Fix issue 553: function frame is skipped in profile when compare stub is called.

Fix issue 553: function frame is skipped in profile when compare stub is called.

The problem appeared due to a fact that stubs doesn't create a stack
frame, reusing the stack frame of the caller function. When building
stack traces, the current function is retrieved from PC, and its
callees are retrieved by traversing the stack backwards. Thus, for
stubs, the stub itself was discovered via PC, and then stub's caller's
caller was retrieved from stack.

To fix this problem, a pointer to JSFunction object is now captured
from the topmost stack frame, and is saved into stack trace log
record. Then a simple heuristics is applied whether a referred
function should be added to decoded stack, or not, to avoid reporting
the same function twice (from PC and from the pointer.)

BUG=553
TEST=added to mjsunit/tools/tickprocessor

Committed: http://code.google.com/p/v8/source/detail?r=3673

[Søren Thygesen Gjesse, 2010-01-21 08:28:52]

LGTM!

http://codereview.chromium.org/546089/diff/1/3
File src/handles.cc (right):

http://codereview.chromium.org/546089/diff/1/3#newcode685
src/handles.cc:685: LOG(FunctionInfoEvent(*function));
How about FunctionCreateEvent? The actual JS function object is created here.

http://codereview.chromium.org/546089/diff/1/5
File src/log.h (right):

http://codereview.chromium.org/546089/diff/1/5#newcode118
src/log.h:118: V(CODE_DELETE_EVENT,              "code-delete",            "cd")
      \
Do we need a function delete event?

http://codereview.chromium.org/546089/diff/1/5#newcode119
src/log.h:119: V(FUNCTION_INFO_EVENT,            "function-info",          "fi")
      \
INFO -> CREATION, info -> creation?

http://codereview.chromium.org/546089/diff/1/6
File src/mark-compact.cc (right):

http://codereview.chromium.org/546089/diff/1/6#newcode1954
src/mark-compact.cc:1954: LOG(CodeMoveEvent(old_addr, new_addr));
Should we have a FunctionMoveEvent? Or just a generic ObjectMoveEvent (perhaps
with a type)? Using the CodeMoveEvent for a JSFunction object seems confusing.

http://codereview.chromium.org/546089/diff/1/20
File tools/tickprocessor.js (right):

http://codereview.chromium.org/546089/diff/1/20#newcode286
tools/tickprocessor.js:286: this.profile_.addCodeAlias(functionAddr, codeAddr);
Are these aliases ever removed?

[mnaganov (inactive), 2010-01-21 16:02:39]

Thanks for thoughtful comments. I also discovered that I need to add some stuff
for DevTools to work after this patch.

http://codereview.chromium.org/546089/diff/1/3
File src/handles.cc (right):

http://codereview.chromium.org/546089/diff/1/3#newcode685
src/handles.cc:685: LOG(FunctionInfoEvent(*function));
On 2010/01/21 08:28:52, Søren Gjesse wrote:
> How about FunctionCreateEvent? The actual JS function object is created here.

Done.

http://codereview.chromium.org/546089/diff/1/5
File src/log.h (right):

http://codereview.chromium.org/546089/diff/1/5#newcode118
src/log.h:118: V(CODE_DELETE_EVENT,              "code-delete",            "cd")
      \
On 2010/01/21 08:28:52, Søren Gjesse wrote:
> Do we need a function delete event?

Added.

http://codereview.chromium.org/546089/diff/1/5#newcode119
src/log.h:119: V(FUNCTION_INFO_EVENT,            "function-info",          "fi")
      \
On 2010/01/21 08:28:52, Søren Gjesse wrote:
> INFO -> CREATION, info -> creation?

Done.

http://codereview.chromium.org/546089/diff/1/6
File src/mark-compact.cc (right):

http://codereview.chromium.org/546089/diff/1/6#newcode1954
src/mark-compact.cc:1954: LOG(CodeMoveEvent(old_addr, new_addr));
On 2010/01/21 08:28:52, Søren Gjesse wrote:
> Should we have a FunctionMoveEvent? Or just a generic ObjectMoveEvent (perhaps
> with a type)? Using the CodeMoveEvent for a JSFunction object seems confusing.

Added dedicated function-related events. Creating a generic events for "move"
and "delete" operations is OK, but for "create" operations is more elaborated
because it is assumed that all event instances have the same signature, while
for code objects and function objects they will be different:

ObjectCreateEvent,Code,Tag,address,size,...
ObjectCreateEvent,Function,address,code_address

So I decided not to create generic events in favor of dedicated ones.

http://codereview.chromium.org/546089/diff/1/20
File tools/tickprocessor.js (right):

http://codereview.chromium.org/546089/diff/1/20#newcode286
tools/tickprocessor.js:286: this.profile_.addCodeAlias(functionAddr, codeAddr);
On 2010/01/21 08:28:52, Søren Gjesse wrote:
> Are these aliases ever removed?

A good catch, thank you! Yes, I completely forgot reporting of function object
deletes.

Fixed.