Reference: https://groups.google.com/forum/#!topic/golang-announce/Zc6kgP2VzhY.

Helps #725.

…ncname.

These are used by the new panicwrap implementation. panicwrap wasn't
implemented previously either, so this is not a regression.

Related to golang/go@354fa9a.

Fixes:

	runtime/error.go:134:8: undeclared name: getcallerpc
	runtime/error.go:135:10: undeclared name: funcname
	runtime/error.go:135:19: undeclared name: findfunc

We can now use the upstream one from package math.

In Go 1.10, it's now implemented directly. Previously, it relied on max
and had assembly implementations, which is why a custom implementation
for GopherJS was needed.

Related to golang/go@b97688d.

Fixes:

	math/math.go:67:9: undeclared name: dim

In Go 1.10, package time has a new LoadLocationFromTZData API addition.
There were some internal code changes, and this change makes appropriate
updates to GopherJS.

Fixes:

	/src/time/time.go:83:9: undeclared name: loadZoneFile
	/usr/local/go/src/time/zoneinfo.go:302:28: undeclared name: zoneSources

It always had the zero value and was unused. It was removed upstream.

Keep name.pkgPath method in order to override the upstream method that
remains with non-trivial implementation (one that would fail).
See https://github.com/golang/go/blob/bfb8f2a765d6097c82b30ae2c19c99fa1082add3/src/reflect/type.go#L515-L529
and https://go-review.googlesource.com/c/go/+/54331/3/src/reflect/type.go#b515.

Related to golang/go@9c9df65 (/cc @mvdan).

Fixes:

	reflect/type.go:1451:45: cannot convert false (untyped bool constant) to string
	reflect/type.go:1451:50: too few arguments in call to newName
	reflect/type.go:1854:45: cannot convert false (untyped bool constant) to string
	reflect/type.go:1854:50: too few arguments in call to newName
	reflect/type.go:1897:45: cannot convert false (untyped bool constant) to string
	reflect/type.go:1897:50: too few arguments in call to newName
	reflect/type.go:2065:47: cannot convert false (untyped bool constant) to string
	reflect/type.go:2065:52: too few arguments in call to newName
	reflect/type.go:2260:44: cannot convert false (untyped bool constant) to string
	reflect/type.go:2260:49: too few arguments in call to newName
	reflect/type.go:2260:49: too many errors

…m uint64 to int.

It has changed upstream.

Related to golang/go@8a6e51a.

Fixes:

	reflect/value.go:2119:31: cannot use buffer (variable of type int) as uint64 value in argument to makechan

syscall.Exit implementation has been removed upstream. In Go 1.10, it's
provided via runtime.exit. We can't do that, so just implement
syscall.Exit ourselves as it was previously implemented.

Related to golang/go@438c8f6.

Fixes:

	Error: runtime error: native function not implemented: syscall.Exit

Copy semaphore implementation from sync package. In Go 1.10,
these are used in FD.destroy and FD.Close methods.

Related to golang/go@382d492.

Fixes:

	Error: runtime error: native function not implemented: internal/poll.runtime_Semrelease

The behavior or reflect.Copy has changed upstream to somewhat mirror
the special case behavior of the copy built-in:

	// As a special case, src can have kind String
	// if the element type of dst is kind Uint8.

Related to golang/go@245e386.

Fixes:

	Error: reflect: call of reflect.Copy on string Value

…ded fields.

Update for upstream changes.

Related to golang/go@6471ace.
Related to golang/go@d349fa2.

Fixes:

	=== RUN   TestCallPanic
	--- FAIL: TestCallPanic (0.01s)

	/Users/Dmitri/Dropbox/Work/2013/GoLand/src/github.com/gopherjs/gopherjs/
	test.043844921:1412
	          throw new Error(msg);
	                ^
	Error: did not panic

…ported methods.

Update overridden method to keep up to date with an upstream change.

Related to golang/go@6f1724f.

At least this helps narrow down where the issue was. Need to come up
with a general, correct fix.

Related to golang/go@6f1724f.

…ypes.

I don't see a good reason to skip string internalization in reflect
package. It was introduced as part of other changes for Go 1.7 support
in commit f1514b4, but no rationale
was provided.

Skipping internalization is inconsistent and causes issues because
string comparison no longer works as expected.

Specifically, the test TestMethodByNameUnExportedFirst was failing with:

	got , expected ΦExported

The reason it was failing is because the test makes a
typ.MethodByName("ΦExported") call. Once that string is internalized,
it becomes "\xCE\xA6Exported".

