On 2018-11-14 12:23, Ludovic Courtès wrote:
Hello,
swedebugia <swedebu...@riseup.net> skribis:
"Note: if your have wound up with a faulty guix after a "guix pull" you can
manually run an older version by looking at the symlinks in your /user/home/.config/guix/
directory.
E.g. if it lists:
$ ls -l /root/.config/guix/
total 0
lrwxrwxrwx 1 root root 33 Nov 4 01:24 current ->
/root/.config/guix/current-1-link
...
Then your faulty guix is in current-1-link, and your former probably well
working guix is in an older symlink, e.g. current-2-link.
To run the guix-version before the last pull in the example above, run:
/root/.config/guix/current-2-link/bin/guix
Now go ahead and use that older guix to roll back as described above."
What do you think?
I think you’re describing a terrible bug, but a bug that’s behind us
AFAIK. I’m reluctant to documenting an old bug that’s hopefully no
longer relevant; I think it would clutter the manual.
What makes you think it is behind us?
There are probably a handful of our users who has old guix installations.
This will not go away anytime soon unless we state very clearly how to
deal with it when they try to "pull". Maybe the manual is not the best
place for this.
We could start by putting a big WARNING to users somewhere noticeable
when the guix they run is deprecated beyond repair.
E.g. before 1.0 implement something that we can trigger to make older
guix signal to the user that they should reinstall instead of running
"guix pull" to no avail.
E.g.
"*** Warning: Your guix is so outdated that guix pull is not supported.
Please reinstall or otherwise obtain a newer guix binary e.g. by guix
copy from another machine.***
Thoughts?
--
Cheers
Swedebugia
