Re: pkgbase and config file merges
- In reply to: Baptiste Daroussin : "Re: pkgbase and config file merges"
- Go to: [ bottom of page ] [ top of archives ] [ this month ]
Date: Wed, 06 Mar 2024 20:53:18 UTC
On Mar 6, 2024, at 10:12, Baptiste Daroussin <bapt@FreeBSD.org> wrote: > > Le 6 mars 2024 19:07:51 GMT+01:00, Mark Millard <marklmi@yahoo.com> a écrit : >> Baptiste Daroussin <bapt_at_freebsd.org> wrote on >> Date: Wed, 06 Mar 2024 17:24:18 UTC : >> >>> On Wed, Mar 06, 2024 at 08:59:36AM -0800, Roger Marquis wrote: >>>> Automatically overwriting files like passwd is clearly a problem. >>>> Preferred solution would be a list, say [never_overwrite] and/or [ask >>>> before overwriting] and/or [write diff to /var/db/pkgbase/...] in >>>> /etc/defaults/pkgbase specifying existing files that should never be >>>> modified without at least a dialog. >>>> >>>> This plus the cleanup of pkgsave files which should only be written to >>>> /var/db/pkgbase/pkgsave/`date +%G%m%d%H%M` should address the blockers >>>> to more widespread adoption of packagebase and, ideally, make it >>>> mandatory or at least the default option in the next version of FreeBSD. >>>> >>> this can only happen in the bootstrap phase and not otherwise, in normal usage >>> there is always a 3 way merge happening and if the 3 way merge fails there the >>> original is preserved and a file.pkgnew is created with the new content. >>> >>> pkg should always be and remain unattended, once it enter a transaction so there >>> is no plan to make it interactive. >>> >>> I don't understand what you are asking for about the .pkgsave files cleanup. >> >> I've been using find to check for *.pkgsave files to know to deal with >> them if they occur. >> >> Checking now with find, I did not find any *.pkgnew files. I had not >> been aware to look for those. >> >> But I've discovered .pkgtemp.* as a naming convention and for >> it I find: >> >> # find -x / -name .pkgtemp.\* -print | wc -l >> 80 >> >> So it looks like I've more checking and cleanup to do in general for >> my PkgBase experiments. >> >> >> === >> Mark Millard >> marklmi at yahoo.com >> > > .pkgtemp. Should never occur except if pkg crashes I looked at the file dates/times: All from 2024-Jan-20. That would be when I was discovering that use of the RPi5 EDK2 draft ends up with unreliable USB file I/O. So it fits your description. > .pkgsave once you have boostrapped you should not have any. > === Mark Millard marklmi at yahoo.com