Help with upgrading old portage while using EAPI=5 profile

colag · Tux's lil' helper Joined: 26 Sep 2009 Posts: 105

After one year i did emerge --sync

Then did emerge dhcpcd

Got error:

colag · Tux's lil' helper Joined: 26 Sep 2009 Posts: 105

emerge --sync

eselect profile list

NeddySeagoon · Posted: Sun Apr 06, 2014 10:54 am Post subject:

colag,

Your emerge --sync removed the profile you had selected and you have not yet set a new profile.

khayyam · Posted: Sun Apr 06, 2014 1:41 pm Post subject:

colag ...

it looks as though your currently installed portage doesn't support EAPI5 ... the following should resolve the issue (you may want to re-merge subsequent to using the binary package) ...

Hu · Moderator Joined: 06 Mar 2007 Posts: 21631

OP: what is the output of emerge --info? How long has it been since your last system update?

desultory · Bodhisattva Joined: 04 Nov 2005 Posts: 9410

Merged three posts, and updated the topic title.

While I generally agree with khayyam's approach, there are a few refinements that I would suggest:

A different PORTAGE_BINHOST, specifically: http://tinderbox.dev.gentoo.org/default-linux/x86/.
If, as expected, the install fails when using emerge directly, unpack it as per khayyam's directions, then use the newly unpacked portage to itself install the the downloaded package, so as to correctly handle orphans and general package manager book keeping.

NeddySeagoon · Posted: Mon Apr 07, 2014 6:23 pm Post subject:

In this situation, I would employ a hack.

Edit /usr/portage/profiles/eapi and change the content from 5 to 4.

Any EAPI=5 ebuilds will fail horribly as your portage cannot parse them. Its very ugly but harmless.
Provided that portage and its dependencies are EAPI <= 4 then this will allow you to

desultory · Bodhisattva Joined: 04 Nov 2005 Posts: 9410

wrc1944 · Posted: Sat Oct 04, 2014 12:41 am Post subject:

I just ran into this updating an old Gentoo amd64 install. I had the old "default/linux/amd64/10.0/no-multilib" profile, and after syncing switched to default/linux/amd64/13..0/no-multilib.
Then I got got \:

dol-sen · Posted: Sat Oct 04, 2014 2:50 am Post subject:

You should find an old enough tree in ZeroChaos dev space. He has some old trees for such a purpose.
_________________
Brian
Porthole, the Portage GUI frontend irc@freenode: #gentoo-guis, #porthole, Blog
layman, gentoolkit, CoreBuilder, esearch...

Zero_Chaos · Developer Joined: 22 Jun 2013 Posts: 6

There should have been a news article that showed up, but with a system this old I wonder if it doesn't support news either...

wrc1944 · Posted: Sat Oct 04, 2014 3:29 am Post subject:

Thanks for the replies!
I got the 20120720 snapshot, extracted it, then changed back to my old profile, and tried emerge -1 portage, but it seems the version of portage it wants is no longer available (on several mirrors).

Zero_Chaos · Developer Joined: 22 Jun 2013 Posts: 6

...we suggested the snapshots on my devspace for a reason. They were taken just before the eapi profile transition and meant exactly for what you are doing...

wrc1944 · Posted: Sat Oct 04, 2014 4:53 am Post subject:

Zero_Chaos,
Thanks much! I just got your 20140311, and will try it. I've had no luck with 4 other older 2013 versions.
_________________
Main box- AsRock x370 Gaming K4
Ryzen 7 3700x, 3.6GHz, 16GB GSkill Flare DDR4 3200mhz
Samsung SATA 1000GB, Radeon HD R7 350 2GB DDR5
OpenRC Gentoo ~amd64 plasma, glibc-2.36-r7, gcc-13.2.1_p20230304
kernel-6.8.4 USE=experimental python3_11

wrc1944 · Posted: Sat Oct 04, 2014 5:18 am Post subject:

Hmmm. I must be doing something wrong- it doesn't seem to work whatever I try (profile changes, portage tree versions, EAPI file edits, etc) I still get:

Zero_Chaos · Developer Joined: 22 Jun 2013 Posts: 6

update portage first, if you can't update that then nothing else matters.

Naib · Posted: Sun Oct 05, 2014 8:51 am Post subject:

Would be good if there was a policy that portage keeps with a very basic eapi standard (even if it meant an overly verbose ebuild);as this should help with things like this.

