Output rounding bug

Schmart

It seems I've run into strange behavior where floats are rounded off when used in GCODE or output with echo. It happens under specific conditions only, and I think I have determined which. Please see the isolated code fragments below, and especially the in-line comments:

Case #1

Note that the number of decimals when displaying variable 'c' in the output will be limited by the number of decimals of variable 'a'.

var a = 0.1
; The problem seems to be triggered by the division operator
var b = pi / 42
var c = var.a + var.b

echo "a: " ^ var.a
echo "b: " ^ var.b

; Expected 0.1747998, but getting 0.2
echo "c: " ^ var.c

; Shows 1.7, 17.5 and 174.8 respectively, with one decimal.
; But the internal precision seems to be there still!
echo "c * 10: " ^ var.c * 10
echo "c * 100: " ^ var.c * 100
echo "c * 1000: " ^ var.c * 1000

Case #2: One more decimal

Note that this is identical to case #1, however, variable 'd' now has two decimals. Even when the trailing decimal is a zero, this affects the number of decimals in the output of variable 'f'.

var d = 0.10
var e = pi / 42
var f = var.d + var.e

echo "d: " ^ var.d
echo "e: " ^ var.e

; Expected 0.1747998 as output, but getting 0.17 now
echo "f: " ^ var.f

; Shows 1.75, 17.48 and 174.80 respectively, with two decimals
echo "f * 10: " ^ var.f * 10
echo "f * 100: " ^ var.f * 100
echo "f * 1000: " ^ var.f * 1000

Case #3: Workaround

Note that there are no problems when the division operator is replaced by a multiplication operator when calculating variable 'h'.

var g = 0.1
var h = pi * 0.023809524
var i = var.g + var.h

echo "g: " ^ var.g
echo "h: " ^ var.h

; Expecting 0.1747998 and getting it
echo "i: " ^ var.i

I hope you can reproduce the problem. I'm on RRF 3.4.0 beta 5.

T3P3Tony

@schmart thanks for this report. I have just checked and I can reproduce the issue so will flag it up to @dc42.

Schmart

@t3p3tony Thanks Tony!

dc42

@schmart thanks for reporting this. It is now fixed.

25/10/2021, 18:17:02 	M98 P"0:/macros/Rounding bug 1"
a: 0.1
b: 0.0747998
c: 0.1747998
c * 10: 1.7479982
c * 100: 17.4799824
c * 1000: 174.7998352
25/10/2021, 18:16:45 	Connection established
25/10/2021, 18:16:27 	Upload of Duet2CombinedFirmware.bin successful after 1s
25/10/2021, 18:07:44 	M98 P"0:/macros/Rounding bug 1"
a: 0.1
b: 0.0747998
c: 0.2
c * 10: 1.7
c * 100: 17.5
c * 1000: 174.8

dc42

I made a further improvement to avoid printing excess decimal places:

25/10/2021, 18:29:53 	M98 P"0:/macros/Rounding bug 1"
a: 0.1
b: 0.0747998
c: 0.1747998
c * 10: 1.747998
c * 100: 17.47998
c * 1000: 174.7998

Schmart

@dc42 @T3P3Tony Wow! Fixed in minutes, in spite of all the 3.4 development happening: that's next level. I really appreciate this!