- made GoroutineStacktrace a method of struct G
- made stacktrace a method of StackIterator
- renamed StackIterator to stackIterator
- factored out logic to obtain a stackIterator from a goroutine that's
used by both (*G).Stacktrace and by (*G).UserCurrent

Instead of returning an error when FDE of a frame can not be found,
just truncate the stack trace.

Fixes #462

On linux kernel 2.6.x, the Trace/Stop status is 'T'

Fixes #209

Automatically sets a breakpoint on runtime.startpanic, the function
that gets called by runtime.dopanic when a panic is not recovered.

Implements #317

Caches the mapping of type names to offset in debug_info to speed up
variable evaluation.

BEFORE:
	BenchmarkArray-4         	     100	  13'238'441 ns/op	   0.62 MB/s
	BenchmarkArrayPointer-4  	     200	  10'044'093 ns/op	   0.87 MB/s
	BenchmarkMap-4           	    1000	   1'332'530 ns/op	   0.77 MB/s
	BenchmarkGoroutinesInfo-4	      10	 114'677'462 ns/op
	BenchmarkLocalVariables-4	    2000	   1'223'975 ns/op
AFTER:
	BenchmarkArray-4         	     200	   9'925'686 ns/op	   0.83 MB/s
	BenchmarkArrayPointer-4  	     100	  11'143'930 ns/op	   0.78 MB/s
	BenchmarkMap-4           	    2000	   1'302'520 ns/op	   0.79 MB/s
	BenchmarkGoroutinesInfo-4	      30	  35'079'549 ns/op
	BenchmarkLocalVariables-4	    1000	   1'137'299 ns/op

Note in particular the speedup of BenchmarkGoroutinesInfo, since
proc.(*Variable).parseG is a function we call a lot.

The go compiler changed and break statements no longer get compiled to
nothing when -N is passed:

https://go-review.googlesource.com/#/c/19848/
https://go-review.googlesource.com/#/c/19854/

Lists all the types defined in the debugged program.

Fixes #414

Fixes #419 (partial)

Unrelated to conversion, I have also changed (*Thread).readMemory
to return only first count bytes of memory just as advised
by ReadProcessMemory.

Fixes #409
Fixes #412
Fixes #416

Moved the code in testvariables2 and testvariables4 into
testvariables3, renamed testvariables3 into testvariables2

Step disassembles the current instruction, if it is a CALL sets a
temp breakpoint inside the called function, after the prologue and
calls Continue.

Fixes #332

- Unlike FunctionEntryToFirstLine can skip the prologue on functions
that are defined on a single line, either because they weren't
formatted or because they were autogenerated
- Can skip the prologue on most functions when setting a breakpoint
with the filename:line syntax

Fixes #396

Implements #109 and #120

Implements #368

1. A running goroutine is by definition not parked waiting for a
chan recv
2. The FDE end address is intended to be exclusive, the code
interpreted as inclusive and sometimes ended up setting a breakpoint
on a function other than the current one.

Past the maximum recursion depth maps shouldn't be loaded at all,
adding map children and not loading them breaks assumptions in
the prettyprinter.

Fixes #406

I was hoping this will fix the issue below.
But there is more code that break new cgo rules.

Fixes #409 (partial)

Fixes #398.

Fixes #370.

Fixes #384

Fixes #355

If uninitialized memory is read loadArrayValues could try to call
cacheMemory with a negative size, which could cause a 'makeslice:
len out of range' panic.

Fixes #354 (partial)

Fixes #370.

Fixes #361

This patch modifies the `step` command to step to the next source line,
stepping into any function encountered along the way.

Fixes #360

Typedefs that resolve to slices are not recorded in DWARF as typedefs
but instead as structs in a way that there is no way to know they
are really slices using debug/dwarf.
Using golang.org/x/debug/dwarf instead this problem is solved and
as a bonus some types are printed with a nicer names: (struct string
→ string, struct []int → []int, etc)

 Fixes #356 and #293

Instead of doing a bunch of manual DWARF / memory parsing we instead can
just dogfood the internal Variable API to retrieve the G struct.

Backend only, no UI

Implements #120 (partial)