I couldn't find any mention of setting up /etc/make.conf to work properly w/ emerge. Here are the settings that seem to be needed - a portage guru should be consulted to make sure the list is definitive. SYNC="cvs://stuart@cvs.gentoo.org:/var/cvsroot" CVS_RSH=ssh CVSROOT=":ext:stuart@cvs.gentoo.org:/var/cvsroot" FEATURES="autoaddcvs cvs digest manifest" Best regards, Stu
SYNC shouldn't be listed in the handbook; not everybody wants to use their cvs checkout as their primary Portage tree. The CVS* variables are unnecessary. Once you've checked out the tree, you don't need CVSROOT in your environment anymore, and ssh is the default CVS_RSH.
FEATURES="autoaddcvs cvs digest manifest" isn't really required either - autoaddcvs is already mentioned there, 'digest' is only a convenience to prevent running ebuild ... digest [ Which not all may want! ] and repoman always updates manifests anyway. I'll add in a note about FEATURES="cvs" though.
avenj, *Strongly* disagree with you w.r.t not mentioning SYNC in the Handbook. It should at least be listed, so that devs have the choice if they wish. Otherwise, how do you expect them to know? plasmaroo, The reference to 'autoaddcvs' is very well hidden, and is easily missed when reading the document. At the very least it should be moved into a more prominent section. All, Two FEATURES that I recommend all devs should run with are 'sandbox' (on supported platforms) and 'buildpkg'. sandbox speaks for itself. 'buildpkg' has an undocumented side effect on ebuilds. With 'buildpkg' set, Portage calls the individual functions of the ebuild in separate processes. This shows up bugs in ebuilds which rely on variables being initialised only in pkg_setup(). Best regards, Stu
I don't disagree that it should be documented, merely that it should be documented as something needed for things to work properly. If we're just talking about things we can make people aware of, then I certainly agree.
"buildpkg" creates binary packages for everything which not all may want - I'll leave that one out for now. Perhaps the "undocumented" behaviour should moved out to a separate FEATURES flag? The rest of the requested changes should be visible at http://dev.gentoo.org/~plasmaroo/gentoo-docs/hb-introduction-new-devs.html - tell me if anything else needs adding or changing... Thanks!
I've just done some testing w/ Portage 2.0.50-r9 and .51-pre13. The buildpkg FEATURE no longer affects whether variables defined in pkg_setup() persist through to src_install() or not. Might be worth thinking about whether or not we should be encouraging devs to test that their ebuilds correctly create binary packages correctly, tho. Best regards, Stu
Ok, the changes requested here seem to be in the document for some time now so I'm closing this as FIXED.