Include Szabolcs Nagy's patch which adds serialisation for more
signature subpackets.
2) Include Szabolcs Nagy's patch which adds functions for making DSA
keys.
3) Make the random io.Reader an argument to the low-level signature
functions rather than having them use crypto/rand.
4) Rename crypto/openpgp/error to crypto/openpgp/errors so that it
doesn't clash with the new error type.
R=bradfitz, r
CC=golang-dev
https://golang.org/cl/5528044
pkg/runtime/sys_darwin_amd64.s: fixes syscall select nr
pkg/runtime/sys_linux_arm.s: uses newselect instead of the now unimplemented
(old) select, also fixes the wrong div/mod statements in runtime.usleep.
Fixes#2633
R=golang-dev, dave, rsc
CC=golang-dev
https://golang.org/cl/5504096
What package image currently provides is a larger image consisting
of many copies of a smaller image.
More generally, a tiled image could be a quilt consisting of different
smaller images (like Google Maps), or a technique to view a portion of
enormous images without requiring the whole thing in memory.
This richer construct might not ever belong in the standard library (and
is definitely out of scope for Go 1), but I would like the option for
image.Tiled to be its name.
R=r, rsc
CC=golang-dev
https://golang.org/cl/5530062
is consistent with what the Go compiler returns when such sequences
appear in string literals.
Fixes#2658.
R=golang-dev, rsc, r, r, nigeltao
CC=golang-dev
https://golang.org/cl/5530051
