First of all: thanks again, also for running your numbers again!
Second, I won't be doing this for every point update: I don't want to waste your time.
This PR does not (should not) affect benchmark numbers in any significant way.
I'm updating because again your code helped me detect a defect in my driver (thanks!), this time macOS specific, affecting only the large workloads.
This workload could (momentarily) need orders of magnitude (100s of GB) more disk space than actually required.
This was due to my implementation of fallocate for macOS, which I ported from mozilla. The buggy code is in Gists, StackOverflow answers, KDE source code, has been independently ported to other languages multiple times over.
Hard to believe it's wrong, but a new unit test demonstrates it is.
First of all: thanks again, also for running your numbers again! Second, I won't be doing this for every point update: I don't want to waste your time.
This PR does not (should not) affect benchmark numbers in any significant way.
I'm updating because again your code helped me detect a defect in my driver (thanks!), this time macOS specific, affecting only the large workloads. This workload could (momentarily) need orders of magnitude (100s of GB) more disk space than actually required.
This was due to my implementation of
fallocate
for macOS, which I ported from mozilla. The buggy code is in Gists, StackOverflow answers, KDE source code, has been independently ported to other languages multiple times over.Hard to believe it's wrong, but a new unit test demonstrates it is.