When time command is given data on stdin, it either prints the man page for time or doesn't notice non-zero subprocess exit or show stderr

[terrycojones]

fish version 3.4.1 on OS X 12.2.1

If I run yes | time cat lkslkjlklkslsj (where the file passed to cat is non-existent), I get a prompt back immediately, nothing appears on standard error, and the exit status ($status) is 2. The same thing happens if I run e.g., cat /etc/passwd | time cat lkslkjlk. When I do the same thing without time I get the expected message about a non-existent file on standard error from the cat and the exit status is 1. So it appears time has detected the non-zero exit status but suppressed the standard error output and it exits with 2.

Bizarrely, if I run echo hi | time cat lskjlsjsjlskjlskj, fish displays what I see when I run man time (including piping it to what seems to be more (not less which is what I have $PAGER set to))! Maybe because these are both builtin commands fish does something unexpected (at least to me).

[faho]

time is a decorator for the entire pipeline.

You need time yes | cat. Same thing applies to not.

[terrycojones]

Hi @faho - thanks for this explanation.

I would suggest (though I don't know who to) that time should not have the current behaviour if stdin is not a tty. It's obviously wrong for a fish user to do what I did (though the man page gives no indication of this, and the fish behaviour is not the behaviour of /usr/bin/time (which passes on the stdin and also prints the stderr of a failing command). So accepting that the fish version of time is going to behave as it does, it seems like it would be a good idea for it to tell you when you've made this error. Silently dropping the stderr output of the failing command seems quite wrong.

Anyway, I'm not sure who I should be saying this to, if anyone! Thanks again for the reply.

[ridiculousfish]

I added an error message for when time is used in a pipeline.

[terrycojones]

Great - thanks!