Fix float to int conversion in base/numerics
This addresses the bug where range checks are incorrect when
converting from a floating point value to an integral value of
higher precision but lower width. It also reverts a previous
attempted fix that eliminated false negatives but introduced
false positives.
BUG=522989
Review URL: https://codereview.chromium.org/1338553003
Cr-Commit-Position: refs/heads/master@{#348706}
CrOS-Libchrome-Original-Commit: fafe071bac241504791c2904af9fbfea476274c8
2 files changed