On 10/20/23 01:05, Rob Landley wrote: > P.S. Count's X/s time output is actually X/period and the period is 250ms, but > it's also zeroing new periods and advancing into the zeroed period immediately > in the calculation as a full period when we've only been there for less than a > full 250ms, and what it SHOULD do is look at the remainder of millitime()%250 > and use that to add a fraction of the oldest period (not otherwise included, > and > not included as a period in the divisor) so "however much data the new one > accumulated in less than a full period of time" plus "the reciprocal portion > of > the old period we left" should equal a coherent count.
After commit 2f93b89b62d1 it's now: $ toybox timeout .1 ./yes | ./count -l >/dev/null 249135104 bytes, 238Mb, 2.3Gb/s, 0m00s $ toybox timeout .2 ./yes | ./count -l >/dev/null 512294912 bytes, 489Mb, 2.3Gb/s, 0m00s $ toybox timeout .6 ./yes | ./count -l >/dev/null 1665302528 bytes, 1.5Gb, 2.5Gb/s, 0m00s $ toybox timeout 1 ./yes | ./count -l >/dev/null 2610102272 bytes, 2.4Gb, 2.4Gb/s, 0m01s I had to explicitly use toybox timeout because debian's timeout doesn't seem to actually call the child process for .2 seconds or less. No, I don't have a clear idea how to add this to the test suite yet, but I'm calling it good enough for now. Rob _______________________________________________ Toybox mailing list Toybox@lists.landley.net http://lists.landley.net/listinfo.cgi/toybox-landley.net