Debugger: fix stepping next with trycatch recursion

Debugger: fix stepping next with trycatch recursion

Added depth check to StepNextContinue. Do step out and queue actual StepNext if check failed.

R=yangguo@chromium.org,danno@chromium.org,vegorov@chromium.org
BUG=v8:1639
TEST=

Committed: http://code.google.com/p/v8/source/detail?r=9264

[Yang, 2011-09-12 10:42:37]

On 2011/09/12 10:09:16, indutny wrote:

Thanks for the patch. I'll take a look into it in a bit.

For outside contributors we require the CLA to be signed before we can accept
any patch. It won't take long if you do it electronically here:
http://code.google.com/legal/individual-cla-v1.0.html

Please tell me as soon as you signed it so I can commit the changes.

[indutny, 2011-09-12 10:43:25]

On 2011/09/12 10:42:37, Yang wrote:
> On 2011/09/12 10:09:16, indutny wrote:
> 
> Thanks for the patch. I'll take a look into it in a bit.
> 
> For outside contributors we require the CLA to be signed before we can accept
> any patch. It won't take long if you do it electronically here:
> http://code.google.com/legal/individual-cla-v1.0.html
> 
> Please tell me as soon as you signed it so I can commit the changes.

Already signed!

[indutny, 2011-09-12 11:13:33]

Sorry, lines 982-985 are useless and came from obsolete implementation of fix.

[Yang, 2011-09-12 12:44:06]

On 2011/09/12 11:13:33, indutny wrote:
> Sorry, lines 982-985 are useless and came from obsolete implementation of fix.

I haven't looked closely at the patch yet, but did a quick test. While the patch
fixes the bug described in 1639, it causes a few existing tests to crash, for
example

> ./d8 --stress-opt --always-opt --expose-debug-as debug test/mjsunit/mjsunit.js
test/mjsunit/debug-continue.js --test

[indutny, 2011-09-12 12:50:02]

Oh, I think I know what causes a problem, I'll push changes in a 10-20 minutes.
Should I recreate ticket once I'll get this problem fixed?

[Yang, 2011-09-12 12:53:05]

On 2011/09/12 12:50:02, indutny wrote:
> Oh, I think I know what causes a problem, I'll push changes in a 10-20
minutes.
> Should I recreate ticket once I'll get this problem fixed?

If you used upload.py to upload you can just upload a new patch set on top of
this one. If not, creating a new issue is probably the easiest way.

[indutny, 2011-09-12 13:08:44]

Oh, git-cl created second patch set, removed old one. 

`tools/test.py  mjsunit/debug*` reports 0 errors now.

Could you please review it?

[Yang, 2011-09-12 17:38:33]

On 2011/09/12 13:19:41, indutny wrote:

The mjsunit tests are not the only ones. If you run tools/test.py, there are
also cctests executed, some of which are still failing with this patch.

Btw. you don't have to delete the previous patch sets, they are supposed to be
incremental.

I haven't had a chance to get familiar with your code yet, but will try to get
to that by tomorrow.

[indutny, 2011-09-12 17:44:20]

Can you please provide me those tests' names?

[indutny, 2011-09-12 17:53:32]

Nevermind, I know names now

[indutny, 2011-09-12 18:26:37]

Fixed patch, 100% passing now

[indutny, 2011-09-13 14:15:43]

So what is a current status of this issue?
That blocks some of my work on node.js debugger, so I'll be glad to know if I
can help to merge it in master somehow.

Thanks

[Yang, 2011-09-13 14:23:16]

Looks pretty good so far, except for one little thing remarked below. You might
want to change that before I land your patch :)

http://codereview.chromium.org/7780033/diff/9002/src/debug.cc
File src/debug.cc (right):

http://codereview.chromium.org/7780033/diff/9002/src/debug.cc#newcode998
src/debug.cc:998: thread_local_.queued_step_action_ = step_action;
From what I can see, the only step action that is ever queued up is StepNext.
Can't we remove queued_step_action_ altogether then? We already use
queued_step_count_ as indicator for whether there is anything queued up, there
is no point in storing queued_step_action_ when we already know that it can only
be StepNext (or StepNone in the uninteresting case, but then queued_step_count_
== 0 is indication enough).

[indutny, 2011-09-13 14:32:03]

Ok, sounds good for me.
Fixed and uploaded changes, tests are still passing.

[Yang, 2011-09-13 14:38:26]

On 2011/09/13 14:32:03, indutny wrote:
> Ok, sounds good for me.
> Fixed and uploaded changes, tests are still passing.

LGTM. Landing in a minute.

[indutny, 2011-09-13 14:39:09]

Great! Thank you!