Consolidate debug stubs

Consolidate debug stubs

The new debug stub BreakpointRuntime gets the original stub address
from the debugger. With this scheme, we don't need a new debug stub
for every runtime stub that the debugger patches.

R=asiva@google.com

Committed: https://code.google.com/p/dart/source/detail?r=23094

[siva, 2013-05-23 01:11:26]

LGTM with one question about target_address_ being a Smi...

https://codereview.chromium.org/15743019/diff/7002/runtime/vm/code_generator.cc
File runtime/vm/code_generator.cc (right):

https://codereview.chromium.org/15743019/diff/7002/runtime/vm/code_generator....
runtime/vm/code_generator.cc:811: // calling a closure.
The comment is a bit dated....

https://codereview.chromium.org/15743019/diff/7002/runtime/vm/code_generator....
runtime/vm/code_generator.cc:822: ASSERT((orig_stub & kSmiTagMask) == kSmiTag);
Is this assert guaranteed to hold always?

https://codereview.chromium.org/15743019/diff/7002/runtime/vm/debugger.cc
File runtime/vm/debugger.cc (right):

https://codereview.chromium.org/15743019/diff/7002/runtime/vm/debugger.cc#new...
runtime/vm/debugger.cc:1726: return bpt->saved_bytes_.target_address_;
Is bpt->saved_bytes_.target_address_ a Smi always?

I didn't see any code that does the shift so I am assuming this code is hoping
that the target_address_ is always even aligned.

https://codereview.chromium.org/15743019/diff/7002/runtime/vm/stub_code_ia32.cc
File runtime/vm/stub_code_ia32.cc (right):

https://codereview.chromium.org/15743019/diff/7002/runtime/vm/stub_code_ia32....
runtime/vm/stub_code_ia32.cc:1719: __ pushl(Immediate(0));
const Immediate& raw_null =
Immediate(reinterpret_cast<intptr_t>(Object::null());
__ pushl(raw_null); // Room for result.

similar to how it is done in other stubs.

https://codereview.chromium.org/15743019/diff/7002/runtime/vm/stub_code_x64.cc
File runtime/vm/stub_code_x64.cc (right):

https://codereview.chromium.org/15743019/diff/7002/runtime/vm/stub_code_x64.c...
runtime/vm/stub_code_x64.cc:1697: __ pushq(Immediate(0));
Ditto comment about pushq(raw_null);

[hausner, 2013-05-23 18:32:12]

https://codereview.chromium.org/15743019/diff/7002/runtime/vm/code_generator.cc
File runtime/vm/code_generator.cc (right):

https://codereview.chromium.org/15743019/diff/7002/runtime/vm/code_generator....
runtime/vm/code_generator.cc:811: // calling a closure.
On 2013/05/23 01:11:26, siva wrote:
> The comment is a bit dated....

Done.

https://codereview.chromium.org/15743019/diff/7002/runtime/vm/code_generator....
runtime/vm/code_generator.cc:822: ASSERT((orig_stub & kSmiTagMask) == kSmiTag);
On 2013/05/23 01:11:26, siva wrote:
> Is this assert guaranteed to hold always?

The assert is here to make sure we catch a case where it's not the case :). The
beginning of a stub or function should always be at an even address.

https://codereview.chromium.org/15743019/diff/7002/runtime/vm/debugger.cc
File runtime/vm/debugger.cc (right):

https://codereview.chromium.org/15743019/diff/7002/runtime/vm/debugger.cc#new...
runtime/vm/debugger.cc:1726: return bpt->saved_bytes_.target_address_;
On 2013/05/23 01:11:26, siva wrote:
> Is bpt->saved_bytes_.target_address_ a Smi always?
> 
> I didn't see any code that does the shift so I am assuming this code is hoping
> that the target_address_ is always even aligned.

Yes, we depend on the fact that target addresses of stubs are aligned. See
assertion in code_generator.cc.

https://codereview.chromium.org/15743019/diff/7002/runtime/vm/stub_code_ia32.cc
File runtime/vm/stub_code_ia32.cc (right):

https://codereview.chromium.org/15743019/diff/7002/runtime/vm/stub_code_ia32....
runtime/vm/stub_code_ia32.cc:1719: __ pushl(Immediate(0));
On 2013/05/23 01:11:26, siva wrote:
> const Immediate& raw_null =
> Immediate(reinterpret_cast<intptr_t>(Object::null());
> __ pushl(raw_null); // Room for result.
> 
> similar to how it is done in other stubs.

Done.

https://codereview.chromium.org/15743019/diff/7002/runtime/vm/stub_code_x64.cc
File runtime/vm/stub_code_x64.cc (right):

https://codereview.chromium.org/15743019/diff/7002/runtime/vm/stub_code_x64.c...
runtime/vm/stub_code_x64.cc:1697: __ pushq(Immediate(0));
On 2013/05/23 01:11:26, siva wrote:
> Ditto comment about pushq(raw_null);

Done.

[hausner, 2013-05-23 18:42:10]

Committed patchset #3 manually as r23094 (presubmit successful).