Use correct lower bounds for geometric distribution bucketers in the new protos.
The old protos said that the underflow bucket was [-Inf, 0) and the first
finite bucket was [0, 1). The new protos say the underflow bucket is
[-Inf, 1), but the code was never updated to do that, so we've been sending
incorrect distributions to the new API.
BUG=
Change-Id: I012b83785defda73c462893df08c862976e118e0
Reviewed-on: https://chromium-review.googlesource.com/459278
Reviewed-by: Vadim Shtayura <vadimsh@chromium.org>
Commit-Queue: Dave Sansome <dsansome@chromium.org>
Cr-Mirrored-From: https://chromium.googlesource.com/infra/infra
Cr-Mirrored-Commit: 6a4e0d04729cf9238730740c13658cef793454ec
4 files changed