When querying browsing history, an "end_time" is passed in as a query option. This end time is supposed to be exclusive, according to the documentation in history_types.h: https://cs.chromium.org/chromium/src/components/history/core/browser/history_types.h?q=queryoptions+end_time&sq=package:chromium&dr=CSs&l=235

It appears that for the regular local HistoryService the end time is exclusive, but for the WebHistoryService the end time is inclusive.

Repro steps:
1) Make a query to BrowsingHistoryService with base::Time() for the end_time (match all results) and a max of X queries (e.g. X = 5)
2) Store the timestamp for the last history entry in C++
3) Issue another query to BrowsingHistoryService with the stored timestamp for the end_time and a max of X queries
4) repeat steps 2-3

Observed behavior: duplicate results from the WebHistoryService appear at the query boundary e.g. if google.com is the last result returned in the query for #1, it will be the first result returned for the query in #3.

Expected behavior: the end time is supposed to be exclusive

I suspect this is a non-regression issue that has been hidden by the fact that our history web ui converts the C++ timestamp to/from JavaScript time. JavaScript uses less bits to represent the number so some precision is lost e.g. rather than 13123713486001564, the conversion to/from JavaScript causes the end_time to be 13123713486001000. This effectively sets the end_time used for the next query to an earlier date and hides WebHistoryService's non-exclusive use of end_time.