IRC channel logs

<ArneBab>sneek: later tell dthompson: I added it in numbers.test. After some minor adjustments, I see errors: https://codeberg.org/spritely/hoot/pulls/806#issuecomment-8404077

<sneek>Will do.

<dthompson>ArneBab: thanks for checking! turns out guile treats signed inexact zeroes in a special way

<sneek>Welcome back dthompson, you have 1 message!

<sneek>dthompson, ArneBab says: I added it in numbers.test. After some minor adjustments, I see errors: https://codeberg.org/spritely/hoot/pulls/806#issuecomment-8404077

<dthompson>(< -0.0 0.0) is #f but (min -0.0 0.0) should be -0.0 even though they're both zero.

<dthompson>and likewise (min 0.0 -0.0) should be -0.0

<dthompson>out of curiosity I wanted to see how another scheme impementation behaved so I used chibi-scheme. min/max are not commutative like this in chibi

<dthompson>r7rs-small has barely any specification for how min/max should work.

<ArneBab>dthompson: floating point implementations are madness, but getting them wrong causes worse madness :-)

<ArneBab>r7rs-small just links to other standards for arithmetic operations, and if min/max are arithmetic, then as far as I can tell Guile gets it right.

<ArneBab>(but I don’t claim to fully understand that)

<dthompson>given that chibi scheme is kind of the official r7rs scheme, I think chibi is probably compliant with the spec. it's just verrrry loosely specified.

<dthompson>r6rs requires args to be real and r7rs doesn't, for example

<dthompson>so things are kinda weird in chibi: (min 1+1i) is fine because it's the trivial single arg case, but (min 1+1i 1) is an error, as is (min 1+1i 2+2i)

<dthompson>guile's behavior feels very sensible

IRC channel logs

2025-11-22.log