I added the clientChan's msg channel to the list of channels that are closed in mainloop when the server sends a channelCloseMsg.
I added an ExitError type that wraps a Waitmsg similar to that of os/exec. I fill ExitStatus with the data returned in the 'exit-status' channel message and Msg with the data returned in the 'exit-signal' channel message.
Instead of having Wait() return on the first 'exit-status'/'exit-signal' I have it return an ExitError containing the status and signal when the clientChan's msg channel is closed.
I added two tests cases to session_test.go that test for exit status 0 (in which case Wait() returns nil) and exit status 1 (in which case Wait() returns an ExitError with ExitStatus 1)
R=dave, agl, rsc, golang-dev, bradfitz
CC=golang-dev
https://golang.org/cl/5452051
It was fragile and non-portable, and then became spammy with
the os.EINVAL removal. Now it just uses the length of the
Peek return value instead.
R=golang-dev, alex.brainman
CC=golang-dev
https://golang.org/cl/5453065
The code in hash functions themselves could write directly into the
output buffer for a savings of about 50ns. But it's a little ugly so I
wasted a copy.
R=bradfitz
CC=golang-dev
https://golang.org/cl/5440111
This CL cleans up the client auth tests, making the
individual test body more manageable.
Also, adds tests for rsa and dsa key negotiation.
Finally, remove the package level use of the variable
strings, which avoids conflicting with the strings pkg.
R=gustav.paul, agl, n13m3y3r, rsc
CC=golang-dev
https://golang.org/cl/5447049
This avoids a problem when creating certificates with parents that
were produce by other code: the Go structures don't contain all the
information about the various ASN.1 string types etc and so that
information would otherwise be lost.
R=golang-dev, rsc
CC=golang-dev
https://golang.org/cl/5453067
Allow the text template to handle the error case of no template
with the given name.
Simplification suggested by Mike Samuel.
R=mikesamuel
CC=golang-dev
https://golang.org/cl/5437147
(Note that the Int and Uint benchmarks use different test sets
and thus cannot be compared against each other. Int and Uint
conversions are approximately the same speed).
Before (best of 3 runs):
strconv_test.BenchmarkFormatInt 100000 15636 ns/op
strconv_test.BenchmarkAppendInt 100000 18930 ns/op
strconv_test.BenchmarkFormatUint 500000 4392 ns/op
strconv_test.BenchmarkAppendUint 500000 5152 ns/op
After (best of 3 runs):
strconv_test.BenchmarkFormatInt 200000 10070 ns/op (-36%)
strconv_test.BenchmarkAppendInt 200000 7097 ns/op (-63%)
strconv_test.BenchmarkFormatUint 1000000 2893 ns/op (-34%)
strconv_test.BenchmarkAppendUint 500000 2462 ns/op (-52%)
R=r, rsc, r
CC=golang-dev
https://golang.org/cl/5449093
When I disallowed map + func comparisons, I only did it
in the static case and missed the comparisons via == on
interface values. Fixing that turned these up.
R=nigeltao, r
CC=golang-dev
https://golang.org/cl/5440103
This is part one of a small set of CL's that aim to resolve
the outstanding TODOs relating to channel close and blocking
behavior.
Firstly, the hairy handling of assigning the peersId is now
done in one place. The cost of this change is the slightly
paradoxical construction of the partially created clientChan.
Secondly, by creating clientChan.stdin/out/err when the channel
is opened, the creation of consumers like tcpchan and Session
is simplified; they just have to wire themselves up to the
relevant readers/writers.
R=agl, gustav.paul, rsc
CC=golang-dev
https://golang.org/cl/5448073
This is a slight change to fmt's semantics, but means that if you use
%d to print an integer with a Stringable value, it will print as an integer.
This came up because Time.Month() couldn't cleanly print as an integer
rather than a name. Using %d on Stringables is silly anyway, so there
should be no effect outside the fmt tests.
As a mild bonus, certain recursive failures of String methods
will also be avoided this way.
R=golang-dev, adg
CC=golang-dev
https://golang.org/cl/5453053
All but 3 cases (in gcimporter.go and hixie.go)
are automatic conversions using gofix.
No attempt is made to use the new Append functions
even though there are definitely opportunities.
R=golang-dev, gri
CC=golang-dev
https://golang.org/cl/5447069
By converting array indices to uint8, they are automatically
constrained in the array range, and the binary AND with 0xff
is no longer needed anymore.
