Russ Cox [Wed, 2 Nov 2011 01:45:21 +0000 (21:45 -0400)]
gofix: error fix
To make the error fix more useful, expand typecheck to gather
more information about struct fields, typecheck range statements,
typecheck indirect and index of named types, and collect information
about assignment conversions.
Also, change addImport to rename top-level uses of a to-be-imported
identifier to avoid conflicts. This duplicated some of the code in
the url fix, so that fix is now shorter.
R=iant, r, r
CC=golang-dev
https://golang.org/cl/5305066
Andrew Balholm [Wed, 2 Nov 2011 00:47:05 +0000 (11:47 +1100)]
html: stop at scope marker node when generating implied </a> tags
A <a> tag generates implied end tags for any open <a> elements.
But it shouldn't do that when it is inside a table cell the the open <a>
is outside the table.
So stop the search for an open <a> when we reach a scope marker node.
Pass tests1.dat, test 78:
<a href="blah">aba<table><tr><td><a href="foo">br</td></tr>x</table>aoe
Back then, every x.go file in a package became an x.6 file
in the archive. It was important to be able to allow the
use of long Go source file names, hence the increase in size.
Today, all Go source files compile into a single _go_.6 file
regardless of their names, so the archive file name length
no longer needs to be long. The longer name causes some
problems on Plan 9, where the native archive format is the
same but with 16-byte names, so revert back to 16.
R=golang-dev, r
CC=golang-dev
https://golang.org/cl/5333050
Andrew Balholm [Tue, 1 Nov 2011 00:42:54 +0000 (11:42 +1100)]
html: process </td> tags; foster parent at most one node per token
Correctly close table cell when </td> is read.
Because of reconstructing the active formatting elements, more than one
node may be created when reading a single token.
If both nodes are foster parented, they will be siblings, but the first
node should be the parent of the second.
Pass tests1.dat, test 77:
<a href="blah">aba<table><a href="foo">br<tr><td></td></tr>x</table>aoe
Scott Lawrence [Mon, 31 Oct 2011 23:07:17 +0000 (16:07 -0700)]
template: fix error checking on execute without parse
Fixed error checking in exec.go to give a sensible error message when
execution is attempted before a successful parse (rather than an
outright panic).
Luuk van Dijk [Mon, 31 Oct 2011 17:09:40 +0000 (18:09 +0100)]
gc: clean up printing.
Got rid of all the magic mystery globals. Now
for %N, %T, and %S, the flags +,- and # set a sticky
debug, sym and export mode, only visible in the new fmt.c.
Default is error mode. Handle h and l flags consistently with
the least side effects, so we can now change
things without worrying about unrelated things
breaking.
Jan Newmarch [Mon, 31 Oct 2011 15:33:14 +0000 (11:33 -0400)]
misc/emacs/go-lang.el: Fix restoration of multiple windows in a frame after gofmt
If a frame has multiple windows then the windows must all be restored
after gofmt has finished and the old windows must be restored.
Before this fix, only the Go code edit window would be left.
Dave Cheney [Sat, 29 Oct 2011 18:22:30 +0000 (14:22 -0400)]
exp/ssh: fix length header leaking into channel data streams.
The payload of a data message is defined as an SSH string type,
which uses the first four bytes to encode its length. When channelData
and channelExtendedData were added I defined Payload as []byte to
be able to use it directly without a string to []byte conversion. This
resulted in the length data leaking into the payload data.
This CL fixes the bug, and restores agl's original fast path code.
Additionally, a bug whereby s.lock was not released if a packet arrived
for an invalid channel has been fixed.
Finally, as they were no longer used, I have removed
the channelData and channelExtedendData structs.
Andrew Balholm [Fri, 28 Oct 2011 23:51:59 +0000 (10:51 +1100)]
html: adjust bookmark in "adoption agency" algorithm
In the adoption agency algorithm, the formatting element is sometimes
removed from the list of active formatting elements and reinserted at a later index.
In that case, the bookmark showing where it is to be reinserted needs to be moved,
so that its position relative to its neighbors remains the same
(and also so that it doesn't become out of bounds).
Also pass tests through test 76:
<test attribute---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------->
Russ Cox [Fri, 28 Oct 2011 02:42:32 +0000 (19:42 -0700)]
crypto/tls: add Error method to alert
alerts get used as both values and errors.
Rather than introduce an alertError wrapper,
this CL just adds an Error method, which will
satisfy the error interface when the time comes.
Russ Cox [Fri, 28 Oct 2011 02:39:25 +0000 (19:39 -0700)]
go/doc: remove os.NewError anti-heuristic
It will be obsolete when error happens.
Submitting this now will make the error transition earlier,
at the cost of making a locally-built godoc viewing
/pkg/syscall or /pkg/os have some functions appear
under the Error type as constructors.
Russ Cox [Fri, 28 Oct 2011 01:04:12 +0000 (18:04 -0700)]
runtime: lock the main goroutine to the main OS thread during init
We only guarantee that the main goroutine runs on the
main OS thread for initialization. Programs that wish to
preserve that property for main.main can call runtime.LockOSThread.
This is what programs used to do before we unleashed
goroutines during init, so it is both a simple fix and keeps
existing programs working.
Julian Phillips [Thu, 27 Oct 2011 08:45:07 +0000 (17:45 +0900)]
goinstall: More intelligent vcs selection for common sites
goinstall has built in support for a few common code hosting sites. The
identification of which vcs tool should be used was based purely on a
regex match against the provided import path. The problem with this
approach is that it requires distinct import paths for different vcs
tools on the same site.
Since bitbucket has recently starting hosting Git repositories under the
same bitbucket.org/user/project scheme as it already hosts Mercurial
repositories, now would seem a good time to take a more flexible
approach.
We still match the import path against a list of regexes, but now the
match is purely to distinguish the different hosting sites. Once the
site is identified, the specified function is called with the repo and
path matched out of the import string. This function is responsible for
creating the vcsMatch structure that tells us what we need to download
the code.
For github and launchpad, only one vcs tool is currently supported, so
these functions can simply return a vcsMatch structure. For googlecode,
we retain the behaviour of determing the vcs from the import path - but
now it is done by the function instead of the regex. For bitbucket, we
use api.bitbucket.org to find out what sort of repository the specified
import path corresponds to - and then construct the appropriate vcsMatch
structure.
Ron Minnich [Wed, 26 Oct 2011 22:27:59 +0000 (15:27 -0700)]
cc: change cas to newcase
Change the name of cas() in cc to newcase() to avoid a NIX conflict.
cas() is used in cc to create a new Case struct. There is a name
conflict in that cas() is a commonly-used
name for compare and swap. Since cas() is only used internally
in the compiler in 3 places, change the name to avoid a wider
conflict with the NIX runtime. This issue might well come up on
other OSes in the future anyway, as the name is fairly common.
Adam Langley [Wed, 26 Oct 2011 14:41:24 +0000 (10:41 -0400)]
crypto/rsa: change public exponent from 3 to 65537
Although there's still no concrete security reason not to use 3, I
think Bleichenbacher has convinced me that it's a useful defense and
it's what everyone else does.