REGRESSION: 10 MB increase in CachedScripts on Membuster. Regression caused by http://trac.webkit.org/projects/webkit/changeset/78297. Fix by adding back removed call to destroy decoded data. <rdar://problem/9146716> REGRESSION: ~10 MB increase in memory under CachedScripts
Created attachment 86873 [details] patch
Comment on attachment 86873 [details] patch r=me, but, as discussed by email, please talk to Michael and make sure you're on the same page here -- he measured r78297 as a membuster improvement!
Stephanie's change in the patch focuses on the during test memory usage while my earlier change addressed retained CachedScript data after MemBuster has finished. From her email comment: Michael was testing the end of the test whereas I am looking at the high water mark. I also saw his change as a win for the end of the test. Michaels change was in response to the bug I filed that said that data from JSC parser was hanging around after all the windows had been closed. Michaels change fixed this by adding a timer to destroy CachedScript decoded data and as part of that to clear the SourceProviderCache. It was a win for the end of membuster. In adding the timer he removed the optimization to immediately throw away decoded data from CachedScripts regressing the high water mark. Since closing the window would trigger his timer it wouldn't have been obvious from the end of the test that the high water mark had increased. I agree with this change.
committed http://trac.webkit.org/projects/webkit/changeset/81991