* change ldt0setup to set GS itself; nacl won't let us do it.
  * change breakpoint to INT $3 so 8l can translate to HLT for nacl.
  * panic if closure is needed on nacl.
  * do not try to access symbol table on nacl.
  * mmap in 64kB chunks.

nacl support:
  * system calls, threading, locks.

R=r
DELTA=365  (357 added, 5 deleted, 3 changed)
OCL=34880
CL=34906

but too slow across the board.

R=rsc
DELTA=315  (50 added, 219 deleted, 46 changed)
OCL=34868
CL=34902

R=rsc
DELTA=57  (57 added, 0 deleted, 0 changed)
OCL=34896
CL=34901

go/test: passes 64% (215/337) tests

metric updated, had total number of tests wrong. number of
failing tests is down from 129 to 122.

R=rsc
APPROVED=rsc
DELTA=228  (12 added, 5 deleted, 211 changed)
OCL=34897
CL=34899

R=r
DELTA=18  (0 added, 0 deleted, 18 changed)
OCL=34894
CL=34894

R=r
DELTA=515  (515 added, 0 deleted, 0 changed)
OCL=34881
CL=34891

similar tweaks to make debug/proc, net, os build.

R=r
DELTA=861  (855 added, 4 deleted, 2 changed)
OCL=34877
CL=34890

deciding what the host process support is.
this makes a cross-compiling (e.g., GOOS=nacl) build
still generate valid host debugger binaries.

R=r
DELTA=5  (0 added, 0 deleted, 5 changed)
OCL=34878
CL=34889

	add jmp to constant pc.
	generate HLT for INT $3
	do not insert NOPs between REP/REPN and subsequent instruction.
	allow very long time for convergence.

R=ken
OCL=34879
CL=34879

R=r
DELTA=305  (305 added, 0 deleted, 0 changed)
OCL=34848
CL=34875

R=rsc
OCL=34871
CL=34871

R=r
DELTA=89  (82 added, 5 deleted, 2 changed)
OCL=34870
CL=34870

can JMP or CALL indirect through a register R
provided the preceding instruction is AND $~31, R.

R=ken
OCL=34863
CL=34867

required for nacl and may be nicer for ffi,
because %gs is the standard register for thread-local storage.

R=ken
OCL=34861
CL=34866

R=ken
OCL=34859
CL=34865

R=r
DELTA=2  (0 added, 0 deleted, 2 changed)
OCL=34860
CL=34864

delete strings.Buffer
add a test for a bug not caught before (mustn't install zero-length blocks)

R=rsc
DELTA=987  (289 added, 587 deleted, 111 changed)
OCL=34850
CL=34850

make strings.Buffer handle strings and bytes with comparable efficiency.
if ok, next step will be to move this code to bytes.Buffer and terminate
strings.Buffer's short happy life.

R=rsc
DELTA=292  (212 added, 0 deleted, 80 changed)
OCL=34837
CL=34849

R=ken
OCL=34845
CL=34845

libpng.org.

R=rsc
APPROVED=r
DELTA=1176  (1175 added, 1 deleted, 0 changed)
OCL=34727
CL=34838

- removed some unused code

R=rsc
DELTA=103  (84 added, 15 deleted, 4 changed)
OCL=34816
CL=34829

R=ken
DELTA=1  (0 added, 0 deleted, 1 changed)
OCL=34828
CL=34828

R=kaib
OCL=34823
CL=34823

R=rsc
APPROVED=rsc
DELTA=300  (297 added, 0 deleted, 3 changed)
OCL=34813
CL=34821

R=rsc
APPROVED=rsc
DELTA=7  (3 added, 0 deleted, 4 changed)
OCL=34661
CL=34820

  - ELF header bits and addresses
  - do not let instructions span 32-byte boundary
  - align CALLs so return is to 32-byte boundary
  - align indirect jump targets to 32-byte boundary
    (only possible indirect jumps are function entries)

still to do:
  - replace indirect jump, indirect call, and ret with
    nacl-approved instruction sequences
  - switch to GS segment for m-local storage

R=ken
OCL=34818
CL=34818

having one copy will cut nacl changes in half.

R=ken
OCL=34815
CL=34815

params. a number of fixes to assembly routines that assumed R0
had the first arg. one stack offset fix, arm pushes the link
register on stack top.

go/test: passes 65% (235/364) tests

R=rsc
APPROVED=rsc
DELTA=20  (11 added, 0 deleted, 9 changed)
OCL=34809
CL=34812

TBR=r
OCL=34803
CL=34803

R=rsc
APPROVED=rsc
DELTA=115  (93 added, 14 deleted, 8 changed)
OCL=34798
CL=34802

