Hold the CrossThreadPersistentMutex across heap poisoning.
Since cross-thread persistents may be included in the poisoning, but must not be observed by another thread to be poisoned, the mutex must be held to prevent them noticing before they are unpoisoned again. The alternative which would narrow the critical section might be to store away a copy of the cross-thread persistents and not poison them at all, but this is consistent with the previous approach and thus seems less risky. Bug: 857311 Change-Id: Ib0ddee582c6ceea98cce0c0b3844fb363aa3c5be Reviewed-on: https://chromium-review.googlesource.com/1118286Reviewed-by:Kentaro Hara <haraken@chromium.org> Commit-Queue: Jeremy Roman <jbroman@chromium.org> Cr-Commit-Position: refs/heads/master@{#571145}
Showing
Please register or sign in to comment