Fail new space promotions only when old space is truly exhausted.

Fail new space promotions only when old space is truly exhausted.

Committed: https://code.google.com/p/dart/source/detail?r=14935

[Ivan Posva, 2012-11-14 21:07:18]

https://codereview.chromium.org/11363226/diff/1/runtime/vm/scavenger.cc
File runtime/vm/scavenger.cc (right):

https://codereview.chromium.org/11363226/diff/1/runtime/vm/scavenger.cc#newco...
runtime/vm/scavenger.cc:188: bytes_promoted_ += size;
Comment above has been separated from the code it refers to.

https://codereview.chromium.org/11363226/diff/1/runtime/vm/scavenger.cc#newco...
runtime/vm/scavenger.cc:191: // Retry.
Please expand on the "Retry." comment by briefly explaining the state machine.

https://codereview.chromium.org/11363226/diff/1/runtime/vm/scavenger.cc#newco...
runtime/vm/scavenger.cc:200: }
What if new_addr is 0?

[cshapiro, 2012-11-14 22:47:36]

https://codereview.chromium.org/11363226/diff/1/runtime/vm/scavenger.cc
File runtime/vm/scavenger.cc (right):

https://codereview.chromium.org/11363226/diff/1/runtime/vm/scavenger.cc#newco...
runtime/vm/scavenger.cc:188: bytes_promoted_ += size;
I do not follow.  Do you want me to transpose lines 188 and 189?  That is
trivial...

https://codereview.chromium.org/11363226/diff/1/runtime/vm/scavenger.cc#newco...
runtime/vm/scavenger.cc:200: }
Same thing that always happens.  We trip the assert below in debug mode and
SIGSEGV in release mode because of the invalid memmove destination argument.

[cshapiro, 2012-11-15 00:51:59]

PTAL

In the mean time I will benchmark the smaller new space.

[Ivan Posva, 2012-11-15 01:02:24]

LGTM with comment.

-Ivan

https://codereview.chromium.org/11363226/diff/1003/runtime/vm/heap.h
File runtime/vm/heap.h (right):

https://codereview.chromium.org/11363226/diff/1003/runtime/vm/heap.h#newcode79
runtime/vm/heap.h:79: PageSpace::GrowthPolicy growth_policy =
PageSpace::kForceGrowth) {
Carl says: "That's wrong!"

[cshapiro, 2012-11-15 01:05:27]

fixed!