Fix compose-discard crasher from 11524. We can't do a call (to a generic

Fix compose-discard crasher from 11524.  We can't do a call (to a generic
stub) unless there is a pointer map.  This does not fix the 3d-raytrace
regression, that will be in another change.
Committed: https://code.google.com/p/v8/source/detail?r=11539

[Michael Starzinger, 2012-05-10 13:06:00]

LGTM (if comments addressed).

https://chromiumcodereview.appspot.com/10382102/diff/1/src/ia32/lithium-codeg...
File src/ia32/lithium-codegen-ia32.cc (right):

https://chromiumcodereview.appspot.com/10382102/diff/1/src/ia32/lithium-codeg...
src/ia32/lithium-codegen-ia32.cc:2294: if (map_count != 0) {
It seems as if this condition always holds, otherwise you would have ended up in
the IC case. Can we replace this with an ASSERT(map_count != 0)?

https://chromiumcodereview.appspot.com/10382102/diff/1/src/ia32/lithium-codeg...
src/ia32/lithium-codegen-ia32.cc:2307: if (map_count != 0) {
Then it makes sense to turn this into an "else if" which would simplify the
structure.

[Erik Corry, 2012-05-10 20:28:33]

https://chromiumcodereview.appspot.com/10382102/diff/1/src/ia32/lithium-codeg...
File src/ia32/lithium-codegen-ia32.cc (right):

https://chromiumcodereview.appspot.com/10382102/diff/1/src/ia32/lithium-codeg...
src/ia32/lithium-codegen-ia32.cc:2294: if (map_count != 0) {
On 2012/05/10 13:06:01, Michael Starzinger wrote:
> It seems as if this condition always holds, otherwise you would have ended up
in
> the IC case. Can we replace this with an ASSERT(map_count != 0)?

No, there are two conditions to get into the IC case now.

https://chromiumcodereview.appspot.com/10382102/diff/1/src/ia32/lithium-codeg...
src/ia32/lithium-codegen-ia32.cc:2307: if (map_count != 0) {
On 2012/05/10 13:06:01, Michael Starzinger wrote:
> Then it makes sense to turn this into an "else if" which would simplify the
> structure.

Same objection.

[Michael Starzinger, 2012-05-11 12:00:50]

As discussed offline, I still think my comments apply. But you already took care
of that in the follow-up cleanup CL. So LGTM if combined with the following.

https://chromiumcodereview.appspot.com/10386089/