On Sat, Mar 04, 2017 at 09:55:30AM -0500, Michael Orlitzky wrote: > On 03/04/2017 09:37 AM, Dutch Ingraham wrote: > > > > Michael, thanks for your response. No, I did not do a one-shot; llvm > > was brought in by way of mesa -> gallium; this is llvm's only use on > > this system as far as I know. > > > > Also, 'emerge -ac' shows no packages to remove. > > > > Well, there goes my one good idea =) > > You can try doing "emerge -pe --tree @world" to see if llvm would get > pulled in by anything in your system. If it is, then a --deep update > --with-bdeps should be updating it. > > One more desperate attempt: the --complete-graph option is weaker than > --deep, I think. What happens if you remove it? (I'm wondering if > --complete-graph overrides --deep). > > If neither of those experiments are illuminating, you should file a bug. > The portage team has a better understanding of why some things are skipped.
Removing --complete-graph doesn't change anything. I don't usually use that, but only added it to see if it would shake something out, but of course it didn't. The emerge -pe --tree @world returns, in relevant part: [nomerge ] mail-client/thunderbird-45.7.0 [nomerge ] x11-libs/gtk+-2.24.31-r1 [ebuild R ] gnome-base/librsvg-2.40.16 [ebuild R ] x11-libs/pango-1.40.3 [ebuild R ] media-libs/harfbuzz-1.4.3 [ebuild R ] x11-libs/cairo-1.14.8 [ebuild R ] media-libs/mesa-17.0.0 [ebuild U ] sys-devel/llvm-3.9.1-r1 [3.7.1-r3] USE=[clip] So, that will bring in the update, just like emerge -1a sys-devel/llvm will. But, why isn't --deep @world doing so? Is it bug-reporting time? (There is one other slight possible anomoly I could find: 'equery depends sys-devel/llvm' returns llvm as a dependency of itself: gentoo3 ~ # equery depends sys-devel/llvm * These packages depend on sys-devel/llvm: media-libs/mesa-17.0.0 [cut massive amnount of non-llvm-related options] sys-devel/llvm-3.7.1-r3 (>=sys-devel/llvm-3.5) gentoo3 ~ # Is this relevant or expected?) Thanks again.