R=rsc
APPROVED=rsc
DELTA=1  (0 added, 1 deleted, 0 changed)
OCL=34799
CL=34801

TBR=r
OCL=34800
CL=34800

- fixed broken link to function literals section
- minor adjustments

R=r
DELTA=20  (1 added, 3 deleted, 16 changed)
OCL=34792
CL=34794

gmp.go:197:4:  type mpz_t          C type  mpz_t
gmp.go:205:2:  call mpz_init       C value func(mpz_ptr) void
gmp.go:206:2:  call mpz_set        C value func(mpz_ptr, mpz_srcptr) void
gmp.go:221:2:  call mpz_init       C value func(mpz_ptr) void
gmp.go:227:7:  call size_t         C type  size_t
gmp.go:228:2:  call mpz_export     C value func(*void, *size_t, int, size_t, int, size_t, mpz_srcptr) *void
gmp.go:235:13: call mpz_sizeinbase C value func(mpz_srcptr, int) size_t
gmp.go:241:2:  call mpz_set        C value func(mpz_ptr, mpz_srcptr) void
gmp.go:252:3:  call mpz_import     C value func(mpz_ptr, size_t, int, size_t, int, size_t, *const void) void
gmp.go:261:2:  call mpz_set_si     C value func(mpz_ptr, long int) void
gmp.go:273:5:  call mpz_set_str    C value func(mpz_ptr, *const char, int) int
gmp.go:282:9:  call mpz_get_str    C value func(*char, int, mpz_srcptr) *char
gmp.go:287:3:  call mpz_clear      C value func(mpz_ptr) void
gmp.go:302:2:  call mpz_add        C value func(mpz_ptr, mpz_srcptr, mpz_srcptr) void
gmp.go:311:2:  call mpz_sub        C value func(mpz_ptr, mpz_srcptr, mpz_srcptr) void
gmp.go:320:2:  call mpz_mul        C value func(mpz_ptr, mpz_srcptr, mpz_srcptr) void
gmp.go:329:2:  call mpz_tdiv_q     C value func(mpz_ptr, mpz_srcptr, mpz_srcptr) void
gmp.go:339:2:  call mpz_tdiv_r     C value func(mpz_ptr, mpz_srcptr, mpz_srcptr) void
gmp.go:348:2:  call mpz_mul_2exp   C value func(mpz_ptr, mpz_srcptr, long unsigned int) void
gmp.go:356:2:  call mpz_div_2exp   C value func(mpz_ptr, mpz_srcptr, long unsigned int) void
gmp.go:367:3:  call mpz_pow_ui     C value func(mpz_ptr, mpz_srcptr, long unsigned int) void
gmp.go:369:3:  call mpz_powm       C value func(mpz_ptr, mpz_srcptr, mpz_srcptr, mpz_srcptr) void
gmp.go:378:2:  call mpz_neg        C value func(mpz_ptr, mpz_srcptr) void
gmp.go:386:2:  call mpz_abs        C value func(mpz_ptr, mpz_srcptr) void
gmp.go:404:9:  call mpz_cmp        C value func(mpz_srcptr, mpz_srcptr) int
gmp.go:413:2:  call mpz_tdiv_qr    C value func(mpz_ptr, mpz_ptr, mpz_srcptr, mpz_srcptr) void
gmp.go:426:2:  call mpz_gcdext     C value func(mpz_ptr, mpz_ptr, mpz_ptr, mpz_srcptr, mpz_srcptr) void

R=r
DELTA=938  (628 added, 308 deleted, 2 changed)
OCL=34733
CL=34791

test both ELF and Mach-O in dwarf package.

R=r
DELTA=83  (44 added, 10 deleted, 29 changed)
OCL=34717
CL=34790

R=r
DELTA=784  (784 added, 0 deleted, 0 changed)
OCL=34715
CL=34788

from inlining newprocreadylocked.  Fix type bridge's handling
of basic types.  Include interpreter's Thread in bridged
native function calls.

; load . "6.out"
Started 6.out
; BpSet("main·merge")
; ContWait()
breakpoint at 0x400800
=>   400800 main·merge /home/austin/src-go1/usr/austin/ptrace/test/sort.go:19
; bt
=>   400800 main·merge /home/austin/src-go1/usr/austin/ptrace/test/sort.go:19
     400b6a main·mergeSort+0x1be /home/austin/src-go1/usr/austin/ptrace/test/sort.go:34
     448313 goexit /home/austin/src-go1/src/pkg/runtime/proc.c:133
; main.merge.a
{1}

; load . "pid:25753"
Attached to 25753
; bt
=>   479ddf syscall·Syscall+0x24 /home/austin/src-go1/src/pkg/syscall/asm_linux_amd64.s:24
     47c011 syscall·Read+0x5d /home/austin/src-go1/src/pkg/syscall/zsyscall_linux_amd64.go:368
     4119e5 os·*File·Read+0x5f /home/austin/src-go1/src/pkg/os/file.go:122
     427bf3 bufio·*Reader·fill+0x116 /home/austin/src-go1/src/pkg/bufio/bufio.go:105
     428361 bufio·*Reader·ReadSlice+0x195 /home/austin/src-go1/src/pkg/bufio/bufio.go:244
     40204a ogle·Main+0x94 /home/austin/src-go1/usr/austin/ogle/cmd.go:226
     40080f main·main+0xf /home/austin/src-go1/usr/austin/ogle/main.go:6
     41c4b8 mainstart+0xf /home/austin/src-go1/src/pkg/runtime/amd64/asm.s:55
     41531f goexit /home/austin/src-go1/src/pkg/runtime/proc.c:133

R=rsc
APPROVED=rsc
DELTA=433  (420 added, 2 deleted, 11 changed)
OCL=34410
CL=34782

R=rsc
APPROVED=rsc
DELTA=282  (281 added, 0 deleted, 1 changed)
OCL=34407
CL=34781

use strings.Buffer instead of bytes.Buffer in some places

R=rsc
DELTA=40  (0 added, 3 deleted, 37 changed)
OCL=34770
CL=34775

similar issues easier in the future.

R=rsc
APPROVED=rsc
DELTA=6  (1 added, 0 deleted, 5 changed)
OCL=34771
CL=34773