Still need to write tests for new syntax
and fix bugs that the tests find, but this
is a good check point.

All tests pass.

Compared against existing regexp:

benchmark                                old ns/op    new ns/op    delta
regexp.BenchmarkLiteral                       1869          620  -66.83%
regexp.BenchmarkNotLiteral                    9489         7823  -17.56%
regexp.BenchmarkMatchClass                   10372         8386  -19.15%
regexp.BenchmarkMatchClass_InRange           10800         7750  -28.24%
regexp.BenchmarkReplaceAll                   13492         8519  -36.86%
regexp.BenchmarkAnchoredLiteralShortNonMatch   747          339  -54.62%
regexp.BenchmarkAnchoredLiteralLongNonMatch    599          335  -44.07%
regexp.BenchmarkAnchoredShortMatch            2137          917  -57.09%
regexp.BenchmarkAnchoredLongMatch             2029          917  -54.81%

R=r, r
CC=golang-dev, sam.thorogood
https://golang.org/cl/4820046

goinstall: report every newly installed package to the dashboard

This makes "goinstall -a" work on systems with GOROOTs that are
not user-writable, as is the case with Debian's Go packages.

This also makes goinstall.log the canonical list of installed
packages, in that only packages new to goinstall.log are reported to
the dashboard.

A side-effect is that writing to goinstall.log is now mandatory.
(A bug in the original implementation meant this was the case, anyway.)

The principal benefit of this change is that multiple packages from the
same repository can now be reported to the dashboard.  It is also less
likely for a user to report multiple installations of the same package
to the dashboard (they would need to remove the package from
goinstall.log first).

R=rsc, n13m3y3r
CC=golang-dev
https://golang.org/cl/4786041

R=golang-dev, r
CC=golang-dev
https://golang.org/cl/4811050

Previously comparisons with NaN led to contradictory results if it was
compared to anything not NaN, since Less always returned false, thus
breaking monotonicity of ordering.
This fix makes NaN less than anything else and adds NaN and (+-)Inf to
testcases.

Fixes #2092.

R=golang-dev, r, rsc, r
CC=golang-dev
https://golang.org/cl/4805051

R=golang-dev, r, r, dsymonds, rsc
CC=golang-dev
https://golang.org/cl/4813043

Replace cas with xadd in scheduler.
Suggested by Dmitriy in last code review.
Verified with Promela model.

When there's actual contention for the atomic word,
this avoids the looping that compare-and-swap requires.

benchmark                            old ns/op    new ns/op    delta
runtime_test.BenchmarkSyscall               32           26  -17.08%
runtime_test.BenchmarkSyscall-2            155           59  -61.81%
runtime_test.BenchmarkSyscall-3            112           52  -52.95%
runtime_test.BenchmarkSyscall-4             94           48  -48.57%
runtime_test.BenchmarkSyscallWork          871          872   +0.11%
runtime_test.BenchmarkSyscallWork-2        481          477   -0.83%
runtime_test.BenchmarkSyscallWork-3        338          335   -0.89%
runtime_test.BenchmarkSyscallWork-4        263          256   -2.66%

R=golang-dev, iant
CC=golang-dev
https://golang.org/cl/4800047

R=golang-dev, rsc
CC=golang-dev
https://golang.org/cl/4794056

This is the last piece (I hope) of the set creation code.
These helpers create sets from files containing individual
template definitions, free of {{define}} clauses. This
design is helpful if the templates live one per file,
undecorated.

R=golang-dev, dsymonds
CC=golang-dev
https://golang.org/cl/4801052

R=r, iant
CC=golang-dev
https://golang.org/cl/4809050

Fixes a problem where Readdir would always return
empty directories (Readdir is only called by godoc
if the usual directory handler is commented out
in godoc.go, and if a zip file system is provided;
thus, this bug never manifested itself in godoc).

Also:
- better choice of variable/field names
- simplified error handling a bit
- better comments

R=bradfitz
CC=golang-dev
https://golang.org/cl/4813047

This surprisingly takes 30 seconds on my fast machine
so disabling by default. Need to optimize the Writer
at some point.

R=golang-dev, r, r
CC=golang-dev
https://golang.org/cl/4815048

This makes the full file paths recorded by 6g
prefer $PWD over the actual directory name
(relevant when $PWD gets to the current directory
via symlinks).  It's what everyone else does, and
what people expect.

R=iant, r, mattn.jp
CC=golang-dev
https://golang.org/cl/4824041

R=golang-dev, mattn.jp, rsc
CC=golang-dev
https://golang.org/cl/4794052

Fixes #2052.

R=r
CC=golang-dev
https://golang.org/cl/4812053

import cycle:
        "testing"
        imports "flag"
        imports "sort"
        imports "testing"

R=golang-dev, r
CC=golang-dev
https://golang.org/cl/4811048

Don't report that a directory was found just because we found
the list index where the directory would be if it were there...

R=iant
CC=golang-dev
https://golang.org/cl/4812051

Also:
- Add parser.SpuriousError flag. If set, the parser reports all (including
  spurious) errors rather then at most one error per line.
- Add -e flag to gofmt and gotype: If set, gofmt and gotype report all
  (including spurious) errors rather than at most one error per line.
- Updated the respective documentation.

Fixes #2088.

R=rsc
CC=golang-dev
https://golang.org/cl/4803047

- remove calls to print
- make units compile again
- make units.y closer to gofmt style

R=golang-dev, dsymonds
CC=golang-dev
https://golang.org/cl/4802052

Clean up glitch created by flurry of editing.

