On 23/02/2014 18:13, Alan Mackenzie wrote: > Hello, Alan. > > On Sun, Feb 23, 2014 at 05:22:15PM +0200, Alan McKinnon wrote: >> On 23/02/2014 14:13, Alan Mackenzie wrote: >>>> - are you sure that's an emerge failure and not just a convoluted info >>>>> message? Perhaps post the entire emerge output. >>> I tried it again without the -p, and got the same output. > >>> I think this is a portage bug. At the very least, it's poor >>> documentation. I've reported the situation to bugs.gentoo.org, bug >>> #502236. > >>> Thanks for the help. > > >> I don't think you have a portage bug as such (other than the sloppy >> bizarre output messages that are going into recent versions). I think we >> have bug in an ebuild, probably a maintainer that doesn't quite know how >> to navigate these new subslots waters, > > OK. This is a bit philosophical. The way I see it is even if the main > bug is in the libpng ebuild, portage should have a way of protecting > itself against whatever is in the ebuild. Currently it's wedged.
I know what you mean. emerge doesn't work, therefore the system is broken. > >> One of the other replies suggested to unmerge libpng, emerge it back, >> and continue with emerge world, @preserved-rebuild, revdep-rebuild. > > I'll wait a few days on the response to the bug report, just in case > somebody wants me to probe the current state. > >> Chances are this will work around the issue and let you update >> everything. There *is* a chance some package(s) won't work with or won't >> compile with libpng[1] and you'll have to unwind things again. If this >> happens that will be valuable info to add the entry at bgo > >> [1] This happened to me at least once before, I had to package.mask the >> latest version of the library until the tree sorted itself out. IIRC, it >> was libpng then too! > > Surely package management shouldn't be this difficult? Indeed. yum is not this difficult. apt is not this difficult. FreeBSD ports are not this difficult. [Windows OTOH often is this difficult]. The big difference is those are binary distros so they have a somewhat stable and predictable base. Gentoo is not, Gentoo's base is "whatever emerge finds happens to be there". All the complexity, new features and weird verbose messages in portage are not there to make things work, they are there to detect problems when it doesn't work and prevent problem situations from going into the works in the first place. Personally, I think portage has gone too far and the complex solutions are causing problems that are worse than what they attempt to solve. Amzing solutions (like sub-slots) aren't really much use in the real world if the package maintainers use them incorrectly, right? Well that's my 2c. I was quite happy with revdep-rebuild -- Alan McKinnon alan.mckin...@gmail.com