As a result, the code in rtype.MethodByName was failing to find the
method by name, because the internalized and non-internalized strings
did not match:

	have: unexported want: ΦExported match: false
	have: ΦExported want: ΦExported match: false

Removing this special behavior seems to not have any adverse effects,
fixes the issue and makes things simpler:

	have: unexported want: ΦExported match: false
	have: ΦExported want: ΦExported match: true

It uses runtime.SetFinalizer, which is not supported by GopherJS.

…verted nil interface.

Not yet sure if this is the right/best fix, needs investigation. But it
allows the TestCallConvert test to pass without breaking any others.

Related to golang/go@eafa29b.

Fixes:

	$ gopherjs test -v --run=TestCallConvert reflect
	=== RUN   TestCallConvert
	--- FAIL: TestCallConvert (0.10s)
		all_test.go:1639: expected [nil], got [<io.Reader Value>]
	FAIL
	FAIL	reflect	0.808s

…lers.

frameSkip relies on runtime.Callers when testing.T.Helper was called.

Fixes:

	$ gopherjs test --run=TestBuilder strings
	--- FAIL: TestBuilder (0.07s)

	github.com/gopherjs/gopherjs/test.481734121:1412
	          throw new Error(msg);
	                ^
	Error: testing: zero callers found

…chitecture.

Can't use package unsafe, so use a simple type conversion from []byte
to string in Builder.String method.

Further optimizations are deferred for later (most likely not needed
because Go->wasm will be ready by the time they're needed).

Related to golang/go#18990.

It's being used in a few places, and previously had no implementation
in GopherJS. (In upstream Go, it's being provided by runtime via
go:linkname directive.)

This restores more helpful error messages as in GopherJS 1.9.

Before:

	runtime error: native function not implemented: sync.throw

After:

	runtime error: sync: inconsistent mutex state

…on for types."

This reverts commit 2f2d5b3.

That commit (and its commit message) was wrong/incomplete. Now I
understand there was a reason for skipping string internalization
for types. The following commit is a better alternative fix.

It has been removed in go1.10beta2.

…gesize.

Fixes:

	--- FAIL: TestIndexByteNearPageBoundary (0.00s)
		Error: runtime error: native function not implemented: syscall.Getpagesize
	FAIL	bytes	19.341s

Reference: https://groups.google.com/forum/#!topic/golang-announce/Pt6Zra9OY3s.

Reference: https://groups.google.com/forum/#!topic/golang-announce/X7N1mvntnoU.

Since the package files are excluded too, the tests can't compile.

Fixes:

	../../../usr/local/go/src/crypto/rand/rand_linux_test.go:13:17: undeclared name: batched
	../../../usr/local/go/src/crypto/rand/rand_linux_test.go:31:7: undeclared name: batched
	../../../usr/local/go/src/crypto/rand/rand_linux_test.go:38:7: undeclared name: batched

This helps know which parts of the package can be relied on, or not.

It's being done upstream to trick the escape analysis and prevent
allocations. We can't rely on it here.

Fixes:

	=== RUN   TestBuilder
	--- FAIL: TestBuilder (0.01s)

	TypeError: Cannot create property '$proxies' on number '0'

TestMultiWriterSingleChainFlatten was added in Go 1.10, and it needs to
be skipped (just like TestMultiReaderFlatten) because it relies on
unsupported runtime features.

Related to golang/go@060d1a5.

Fixes:

	=== RUN   TestMultiWriterSingleChainFlatten
	--- FAIL: TestMultiWriterSingleChainFlatten (0.01s)
	    test.558082205:38: multiWriter did not flatten chained multiWriters: expected writeDepth 12, got 4
	FAIL    io  0.888s

TestTypeRace was added in Go 1.10, and it needs to be skipped because
it relies on sync.WaitGroup to behave correctly under concurrency.
However, encoding/gob is currently configured to use nosync, which
cannot succeed.

Since this is a new test, opt to skip it for now rather than making it
succeed by replacing nosync with sync.

Related to golang/go@1a9f27d.

Fixes:

	=== RUN   TestTypeRace
	--- FAIL: TestTypeRace (0.00s)
		Error: sync: WaitGroup counter not zero
	FAIL    encoding/gob    3.030s

Reference: https://groups.google.com/forum/#!topic/golang-announce/IQeqMp44Qok.

