Improve reliability of NO_COLOR tests.

[dag-erling]

Use env to ensure that NO_COLOR is set to the desired value (or unset) regardless of any initialization that may occur when script forks a shell to run jq in.

env -u is non-standard

[dag-erling]

Do you target any platforms where it's not supported?

I suppose I can remove the -u and revert the removal of the unset line and we'll be no worse off than before. I don't dare try env -i (which is portable) because there might be circumstances where we need PATH etc. to be set up just right to run the tests correctly. Another option is to wrap the command in sh -c "..." so we can set and unset variables without relying on env (e.g. faketty sh -c "unset NO_COLOR; $JQ_NO_B -n ." and faketty sh -c "NO_COLOR=1 $JQ_NO_B -n ."). What do you think?

(The problem I'm trying to fix is that the NO_COLOR=1 case is failing on my machine because NO_COLOR is not being passed through faketty and I can't for the life of me figure out why or reproduce it outside of the context of shtest.)

[emanuele6]

Do you target any platforms where it's not supported?

I don't know about others, but one of the sytems I usually use to test things is my sdf.org shell account that runs a ~15yo NetBSD, and env there doesn't have it.

I suppose I can remove the -u and revert the removal of the unset line and we'll be no worse off than before. I don't dare try env -i (which is portable) because there might be circumstances where we need PATH etc. to be set up just right to run the tests correctly. Another option is to wrap the command in sh -c "..." so we can set and unset variables without relying on env (e.g. faketty sh -c "unset NO_COLOR; $JQ_NO_B -n ." and faketty sh -c "NO_COLOR=1 $JQ_NO_B -n ."). What do you think?

You can't exactly do faketty sh -c '': look at the faketty() { script -qec "$*" /dev/null; } implementation, but you can probably do something like that tweaking the implementations a bit.

(The problem I'm trying to fix is that the NO_COLOR=1 case is failing on my machine because NO_COLOR is not being passed through faketty and I can't for the life of me figure out why or reproduce it outside of the context of shtest.)

This test does not look very robust, I am also noticing that it does not set SHELL to /bin/sh before invoking script.
At least the util-linux implementation of script invokes $SHELL, not /bin/sh, to run the specified command, so it will be invoking the login shell of the current user that may not be an sh shell (e.g. tcsh, zsh, fish) or may be /bin/false or something else. That should also be fixed.