Specifically within it's src_install, there is the following: # Install examples docinto examples dodoc examples/* || die "dodoc failed" docinto examples/simple dodoc examples/simple/* || die "dodoc failed" Think it's pretty obvious from the second dodoc, that the first will encounter a directory; however that ebuild is EAPI=3, and the rules of dodoc are "directories, without -r, are a nonzero exit". You're not seeing it blow up because portage, go figure, broke compliance with the spec about 2 months back. That issue is being sorted in 356389; meanwhile, the ebuild needs fixing. I'd suggest just doing dodoc -r examples/* || die "dodoc failed on examples/*"
See bug 356389, comment #3. I can confirm that dev-python/pyopenssl-0.12 produces the following eqawarn log: QA Notice: dodoc argument 'examples/simple' is a directory
Your suggestion of dodoc -r doesn't actually seem to work: dodoc -r examples/* || die "dodoc failed on examples/*" results in: * QA Notice: file does not exist: * * dodoc: -r does not exist
(In reply to comment #2) > * QA Notice: file does not exist: > * > * dodoc: -r does not exist This is a quirk of trying to use dodoc -r with EAPI < 4. If you use EAPI 4 then it should work.
So dodoc with directories started to fail in EAPI=3, but -r doesn't work until EAPI=4? That seems kind of... silly.
(In reply to comment #4) > So dodoc with directories started to fail in EAPI=3, but -r doesn't work until > EAPI=4? That seems kind of... silly. According to PMS it's supposed to fail in all EAPIs less that 4. It's effectively a break in backward-compatibility, as I mentioned in bug #356389, comment #13.
I replaced dodoc with doins.
(In reply to comment #4) > So dodoc with directories started to fail in EAPI=3, but -r doesn't work until > EAPI=4? That seems kind of... silly. EAPI4 was when dodoc -r was added; pardon, I wasn't clear in the implicit eapi4 bump required for the ebuild...