…itly converted nil interface."

This reverts commit 1fb1aaa.

The change in that commit was not a complete and correct solution.
Instead of resolving the underlying problem, it masked a symptom,
allowing the TestCallConvert test to pass.

A better fix will follow next.

…Kind() == Interface && v.IsNil().

This change adds a copy of Value.assignTo method to reflect package,
identical to upstream, except with golang/go@eafa29b
not applied.

That commit adds a short-circuit check to that skips ifaceE2I, in order
to avoid a panic that occurs in ifaceE2I under that condition in gc and
gccgo compilers.

GopherJS implements a custom version of ifaceE2I which needs to run and
does not panic in that condition. So, there's no need skip it.

Fixes:

	$ gopherjs test --short reflect html/template
	--- FAIL: TestCallConvert (0.01s)
		all_test.go:1639: expected [nil], got [<io.Reader Value>]
	FAIL
	FAIL	reflect	3.084s
	--- FAIL: TestEscapingNilNonemptyInterfaces (0.01s)

	go/src/github.com/shurcooL/play/241/test.243863553:9608
				if (!(($parseInt(methodSet.length) === 0)) || !!(typ.named)) {
	                              ^
	TypeError: Cannot read property 'length' of undefined
	    at reflectType (go/src/github.com/shurcooL/play/241/test.243863553:9608:31)
	    at Object.TypeOf (go/src/github.com/shurcooL/play/241/reflect.go:312:3)
	    at indirect (/html/template/content.go:119:6)
	    at stringify (/html/template/content.go:153:4)
	    at htmlEscaper (/html/template/html.go:43:3)
	    at $call (go/src/github.com/shurcooL/play/241/test.243863553:30:50)
	    at go/src/github.com/shurcooL/play/241/test.243863553:1948:22
	    at Object.$packages.reflect.Value.ptr.call (go/src/github.com/shurcooL/play/241/reflect.go:880:3)
	    at Object.$packages.reflect.Value.ptr.Call (/reflect/value.go:308:3)
	    at Object.$packages.text/template.state.ptr.evalCall (/text/template/exec.go:667:3)
	    at Object.$packages.text/template.state.ptr.evalFunction (/text/template/exec.go:535:3)
	    at Object.$packages.text/template.state.ptr.evalCommand (/text/template/exec.go:432:4)
	    at Object.$packages.text/template.state.ptr.evalPipeline (/text/template/exec.go:405:4)
	    at Object.$packages.text/template.state.ptr.walk (/text/template/exec.go:231:4)
	    at Object.$packages.text/template.state.ptr.walk (/text/template/exec.go:239:5)
	    at Object.$packages.text/template.Template.ptr.execute (/text/template/exec.go:194:3)
	    at Object.$packages.text/template.Template.ptr.Execute (/text/template/exec.go:177:3)
	    at Object.$packages.html/template.Template.ptr.Execute (/html/template/template.go:122:3)
	    at Object.TestEscapingNilNonemptyInterfaces [as $blk] (/html/template/content_test.go:448:3)
	    at Object.tRunner [as $blk] (/testing/testing.go:777:3)
	    at fun (go/src/github.com/shurcooL/play/241/test.243863553:1479:37)
	    at $goroutine (go/src/github.com/shurcooL/play/241/test.243863553:1477:19)
	    at $runScheduled (go/src/github.com/shurcooL/play/241/test.243863553:1517:7)
	    at $schedule (go/src/github.com/shurcooL/play/241/test.243863553:1533:5)
	    at $go (go/src/github.com/shurcooL/play/241/test.243863553:1509:3)
	    at Object.<anonymous> (go/src/github.com/shurcooL/play/241/test.243863553:55363:1)
	    at Object.<anonymous> (go/src/github.com/shurcooL/play/241/test.243863553:55366:4)
	    at Module._compile (module.js:660:30)
	    at Object.Module._extensions..js (module.js:671:10)
	    at Module.load (module.js:573:32)
	    at tryModuleLoad (module.js:513:12)
	    at Function.Module._load (module.js:505:3)
	    at Function.Module.runMain (module.js:701:10)
	    at startup (bootstrap_node.js:190:16)
	    at bootstrap_node.js:662:3
	FAIL	html/template	1.079s

This reverts commit c4db7b2.

That commit was a temporary hacky fix to help find out where the issue
was. It's not a general fix. Revert it, and file issue #763 to resolve
this issue fully and correctly.

go generate ./...