Fix compilation error introduced with r15287.

Fix compilation error introduced with r15287.

REGEXP was added to Code::Kind after TO_BOOLEAN_IC, but NUMBER_OF_KINDS,
which is used as array size for table[] in ReportCodeKindStatistics, was
still TO_BOOLEAN_IC + 1 (indirectly via LAST_IC_KIND).

BUG=
R=svenpanne@chromium.org

Committed: 15315

[Sven Panne, 2013-06-25 06:36:04]

Not LGTM. Although this is a workaround, we already managed to break this
fragile thing, so let's do it right now:

* Nuke FIRST_IC_KIND/LAST_IC_KIND and use a switch with macro-generated cases in
Code::is_inline_cache_stub().

* Calculate LAST_CODE_KIND via a macro (0 + 1 + 1 + ...).

* Use Code::Kind2String in ReportCodeKindStatistics, removing the macro nonsense
there.

After that, order doesn't matter and one can simply add/remove/reorder the
kinds.

[danno, 2013-06-25 06:45:56]

Why not just have two lists? NON_IC_CODE_KIND_LIST and IC_CODE_KIND_LIST? Just
append the lists together for the complete list, and in all other cases (i.e.
calculating first and last) you don't need special macros, just an enum {} with
the macro to calculate the count. This obviates the need for switches and seems
generally cleaner to me.

[Benedikt Meurer, 2013-06-25 07:55:44]

On 2013/06/25 06:36:04, Sven Panne wrote:
> * Nuke FIRST_IC_KIND/LAST_IC_KIND and use a switch with macro-generated cases
in
> Code::is_inline_cache_stub().

Done.

> * Calculate LAST_CODE_KIND via a macro (0 + 1 + 1 + ...).

Done.

> * Use Code::Kind2String in ReportCodeKindStatistics, removing the macro
nonsense
> there.

Done.

[Benedikt Meurer, 2013-06-25 07:56:01]

On 2013/06/25 06:45:56, danno wrote:
> Why not just have two lists? NON_IC_CODE_KIND_LIST and IC_CODE_KIND_LIST? Just
> append the lists together for the complete list, and in all other cases (i.e.
> calculating first and last) you don't need special macros, just an enum {}
with
> the macro to calculate the count. This obviates the need for switches and
seems
> generally cleaner to me.

Done.

[Sven Panne, 2013-06-25 08:08:10]

Almost there...

https://codereview.chromium.org/17636003/diff/5001/src/objects.h
File src/objects.h (right):

https://codereview.chromium.org/17636003/diff/5001/src/objects.h#newcode4474
src/objects.h:4474: LAST_CODE_KIND = -1 CODE_KIND_LIST(COUNT_FLAG)
I don't think we need both LAST_CODE_KIND and NUMBER_OF_KINDS, so let's remove
this pseudo-kind and only use NUMBER_OF_KINDS (calculated via the macro magic
above).

https://codereview.chromium.org/17636003/diff/5001/src/spaces.cc
File src/spaces.cc (right):

https://codereview.chromium.org/17636003/diff/5001/src/spaces.cc#newcode1802
src/spaces.cc:1802: Isolate* isolate = Isolate::Current();
Not from your change, but let's change this while we are here: Nuke the Isolate
and pass code_kind_statistics from the caller.

https://codereview.chromium.org/17636003/diff/5001/src/spaces.cc#newcode1803
src/spaces.cc:1803: const char* table[Code::NUMBER_OF_KINDS] = { NULL };
Nuke the table and just use Code::Kind2String below.

[Benedikt Meurer, 2013-06-25 08:48:09]

https://codereview.chromium.org/17636003/diff/5001/src/objects.h
File src/objects.h (right):

https://codereview.chromium.org/17636003/diff/5001/src/objects.h#newcode4474
src/objects.h:4474: LAST_CODE_KIND = -1 CODE_KIND_LIST(COUNT_FLAG)
On 2013/06/25 08:08:10, Sven Panne wrote:
> I don't think we need both LAST_CODE_KIND and NUMBER_OF_KINDS, so let's remove
> this pseudo-kind and only use NUMBER_OF_KINDS (calculated via the macro magic
> above).

Done.

https://codereview.chromium.org/17636003/diff/5001/src/spaces.cc
File src/spaces.cc (right):

https://codereview.chromium.org/17636003/diff/5001/src/spaces.cc#newcode1802
src/spaces.cc:1802: Isolate* isolate = Isolate::Current();
On 2013/06/25 08:08:10, Sven Panne wrote:
> Not from your change, but let's change this while we are here: Nuke the
Isolate
> and pass code_kind_statistics from the caller.

Done.

https://codereview.chromium.org/17636003/diff/5001/src/spaces.cc#newcode1803
src/spaces.cc:1803: const char* table[Code::NUMBER_OF_KINDS] = { NULL };
On 2013/06/25 08:08:10, Sven Panne wrote:
> Nuke the table and just use Code::Kind2String below.

Done.

[Sven Panne, 2013-06-25 09:02:00]

lgtm

[Benedikt Meurer, 2013-06-25 09:09:33]

Committed patchset #4 manually as r15315 (presubmit successful).