coreutils: Should we really support --coreutils-prog=yes too?

[oech3]

uutils/coreutils is supports --list and coreutils yes (mostly for busybox test suite)? But missing --coreutils-prog=.
It is required by test/misc/coreutils.sh. But in my thought, no one use coreutils directly and supporting it could add some overhead.
Should we really support it (or ignore)?

[oech3]

@pixelb

[pixelb]

Yes I see --coreutils-prog as kind of internal, but a means to support "shebangs" wrappers, which are more general than hardlink or symlinks for multiple reasons. Re --list, GNU should probably support that, as it's generally useful

[pixelb]

Yes I see --coreutils-prog as kind of internal, but a means to support "shebangs" wrappers, which are more general than hardlink or symlinks for multiple reasons. Re --list, GNU should probably support that, as it's generally useful

[oech3]

Maybe, Does busybox's test suite depend on coreutils yes instead of coreutils --coreutils-prog=yes too?

[oech3]

Closing until someone complain with

coreutils/util/build-gnu.sh

Lines 173 to 174 in 3108c45

	# We use coreutils yes
	"${SED}" -i "s|--coreutils-prog=||g" tests/misc/coreutils.sh

[oech3]

@pixelb How about adding coreutils cmdname (busybox compat) and removing the flag?

[pixelb]

Yes maybe. It's on my list to look at

[oech3]

Some people are interested in running GNU coreutils against busy(toy)box tests to achieve coveredge.
But, it seems coreutils should not exit 1 when called as *box to run against *box test suite. Related: #10244 #10258

@ChrisDryden

[oech3]

Reopened for the usecase invalid_name --coreutils-prog=true exits 0.