math reports negative values as positive when in octal or hex, also mishandles overflow

[ridiculousfish]

On fish master commit 1261b53:

> math --base hex -10
0x0
> math --base hex 'pow(2, 40)'
0xffffffff

this is because the value is converted to a (presumably 32 bit) int, and then cast to an unsigned int before it is printed.

[floam]

I think you have this backwards.

[floam]

Current master:

floam@M1 ~> math --base hex -10
0x0
floam@M1 ~> math --base hex 'pow(2, 40)'
0xffffffff
floam@M1 ~> echo $version
3.3.1-573-g3de63f7e2

Because I just fixed it, no? Am I nuts?

[floam]

older versions had the output you are quoting, right?

[floam]

Pulling up the last .app bundle, or an old build from last week:

floam@M1 ~> echo $version
3.3.1
floam@M1 ~> math --base hex 'pow(2, 40)'
0x0
floam@M1 ~> math --base hex -10
0xfffffff6

[ridiculousfish]

Yeah, was testing with the wrong fish. I'll update the issue.

[floam]

Oh gotcha, you actually fixed the output to not be naive, rather than not just totally backwards.

[floam]

Math can always understand its own output regardless of which --base it is running in, this just regressed for negative octal:

> math -0x50 # works
-11
> math -050 # should be -40
-50

[floam]

Nevermind; it was never working/doing that for octal input, negative or positive.