I thought I'd use 'equery hasuse test' to find all packages that have a test phase, but $SUMMARY. dev-lang/python dev-util/dejagnu sys-apps/coreutils sys-apps/grep sys-apps/sandbox sys-apps/sed sys-apps/texinfo sys-devel/autoconf sys-devel/automake sys-process/procps Reproducible: Always
If there is no need for IUSE=test, then it should not be set. One particular use case would be to set test dependencies (DEPEND="... test? ( cat/pkg )") but I don't see how this would apply in the packages you pointed out. Having a test phase does not mean IUSE=test should be set.
Please suggest a means of finding all packages that have a test phase.
Moreover, the current state of things is confusing: a lot of packages have IUSE=test. That might intuitively suggest something to the users: when a package has a test phase, it exports the "test" IUSE. They might make decisions based on that. But misleadingly, this turns out to be false. If "test" should not exist as a IUSE, please, remove it from all packages that do export it. And there are many, because the situation of "test" missing in IUSE has been considered an actual bug in the past and fixed as such repeatedly.
(In reply to comment #2) > Please suggest a means of finding all packages that have a test phase. `grep --include='*.ebuild' --include='*.eclass' -r src_test /usr/portage` (In reply to comment #3) > Moreover, the current state of things is confusing: a lot of packages have > IUSE=test. That might intuitively suggest something to the users: when a > package has a test phase, it exports the "test" IUSE. They might make > decisions based on that. But misleadingly, this turns out to be false. > > If "test" should not exist as a IUSE, please, remove it from all packages > that do export it. And there are many, because the situation of "test" > missing in IUSE has been considered an actual bug in the past and fixed as > such repeatedly. If a reasonable problem, please file such things against individual maintainers.