code review 3477041: [68]l and runtime: GDB support for interfaces and gorou...

[68]l and runtime: GDB support for interfaces and goroutines.

[lvd, 2010-12-08 12:36:44]

Hello rsc (cc: golang-dev@googlegroups.com),

I'd like you to review this change.

[lvd, 2010-12-08 12:42:36]

On 2010/12/08 12:36:44, lvd wrote:
> Hello rsc (cc: mailto:golang-dev@googlegroups.com),
> 
> I'd like you to review this change.

sample session:
lvd@hpgntai-ubiq89:~/play$ GOROOT=~/g PATH=$PATH:$GOROOT/bin GOARCH=amd64 6l
convert.6 
lvd@hpgntai-ubiq89:~/play$ gdb 6.out

....

Loading Go Runtime support.
(gdb) b 143
Breakpoint 1 at 0x401619: file /home/lvd/play/../g/test/interface/convert.go,
line 143.
(gdb) r
Starting program: /home/lvd/play/6.out 

Breakpoint 1, main.main () at /home/lvd/play/../g/test/interface/convert.go:144
144             e = u
(gdb) p 'main.e'
$1 = {str = "hello", len = 5}
(gdb) whatis 'main.e'
type = main.Empty
(gdb) iface 'main.e'
main.Empty: struct main.T *
(gdb) p $dtype('main.e')
$2 = (struct main.T *) 0x7ffff7ed81e0
(gdb) p *$dtype('main.e')
$3 = {str = 0x40d3b4 "hello", len = 5}



Goroutines:
Loading Go Runtime support.
(gdb) b 39
Breakpoint 1 at 0x400d85: file /home/lvd/g/test/chan/goroutines.go, line 39.
(gdb) r 12
Starting program: /home/lvd/play/6.out 12

Breakpoint 1, main.main () at /home/lvd/g/test/chan/goroutines.go:39
39              go func(c chan int) { c <- 1 }(right)
(gdb) info goroutines 
  13 runnable runtime.goexit
  12 runnable runtime.goexit
  11 runnable runtime.goexit
  10 runnable runtime.goexit
  9 runnable runtime.goexit
  8 runnable runtime.goexit
  7 runnable runtime.goexit
  6 runnable runtime.goexit
  5 runnable runtime.goexit
  4 runnable runtime.goexit
  3 runnable runtime.goexit
  2 runnable runtime.goexit
* 1  running scheduler
(gdb) goroutine 12 bt
#0  runtime.goexit () at /home/lvd/g/src/pkg/runtime/proc.c:146
#1  0x00007ffff7eea840 in ?? ()
#2  0x00007ffff7eea7e0 in ?? ()
#3  0x0000000000000000 in ?? ()
(gdb) goroutine 12 info local
No locals.

(that bt isnt accurate near the top of stacs is a separate bug)

[adg, 2010-12-08 20:24:41]

Hah, very cool!

On 8 December 2010 23:42,  <lvd@google.com> wrote:
> On 2010/12/08 12:36:44, lvd wrote:
>>
>> Hello rsc (cc: mailto:golang-dev@googlegroups.com),
>
>> I'd like you to review this change.
>
> sample session:
> lvd@hpgntai-ubiq89:~/play$ GOROOT=~/g PATH=$PATH:$GOROOT/bin
> GOARCH=amd64 6l convert.6
> lvd@hpgntai-ubiq89:~/play$ gdb 6.out
>
> ....
>
> Loading Go Runtime support.
> (gdb) b 143
> Breakpoint 1 at 0x401619: file
> /home/lvd/play/../g/test/interface/convert.go, line 143.
> (gdb) r
> Starting program: /home/lvd/play/6.out
>
> Breakpoint 1, main.main () at
> /home/lvd/play/../g/test/interface/convert.go:144
> 144             e = u
> (gdb) p 'main.e'
> $1 = {str = "hello", len = 5}
> (gdb) whatis 'main.e'
> type = main.Empty
> (gdb) iface 'main.e'
> main.Empty: struct main.T *
> (gdb) p $dtype('main.e')
> $2 = (struct main.T *) 0x7ffff7ed81e0
> (gdb) p *$dtype('main.e')
> $3 = {str = 0x40d3b4 "hello", len = 5}
>
>
>
> Goroutines:
> Loading Go Runtime support.
> (gdb) b 39
> Breakpoint 1 at 0x400d85: file /home/lvd/g/test/chan/goroutines.go, line
> 39.
> (gdb) r 12
> Starting program: /home/lvd/play/6.out 12
>
> Breakpoint 1, main.main () at /home/lvd/g/test/chan/goroutines.go:39
> 39              go func(c chan int) { c <- 1 }(right)
> (gdb) info goroutines
>  13 runnable runtime.goexit
>  12 runnable runtime.goexit
>  11 runnable runtime.goexit
>  10 runnable runtime.goexit
>  9 runnable runtime.goexit
>  8 runnable runtime.goexit
>  7 runnable runtime.goexit
>  6 runnable runtime.goexit
>  5 runnable runtime.goexit
>  4 runnable runtime.goexit
>  3 runnable runtime.goexit
>  2 runnable runtime.goexit
> * 1  running scheduler
> (gdb) goroutine 12 bt
> #0  runtime.goexit () at /home/lvd/g/src/pkg/runtime/proc.c:146
> #1  0x00007ffff7eea840 in ?? ()
> #2  0x00007ffff7eea7e0 in ?? ()
> #3  0x0000000000000000 in ?? ()
> (gdb) goroutine 12 info local
> No locals.
>
> (that bt isnt accurate near the top of stacs is a separate bug)
>
>
>
>
>
>
>
> http://codereview.appspot.com/3477041/
>