Before: aes.BenchmarkEncrypt 363 ns/op
After: aes.BenchmarkEncrypt 273 ns/op
R=golang-dev, gri, agl
CC=golang-dev, remy
https://golang.org/cl/5450084
Equality on structs will require arbitrary code for type equality,
so change algorithm in type data from uint8 to table pointer.
In the process, trim top-level map structure from
104/80 bytes (64-bit/32-bit) to 24/12.
Equality on structs will require being able to call code generated
by the Go compiler, and C code has no way to access Go return
values, so change the hash and equal algorithm functions to take
a pointer to a result instead of returning the result.
R=ken
CC=golang-dev
https://golang.org/cl/5453043
The SSH spec allows for the server to send a banner message to the client at any point during the authentication process. Currently the ssh client auth types all assume that the first response from the server after issuing a userAuthRequestMsg will be one of a couple of possible authentication success/failure messages. This means that client authentication breaks if the ssh server being connected to has a banner message configured.
This changeset refactors the noneAuth, passwordAuth and publickeyAuth types' auth() function and allows for msgUserAuthBanner during authentication.
R=golang-dev, rsc, dave, agl
CC=golang-dev
https://golang.org/cl/5432065
For now a pair of socket options SOL_SOCKET and SO_BINDTODEVICE
is supported on Linux only. I'd like to demote BindToDevice API
to syscall level because it's Linux dependent one.
In the near future, probably we may have a bit more portable
API that using IPROTO_IP/IPV6 level socket options to specify,
identify an inbound, outbound IP interface on incoming, outgoing
UDP and raw IP packets.
R=cw, golang-dev
CC=golang-dev
https://golang.org/cl/5447071
cmd.exe implicitly looks in "." before consulting PATH.
LookPath should match this behavior.
R=alex.brainman, rsc
CC=golang-dev
https://golang.org/cl/5434093
This is a more conservative approach to heading detection and
removes 11 headings from the current repository (several in
fmt). The current headscan output is:
/home/gri/go3/src/cmd/goinstall (package documentation)
Remote Repositories
The GOPATH Environment Variable
/home/gri/go3/src/pkg/exp/gotype (package documentation)
Examples
/home/gri/go3/src/pkg/html/template (package template)
Introduction
Contexts
Errors
A fuller picture
Contexts
Typed Strings
Security Model
/home/gri/go3/src/pkg/text/template (package template)
Actions
Arguments
Pipelines
Variables
Examples
Functions
Associated templates
Nested template definitions
18 headings found
R=golang-dev, adg, rsc
CC=golang-dev
https://golang.org/cl/5437105
- this removes extra conversions from strings to bytes and vice versa
for each comment
- minor cleanups
R=golang-dev, rsc
CC=golang-dev
https://golang.org/cl/5434096
- scan all comments not just the package documentation
- declutter output so that false positives are more easily spotted
- count the number of headings to quickly see differences
- minor tweaks
R=golang-dev, r, r
CC=golang-dev
https://golang.org/cl/5450061
To structure larger sections of comments in html output headings
are detected in comments and formated as h3 in the generated html.
A simple heuristic is used to detect headings in comments:
A heading is a non-blank, non-indented line preceded by a blank
line. It is followed by a blank and a non-blank, non-indented line.
A heading must start with an uppercase letter and end with a letter,
digit or a colon. A heading may not contain punctuation characters.
R=jan.mercl, gri, adg, rsc, r
CC=golang-dev
https://golang.org/cl/5437056
This is the result of running `gofix -r hashsum` over the tree, changing
the hash function implementations by hand and then fixing a couple of
instances where gofix didn't catch something.
The changed implementations are as simple as possible while still
working: I'm not trying to optimise in this CL.
R=rsc, cw, rogpeppe
CC=golang-dev
https://golang.org/cl/5448065
The problem is that execution can modify the template, so it needs
interlocking to have the same thread-safe guarantee as text/template.
Fixes#2439.
R=golang-dev, adg
CC=golang-dev
https://golang.org/cl/5450056
Not quite done yet but enough is here to review.
Embedding is eliminated so clients can't accidentally reach
methods of text/template.Template that would break the
invariants.
TODO later: Add and Clone are unimplemented.