But this raises an old topic... How old of a system is the Gentoo toolsuite meant to be upgradable from
_________________

Navar · Guru Joined: 20 Aug 2012 Posts: 353

wrc1944,

Just tossing out some ideas since I'm running into this today on an older x86 backup (2012 and this time I didn't try stripping it down pre-sync--probably a mistake). I followed Zero_Chaos' steps to revert back to an old snapshot. Still ran into the EAPI 5 issues, hacked those to 4 in respective files and ended up running emerge in debug to see why backtracking was failing. In my case it was struggling to update to dev-lang/python-3.3.5-r1 and that was due to tk useflag being set which ends up being a specific EAPI 5 dependency. Maybe see if you have any specifics set for python in /etc/portage.

Now it's moved forward to the sandbox issue, which seems familiar.

wrc1944 · Posted: Mon Oct 06, 2014 10:14 pm Post subject:

I finally succeeded in a full upgrade, but what a nightmare! My main problems were aggrevated by the fact I forgot this was a no-multilib 64bit system that used to run kdesvn live at some point about 2-3 years ago (not to mention a 3.1.0 kernel, with old radeon drivers which made the gui virtually unusable). Plus, until I saw a post by dilfridge on another thread saying eselect profile now couldn't revert to a 10.0 profile and it had to be done manually, I was going in circles, even with the Zero_Chaos portage tree.

After I got the profile to the old one, cleaned up package.use and package.mask, etc., I still was stuck. Had countless unsolvable udev/systemd blocks (among many other qt, kde, and gnome deps blocks/conflicts) even though I had applied all the config edits that worked before on other systems. Couldn't run --depclean at all, so I manually uninstalled all of the 2-3 year old kde and gnome related stuff, plus any deps of those desktop environments I could find. Still no --depclean, or no emerge -upDNv --with-bdeps=y @ system was possible. Then I found and removed about 20 other systemd and gnome/kde deps I had missed before. Then I decided to emerge --sync, did so, and then changed back to the newer 13.0 no-multilib profile.

After fixing some more random USE flag problems, things were going better. I was able to emerge -u gcc, glibc, binutils, libtool, and then got through an emerge --update --newuse --deep --with-bdeps=y @system. Then I reinstalled kdebase-meta, and a few other kde metas, which went well.

I thought I was finally OK, but I still couldn't run emerge @preserved-rebuild or revdep-rebuild, perl-cleaner, python-updater, and the emerge --update --newuse --deep --with-bdeps=y @world update kept presenting numerous blockages and conflicts I gradually sorted out (USE flag, world file, mask problems). Finally got it down to only the dreaded udev/systemd conflict, and after fumbling around I realized that udev and a few others were built with the wrong USE flags. After fixing those, I could finally run --depclean, which removed 467 packages, including the just installed kdebase-meta (at that point, I didn't care about kde- just wanted a full --depclean). After that, I could finally run revdep-rebuild, perl-cleaner, python-updater, and a full @world update, and the old 2-3 year old installation was back to current, with revdep-rebuild all consistent.

I probably have left out a few problems I encountered, but I finally reinstalled a full kde-meta desktop, and Case closed.

One thing I know for sure- I'm never doing this again. Any over 1 year-old neglected Gentoo install will now warrant a data backup and complete format/reinstall with the latest live cd.

EDIT: Forgot to mention- I did get a 3.16.3-gentoo kernel built IIRC right after I finally did the full @system update, and a reboot made the gui usable with an LXDE desktop.
_________________
Main box- AsRock x370 Gaming K4
Ryzen 7 3700x, 3.6GHz, 16GB GSkill Flare DDR4 3200mhz
Samsung SATA 1000GB, Radeon HD R7 350 2GB DDR5
OpenRC Gentoo ~amd64 plasma, glibc-2.36-r7, gcc-13.2.1_p20230304
kernel-6.8.4 USE=experimental python3_11

Spankin Partier · n00b Joined: 15 Jan 2007 Posts: 33

Just ran into this problem on an old headless file/web (internal use only) server. By the looks of things it wasn't synced in over 5 years.

So from what I see I can attempt the above trick of using an older portage snapshot and go from there. Or since the system is so old, would I be better off backing up the data files, /etc, and /var/lib/portage so that I can download and install the latest Stage3 and install it over top of the existing system. Then restore /var/lib/portage, check my /etc/portage/make.config and run a emerge -aDuN @world?

I'm guessing my biggest problem would be my /etc directory. Since extracting a Stage3 will overwrite all my configs.

The only other option would it be better to simply start completely from scratch.

Any thoughts?

khayyam · Posted: Tue Dec 16, 2014 11:15 am Post subject:

Spankin Partier · n00b Joined: 15 Jan 2007 Posts: 33

Cool thanks.

The one thing I've always loved about Gentoo is the fact that out of the dozens of installations I've managed over the years I've only lost one installation do to software issues (and that to was due to a lack of timely maintenance). Time to say a few words and send him off to the frigid Zeroed Sea.