Avoid fast short-cut in Map::GeneralizeRepresentation() for literals with non-simple transitions.

Avoid fast short-cut in Map::GeneralizeRepresentation() for literals with non-simple transitions.
It started showing after r25253.

BUG=v8:3687
LOG=N
R=verwaest@chromium.org

Committed: https://chromium.googlesource.com/v8/v8/+/bc8c41c08d64f5d051ae82db54d59ae2a167f430

[Igor Sheludko, 2014-11-12 18:34:37]

ishell@chromium.org changed reviewers:
+ verwaest@chromium.org

[Igor Sheludko, 2014-11-12 18:34:37]

PTAL

[Toon Verwaest, 2014-11-12 22:52:48]

Thanks for finding the problem!

https://codereview.chromium.org/715313003/diff/1/src/objects.cc
File src/objects.cc (right):

https://codereview.chromium.org/715313003/diff/1/src/objects.cc#newcode2480
src/objects.cc:2480: old_map->transitions()->IsSimpleTransition())) {
What about instead updating the fieldtype in the entire tree? There's
infrastructure to do that. Knowing we have a simple transition doesn't guarantee
(I believe) that we are sharing the descriptor array...

[Igor Sheludko, 2014-11-13 09:58:21]

Patchset #2 (id:20001) has been deleted

[Igor Sheludko, 2014-11-13 10:00:41]

PTAL

https://codereview.chromium.org/715313003/diff/1/src/objects.cc
File src/objects.cc (right):

https://codereview.chromium.org/715313003/diff/1/src/objects.cc#newcode2480
src/objects.cc:2480: old_map->transitions()->IsSimpleTransition())) {
On 2014/11/12 22:52:48, Toon Verwaest wrote:
> What about instead updating the fieldtype in the entire tree? There's
> infrastructure to do that. Knowing we have a simple transition doesn't
guarantee
> (I believe) that we are sharing the descriptor array...

Done, thanks!

[Toon Verwaest, 2014-11-13 10:39:28]

lgtm

[Igor Sheludko, 2014-11-13 10:56:36]

Committed patchset #2 (id:40001) manually as
bc8c41c08d64f5d051ae82db54d59ae2a167f430 (presubmit successful).