TODO later: address issue 2349
R=golang-dev, r, rsc
CC=golang-dev
https://golang.org/cl/5434077
Sincos via sincos.go is 35.4 ns/op, via sincos_amd64.s is 37.4 ns/op on 2.53 GHz Intel Core 2 Duo (Mac OS X).
R=rsc, golang-dev
CC=golang-dev
https://golang.org/cl/5447045
This is not the right fix, but it is what used to happen
before the FileInfo conversion, and it should get the
build working again (at least that part).
TBR=brainman
CC=golang-dev
https://golang.org/cl/5434090
time.Parse uses time.Local if it has the right zone offset,
otherwise it calls time.FixedZone. The test's use of reflect.DeepEqual
meant that the test expected time.FixedZone always, failing
when the local time zone really would have used -0700 for
that time. The fix is to format the time to display only the
pieces we intend to test.
R=golang-dev, agl, iant
CC=golang-dev
https://golang.org/cl/5437088
Exec() has been renamed to Run() in keeping with the os/exec API.
Added func (*Session) Start(cmd string) which starts a remote process but unlike Run() doesn't wait for it to finish before returning.
Run() has been refactored to use Start internally. Its really just a refactoring, no new code but some extra functionality was won.
Also added func (*Session) Signal(sig signal) which sends a UNIX signal to a remote process. This is espcially useful in conjunction with Start() as the two allow you to start a remote process, monitor its stdout/stderr, and send it a TERM/HUP/etc signal when you want it to close.
R=dave, rsc, agl, bradfitz, n13m3y3r, gustavo
CC=golang-dev
https://golang.org/cl/5437058
Previously we were using the map iteration order to set the order of
the cipher suites in the ClientHello.
R=bradfitz
CC=golang-dev
https://golang.org/cl/5440048
- allow Lookup to work on uninitialized templates
- fix bug in add: can't error after parser is stopped
- add Add method for html/template
R=adg, rogpeppe, r, rsc
CC=golang-dev
https://golang.org/cl/5436080
The wrong value made Nconv() show "1" for node "-1", and "2" from
node "2+3".
Fixes#2452.
R=gri, lvd, rsc
CC=golang-dev, remy
https://golang.org/cl/5435064
This change adds the second aspect to the conversion code, the
use of large divisiors (powers of big base) to greatly speed up
the divsion of large numbers. Speedups of 30x are common in the
large cases. Also includes new tests and tuning code for the
key internal parameters.
R=gri
CC=golang-dev
https://golang.org/cl/5438058
According to http://www.ietf.org/rfc/rfc4254.txt most channel messages contain the channel id of the recipient channel, not the sender id. This allows the recipient connection multiplexer to route the message to the correct channel.
This changeset fixes several messages that incorrectly send the local channel id instead of the remote channel's id.
While sessions were being created and closed in sequence channels in the channel pool were freed and reused on the server side of the connection at the same rate as was done on the client, so the channel local and remote channel ids always corresponded. As soon as I had concurrent sessions on the same clientConn the server started to complain of 'uknown channel id N' where N is the local channel id, which is actually paired with server channel id K.
R=golang-dev, dave, rsc, agl
CC=golang-dev
https://golang.org/cl/5433063
IIRC, package sql used to be called db. There was one occurrence
of the old name in a comment.
R=golang-dev, bradfitz
CC=golang-dev
https://golang.org/cl/5431075
The Set type is gone. Instead, templates are automatically associated by
being parsed together; nested definitions implicitly create associations.
Only associated templates can invoke one another.
This approach dramatically reduces the breadth of the construction API.
For now, html/template is deleted from src/pkg/Makefile, so this can
be checked in. Nothing in the tree depends on it. It will be updated next.
R=dsymonds, adg, rsc, r, gri, mikesamuel, nigeltao
CC=golang-dev
https://golang.org/cl/5415060
Buffer intermediate output via a bytes.Buffer and thus avoid
calling through the entire Writer stack for every item printed.
There is more opportunity for improvements along the same lines.
