> > I expect that on Windows, qsort() is not stable and the output will change. Could > you > > please check if you have time? > > Thanks, that patch looks very reasonable. Let me know what you find out regarding the > other issue.
Right, good find. The simplest fix as you mention is to ensure that we never return "equal" from the qsort callback.
Fortunately, no matter the ordering, all CHDs produced the existing "bad" way should be readable by all other implementations. But we should make sure we always generate the same output regardless. Thanks for digging into it.
|