R=golang-dev, dsymonds
CC=golang-dev
https://golang.org/cl/4808052

R=golang-dev, dsymonds
CC=golang-dev
https://golang.org/cl/4805048

R=r
CC=golang-dev
https://golang.org/cl/4812048

Drops mallocrep1.go back to a reasonable
amount of time.  (154 -> 0.8 seconds on my Mac)

Fixes #2085.

R=golang-dev, dvyukov, r
CC=golang-dev
https://golang.org/cl/4811045

Also make the Set.Parse methods variadic so you can parse static lists of files without loops.

R=rsc, dsymonds, r
CC=golang-dev
https://golang.org/cl/4802051

R=gri, r, r, rsc
CC=golang-dev
https://golang.org/cl/4794047

This is already the behavior for pipelines producing values for if, with, and range.

R=golang-dev, gri
CC=golang-dev
https://golang.org/cl/4808050

R=r
CC=golang-dev
https://golang.org/cl/4807043

R=mtj, r
CC=golang-dev
https://golang.org/cl/4814044

Now handles standard precision specifications, standard interactions of
redundant specifications (such as precision and zero-fill), handles the
special case of precision specified but equal to zero, and generates the
output without recursive calls to format/printf to be clearer and faster.

R=gri, mtj, gri
CC=golang-dev
https://golang.org/cl/4703050

R=golang-dev, bradfitz, gri
CC=golang-dev
https://golang.org/cl/4807044

R=ken2
CC=golang-dev
https://golang.org/cl/4794049

Make selectsend() accept pointer to the element,
it makes it possible to make Scase fixed-size
and allocate/free Select, all Scase's and all SudoG at once.
As a consequence SudoG freelist die out.

benchmark                       old,ns/op  new,ns/op
BenchmarkSelectUncontended	     1080        558
BenchmarkSelectUncontended-2	      675        264
BenchmarkSelectUncontended-4	      459        205
BenchmarkSelectContended	     1086        560
BenchmarkSelectContended-2	     1775       1672
BenchmarkSelectContended-4	     2668       2149
(on Intel Q6600, 4 cores, 2.4GHz)

benchmark                       old ns/op    new ns/op    delta
BenchmarkSelectUncontended         517.00       326.00  -36.94%
BenchmarkSelectUncontended-2       281.00       166.00  -40.93%
BenchmarkSelectUncontended-4       250.00        83.10  -66.76%
BenchmarkSelectUncontended-8       107.00        47.40  -55.70%
BenchmarkSelectUncontended-16       67.80        41.30  -39.09%
BenchmarkSelectContended           513.00       325.00  -36.65%
BenchmarkSelectContended-2         699.00       628.00  -10.16%
BenchmarkSelectContended-4        1085.00      1092.00   +0.65%
BenchmarkSelectContended-8        3253.00      2477.00  -23.85%
BenchmarkSelectContended-16       5313.00      5116.00   -3.71%
(on Intel E5620, 8 HT cores, 2.4 GHz)

R=rsc, ken
CC=golang-dev
https://golang.org/cl/4811041

ServeMux depends on having a URL
in order to mux.  It might be that the right
fix is to have CONNECT handlers just not
look at URL.

««« original CL description
http: do not parse req.URL for CONNECT

CONNECT's argument is not a URL.

R=golang-dev, bradfitz, rsc
CC=golang-dev
https://golang.org/cl/4808044

»»»

TBR=bradfitz
CC=golang-dev
https://golang.org/cl/4798046

CONNECT's argument is not a URL.

R=golang-dev, bradfitz, rsc
CC=golang-dev
https://golang.org/cl/4808044

It makes it clear what benchmark is currently running.
Especially useful in case of hangup or crash.

R=golang-dev, rsc
CC=golang-dev
https://golang.org/cl/4816043

Use atomic.LoadUint32(&done) instead of
atomic.AddInt32(&done, 0) on fast path.

benchmark            old ns/op    new ns/op    delta
BenchmarkOnce            13.40         7.26  -45.82%
BenchmarkOnce-2          22.90         4.04  -82.36%
BenchmarkOnce-4          25.60         2.16  -91.56%
BenchmarkOnce-8          25.80         1.38  -94.65%
BenchmarkOnce-16         24.40         1.33  -94.55%

(on HP Z600, 2 x Intel Xeon E5620, 8 HT cores, 2.4 GHz)

R=golang-dev, rsc
CC=golang-dev
https://golang.org/cl/4787041

R=golang-dev, rsc
CC=golang-dev
https://golang.org/cl/4809042

Something is broken, and investigation is underway.
In the meantime, godoc is broken, so disable sniffing for now
by reverting to the pre-sniffer state.

R=r
CC=golang-dev
https://golang.org/cl/4809046

R=golang-dev, r
CC=golang-dev
https://golang.org/cl/4794046

The C-stdlib heritage of printf/fprintf/sprintf has two odd
aspects for precisions of zero with integers. First, the zero
can be specified in any of these ways, "%4.0d", "%.0d" and
"%.d" which was not previously supported here. Secondly, the
seemingly universal interpretation of precision for integers
is that precision==0 and value==0 means print nothing at all.
The code here now handles this for integers just as the code
in big/int.c does the same for the Int type. New tests are
added to fmt_test.go to verify these changes.

R=r, r
CC=golang-dev
https://golang.org/cl/4717045

This simplifies the API and makes it easier to make the template
invocation statically secure, at the cost of some minor flexibility.

R=golang-dev, dsymonds, r
CC=golang-dev
https://golang.org/cl/4794045