Vincent Kang Fu

The default value for invalidate is 1. So more investigation needs to happen to understand what is going on. Consider running a small job with `--debug=io,file` to gain insight into...

If you are using python, here is one work-around for this issue: https://github.com/axboe/fio/blob/fio-3.17/t/run-fio-tests.py#L281

I tried the same installer in WS 2019 and saw the same behavior you did. Consider trying the installers from appveyor. This page has a link to the cygwin fio...

@bcran This appears in the event viewer after running fio.exe from the installer on bsdio.com: ``` Faulting application name: fio.exe, version: 0.0.0.0, time stamp: 0x624c97a8 Faulting module name: fio.exe, version:...

I was able to reproduce this problem and bisect for me identified f31feaa21642929b6d9d5396b73669372fda9a0a as the first bad commit. ``` commit f31feaa21642929b6d9d5396b73669372fda9a0a Author: Jens Axboe Date: Tue Apr 17 21:50:55 2018...

@sitsofe I don't fully understand your suggestion, but I think any sort of rounding or truncation might also lead to undesirable outcomes. If the Xth percentile legitimately maps to the...

I don't know of a good solution to this problem, but it is diagnosed and documented here in case someone comes along and is able to fix it.

Thanks for getting to the bottom of this. I made the change (298d751e5f7dcde630d90a38f9ad5ab6670483a1) to use a long double because users were encountering a floating point precision issue with 64-bit doubles...

Here is another latency percentile mismatch test failure: https://ci.appveyor.com/project/axboe/fio/builds/34314077/job/v5wh22tgqqgdoly5 As long as fio and python calculate percentiles with the same floating point representation the tests will pass, although the fundamental...

Possible recent occurrences of this problem: https://ci.appveyor.com/project/axboe/fio/builds/44283157/job/19b037ummt52bmov https://ci.appveyor.com/project/axboe/fio/builds/44417966/job/kffdycupod0cboca https://ci.appveyor.com/project/axboe/fio/builds/44416420/job/l8x12p3m138ee67k It's strange to have so many sightings in such a short period (two today, one two weeks ago), but we haven't...