Before (best of 3 runs):
- printer.BenchmarkPrint 50 47959760 ns/op
- time gofmt -l $GOROOT/src real 0m11.517s
After (best of 3 runs):
- printer.BenchmarkPrint 50 32056640 ns/op (= -33%)
- time gofmt -l $GOROOT/src real 0m9.070s (= -21%)
R=r
CC=golang-dev
https://golang.org/cl/5432054
The allowed conversions before and after are:
type Tstring string
type Tbyte []byte
type Trune []rune
string <-> string // ok
string <-> []byte // ok
string <-> []rune // ok
string <-> Tstring // ok
string <-> Tbyte // was illegal, now ok
string <-> Trune // was illegal, now ok
Tstring <-> string // ok
Tstring <-> []byte // ok
Tstring <-> []rune // ok
Tstring <-> Tstring // ok
Tstring <-> Tbyte // was illegal, now ok
Tstring <-> Trune // was illegal, now ok
Update spec, compiler, tests. Use in a few packages.
We agreed on this a few months ago but never implemented it.
Fixes#1707.
R=golang-dev, gri, r
CC=golang-dev
https://golang.org/cl/5421057
Pass tests2.dat, test 47:
" \n "
(That is, two spaces separated by a newline)
| <html>
| <head>
| <body>
Also pass tests through test 49:
<!DOCTYPE html><script>
</script> <title>x</title> </head>
R=nigeltao
CC=golang-dev
https://golang.org/cl/5422043
exec_plan9.go:
. Adjusted return argument to match other changes.
#mksyscall.pl:
. Replaced "err = e1" with "err = NewError(e1)".
* Change abandoned, Russ made a better suggestion involving
syscall_plan9.go.
syscall_plan9.go:
. Removed redundant "err = nil" lines.
. Adjusted //sys lines for mksyscall.pl.
* Replaced "err string" with "err ErrorString" in return arguments.
zsyscall_plan9_386.go:
. This module ought to be generated, but as it exists in the
repository, I rebuilt it and checked that it matched expectations.
Anybody is welcome to remove this from the repository if
they feel it should go, but remember that not all Plan 9
installations have a working Perl.
R=rsc
CC=ality, golang-dev
https://golang.org/cl/5411046
Also introduce a new insertion heuristic:
insert new import next to existing import
with the longest matching prefix.
R=golang-dev, adg, gri
CC=golang-dev
https://golang.org/cl/5412053
This CL inverts the direction of the Stdin/out/err members of the
Session struct so they reflect the API of the exec.Cmd. In doing so
it borrows heavily from the exec package.
Additionally Shell now returns immediately, wait for completion using
Wait. Exec calls Wait internally and so blocks until the remote
command is complete.
Credit to Gustavo Niemeyer for the impetus for this CL.
R=rsc, agl, n13m3y3r, huin, bradfitz
CC=cw, golang-dev
https://golang.org/cl/5322055
My excuse for doing this is that test cases with newlines in them didn't
work. But instead of just fixing that, I rearranged everything in
parse_test.go to use fewer channels and pipes, and just call a
straightforward function to read test cases from a file.
R=nigeltao
CC=golang-dev
https://golang.org/cl/5410049
Slight slow-down for printer benchmark (-0.7%) before
applying CL 5416049 (which will wash it out). Code is
cleaner and simpler.
R=rsc
CC=golang-dev
https://golang.org/cl/5417053
Preamble to the simplification of the template API.
Although the signature of Parse (nee Set) changes,
it's really an internal function, used only by
text/template.
R=golang-dev, rsc, gri, r
CC=golang-dev
https://golang.org/cl/5415052
deps.bash does not understand current format,
so make.bash fails (with MAKEFLAGS=-j20).
R=golang-dev, bradfitz
CC=golang-dev
https://golang.org/cl/5410046
Parse {{define}} blocks during template parsing rather than separately as a set-specific thing.
This cleans up set parse significantly, and enables the next step, if we want, to unify the
API for templates and sets.
Other than an argument change to parse.Parse, which is in effect an internal function and
unused by client code, there is no API change and no spec change yet.
R=golang-dev, rsc, r
CC=golang-dev
https://golang.org/cl/5393049
This feature should make it easier to look at very large
directory trees.
- a new mode (URL: /pkg/?m=flat) shows directory listings w/o
indentation and entries with full path (html and text mode)
- in text mode, hierarchical (non-flat) directory listings are
now presented with indentation (/pkg/?m=text)
- in html mode, hierarchical (non-flat) directory listings are
presented with slightly less indentation
- there is an internal hook for programmatic control of the
display mode (for specialized versions of godoc).
R=bradfitz
CC=golang-dev, rsc
https://golang.org/cl/5410043
Work around a bug that was fixed after OpenBSD 5.0 - a request for
kern.hostname or kern.domainname with a nil value for oldp will result
in a length of zero being returned. If we hit this case use a length
of MAXHOSTNAMELEN (256).