[rsc1, 2010-12-14 16:42:09]

LGTM

http://codereview.appspot.com/3477041/diff/5001/src/pkg/runtime/runtime-gdb.py
File src/pkg/runtime/runtime-gdb.py (right):

http://codereview.appspot.com/3477041/diff/5001/src/pkg/runtime/runtime-gdb.p...
src/pkg/runtime/runtime-gdb.py:170: try: return gdb.lookup_type(name)
please put these on lines by themselves.  grep for ": "

http://codereview.appspot.com/3477041/diff/5001/src/pkg/runtime/runtime-gdb.p...
src/pkg/runtime/runtime-gdb.py:191: type_size =  int(dynamic_go_type['size'])
weird spacing.  grep for "=  "

http://codereview.appspot.com/3477041/diff/5001/src/pkg/runtime/runtime-gdb.p...
src/pkg/runtime/runtime-gdb.py:323: Switches PC and SP to the ones in the
goroutine's G structure,
s/Switches/Switch/

http://codereview.appspot.com/3477041/diff/5001/src/pkg/runtime/runtime-gdb.p...
src/pkg/runtime/runtime-gdb.py:346: gdb.execute(cmd)
does this need to be wrapped in a try so that you can
restore pc/sp even if execute fails?
or does execute not throw?
(sure would help to have, oh i don't know, a type system.)

[lvd, 2010-12-14 16:52:18]

all done.  will test some more before submitting.

http://codereview.appspot.com/3477041/diff/5001/src/pkg/runtime/runtime-gdb.py
File src/pkg/runtime/runtime-gdb.py (right):

http://codereview.appspot.com/3477041/diff/5001/src/pkg/runtime/runtime-gdb.p...
src/pkg/runtime/runtime-gdb.py:170: try: return gdb.lookup_type(name)
On 2010/12/14 16:42:09, rsc1 wrote:
> please put these on lines by themselves.  grep for ": "

Done.

http://codereview.appspot.com/3477041/diff/5001/src/pkg/runtime/runtime-gdb.p...
src/pkg/runtime/runtime-gdb.py:191: type_size =  int(dynamic_go_type['size'])
On 2010/12/14 16:42:09, rsc1 wrote:
> weird spacing.  grep for "=  "

Done.

http://codereview.appspot.com/3477041/diff/5001/src/pkg/runtime/runtime-gdb.p...
src/pkg/runtime/runtime-gdb.py:346: gdb.execute(cmd)
On 2010/12/14 16:42:09, rsc1 wrote:
> does this need to be wrapped in a try so that you can
> restore pc/sp even if execute fails?
> or does execute not throw?
> (sure would help to have, oh i don't know, a type system.)

Done.

[lvd, 2010-12-15 11:00:47]

*** Submitted as http://code.google.com/p/go/source/detail?r=2c86bfce35c5 ***

[68]l and runtime: GDB support for interfaces and goroutines.

R=rsc
CC=golang-dev
http://codereview.appspot.com/3477041