R=golang-dev, mikioh.mikioh
CC=golang-dev
https://golang.org/cl/5408041
Move the existing darwin/freebsd specific nametomib implementation
into the respective operating system dependent files.
Provide a nametomib implementation for openbsd, which operates on a
sysctl MIB that has been pre-generated from the various system headers
by mksysctl_openbsd.pl.
R=rsc
CC=golang-dev
https://golang.org/cl/4935044
Making Value opaque means we can drop the interface kludges
in favor of a significantly simpler and faster representation.
v.Kind() will be a prime candidate for inlining too.
On a Thinkpad X201s using -benchtime 10:
benchmark old ns/op new ns/op delta
json.BenchmarkCodeEncoder 284391780 157415960 -44.65%
json.BenchmarkCodeMarshal 286979140 158992020 -44.60%
json.BenchmarkCodeDecoder 717175800 388288220 -45.86%
json.BenchmarkCodeUnmarshal 734470500 404548520 -44.92%
json.BenchmarkCodeUnmarshalReuse 707172280 385258720 -45.52%
json.BenchmarkSkipValue 24630036 18557062 -24.66%
benchmark old MB/s new MB/s speedup
json.BenchmarkCodeEncoder 6.82 12.33 1.81x
json.BenchmarkCodeMarshal 6.76 12.20 1.80x
json.BenchmarkCodeDecoder 2.71 5.00 1.85x
json.BenchmarkCodeUnmarshal 2.64 4.80 1.82x
json.BenchmarkCodeUnmarshalReuse 2.74 5.04 1.84x
json.BenchmarkSkipValue 77.92 103.42 1.33x
I cannot explain why BenchmarkSkipValue gets faster.
Maybe it is one of those code alignment things.
R=iant, r, gri, r
CC=golang-dev
https://golang.org/cl/5373101
Fixes use of c after Dial failure (causes crash).
May fix Dial failure by listening to 127.0.0.1:0
instead of 0.0.0.0:0 (tests should only listen on
localhost).
R=golang-dev, gri
CC=golang-dev
https://golang.org/cl/5395052
I've modified Plan 9's yacc to work with
the grammar in go.y. These are the only
changes necessary on the Go side.
R=rsc
CC=golang-dev
https://golang.org/cl/5375104
The wrong length was being sent, and two parameters
were also transposed. Made the record type be a type
and made the constants typed, to prevent that sort
of bug in the future.
Fixes#2469
R=golang-dev, edsrzf
CC=golang-dev
https://golang.org/cl/5394046
text/template does this (in an entirely different way), so
make html/template do the same. Before this fix, the template
{{.}} given a pointer to a string prints its address instead of its
value.
R=mikesamuel, r
CC=golang-dev
https://golang.org/cl/5370098
Ensure that empty NameLists always return
a zero length []string, not nil.
In practice NameLists are only used in a few
message types and always consumed by a for
range function so the difference between nil
and []string{} is not significant.
Also, add exp/ssh to pkg/Makefile as suggested
by rsc.
R=rsc, agl
CC=golang-dev
https://golang.org/cl/5400042
Move scanner allocation out of loop.
It's the only allocation in the test so it dominates
when it triggers a garbage collection.
R=golang-dev, r
CC=golang-dev
https://golang.org/cl/5369117
Errors in the code under test go to standard output.
Errors in testing or its usage go to standard error.
R=r
CC=golang-dev
https://golang.org/cl/5374090
An experiment: allow structs to be copied even if they
contain unexported fields. This gives packages the
ability to return opaque values in their APIs, like reflect
does for reflect.Value but without the kludgy hacks reflect
resorts to.
In general, we trust programmers not to do silly things
like *x = *y on a package's struct pointers, just as we trust
programmers not to do unicode.Letter = unicode.Digit,
but packages that want a harder guarantee can introduce
an extra level of indirection, like in the changes to os.File
in this CL or by using an interface type.
All in one CL so that it can be rolled back more easily if
we decide this is a bad idea.
Originally discussed in March 2011.
https://groups.google.com/group/golang-dev/t/3f5d30938c7c45ef
R=golang-dev, adg, dvyukov, r, bradfitz, jan.mercl, gri
CC=golang-dev
https://golang.org/cl/5372095
