Check for updates on Current

rvp

pfr I was more wanting to know if the use of pkg_chk was the best option here? It takes a a little while to run but I suppose that's to be expected.

Switch to pkgchkxx (now out of WIP).

pfr [...] feel free to point out any better ways of doing things there too).

There should be no need to do doas su bob. doas should let you switch to bob directly. For example, with sudo: sudo -i -u bob sh -c '...' (The -i is to run as a login-shell: don't want the current user's umask, env. vars., current dir., or other junk to pollute things for bob.)
A.I. has done a better job here: it's thrown away all those redundant `;' ending the lines in your version.

pfr I thought about running a cronjob to intermittently check for updates

Check the mod-time of updates.txt and if it's older than some time, run pkgchkxx again. Like:

needupd.sh

#!/bin/sh

set -eu

DELTA=$((1 * 60 * 60))			# 1 hr.

must_update() {
	local FT NOW

	[ -f "$1" ] || return 0		# non-existent: update
	FT=$(stat -f %Sm -t %s "$1")	# GNU: stat -c %Y
	NOW=$(date +%s)
	[ $((NOW - FT)) -gt $DELTA ]
}

must_update "$1" && { S=yes; N=0; } || { S=no; N=1; }
echo "$1: $S"
exit $N

WS="$(expr $ $AWS + 1 $)" # because workspaces begin at '0'

Just use shell arithmetic here: WS=$((AWS + 1))

pfr

rvp Switch to pkgchkxx (now out of WIP).

Thanks I'll give it a try.

rvp doas should let you switch to bob directly.

Again, thank you! There doesn't appear to be an -i flag in doas. I noticed that the FreeBSD man page for doas includes the -S flag for login shell but this doesn't appear to be present in opendoas man page.. Will check later when I get a chance.

rvp A.I. has done a better job here

I think giving it a simple script to begin with and asking it to improve on it was a better idea than just asking it to write a script from scratch. Perhaps this is where copilot can be useful??

rvp Check the mod-time of updates.txt and if it's older than some time, run pkgchkxx again.

Without yet knowing exactly how pkgchkxx differs from pkg_chk I can't accurately I comment on this, but is there any use in running pkgchkxx more than once after updating the pkgsrc tree as the number of availably updates will remain the same until fetching another updated tree?

rvp Just use shell arithmetic here: WS=$((AWS + 1))

Again, thank you. Your helpful replies to my posts are always greatly appreciated.

rvp

pin [...] and I was wondering if you did some measure?

Never tried it. I just happened to recall @yeti's enthusiastic post--mostly on account of that delightfully ambiguous banner: "pkgchkxx is neigh!". Is that, nigh, neigh (horse), or nay? Our playful animal doesn't clarify.

Anyway, this is your dept. 😉

pfr There doesn't appear to be an -i flag in doas.

The OpenBSD doas says:

      By default, a new environment is created. The variables HOME, LOGNAME, PATH, SHELL,
      and USER and the umask(2) are set to values appropriate for the target user.

So, that looks OK (and I like it!).

pfr Perhaps this is where copilot can be useful?

That's the idea, I think: Co-pilot, not Pilot.

@rvp have you tried pkgchkxx vs. plain pkg_chk?
I remember when pho@ intruduced it but, I can't seem to find benchmarks now.

According to the README it's supposed to be a lot faster and I was wondering if you did some measure?

pfr

rvp The OpenBSD doas says:
By default, a new environment is created. The variables HOME, LOGNAME, PATH, SHELL,
and USER and the umask(2) are set to values appropriate for the target user.

Even if I have keepenv in mydoas.conf... ? I'd like to keep that there if possible.
Perhaps I can add another rule specifically for running doas -u bob command?
something like:

permit nopass keepenv :wheel
permit nopass keepenv setenv { PATH } bob as bob

Is this necessary?

pin
Just tested pkgchkxx and it's certainly quicker than pkg_chk

@rvp Not really statistical, as I only did 2 runs else, I'd be awaiting for hours but, here is a "quick" benchmark,

~> hyperfine --runs 2 'pkg_chk -u -q'
Benchmark 1: pkg_chk -u -q
  Time (mean ± σ):     328.340 s ±  0.474 s    [User: 168.849 s, System: 165.107 s]
  Range (min … max):   328.005 s … 328.676 s    2 runs

~> hyperfine --runs 2 'pkgchkxx -u -q'
Benchmark 1: pkgchkxx -u -q
  Time (mean ± σ):     108.847 s ±  0.542 s    [User: 158.440 s, System: 198.379 s]
  Range (min … max):   108.464 s … 109.231 s    2 runs

pfr Just tested pkgchkxx and it's certainly quicker than pkg_chk

"certainly quicker" is not a benchmark but, yes. See above, approximately 3 times faster.

rvp

pfr Even if I have keepenv in my doas.conf... ?

I don't know why you want to compromise security like that; nor why you want to contaminate Bob's environment from the outside... (Just set any env. vars. you want in Bob's ~/.profile or ~/.shrc, or in the script itself, or source it from another file, ...)

pfr is this avoidable, or should I include some shell arithmetic to [...]

Try this:

xx-massage.awk

#!/usr/bin/awk -f
#
# xx-massage -- no happy ending.

BEGIN {
	if (ARGV[1] == "-l") {	# pass `./-l' for literal file `-l'
		docount = 1	# print only line count
		ARGV[1] = ""	# awk... (sigh)
	}
}
{
	if ($0 == "Rechecking packages after deletions") {	# exact match
		N = 0		# just reuse LINES
		next		# skip sentinel
	}
	LINES[N++] = $0
}
END {
	if (docount)
		printf "%u\n", N
	else {
		for (i = 0; i < N; i++)
			print LINES[i]
	}
}

Use as: pkgchkxx -uq | xx-massage.awk | wc -l or more simply, pkgchkxx -uq | xx-massage.awk -- -l

pfr

So I just ran the script for the first time (with the few revisions to the the original above)

And it completed the first function, running cvs update -dP, but once that was complete the script exited without updating the pkgsrc/wip directory...

besides that, I just checked the output of pkgchkxx -u -q and it lists each package twice after "Rechecking packages after deletions"

~ λ pkgchkxx -u -q
devel/glib2 - glib2-2.80.2 < glib2-2.80.2nb2
devel/pkgconf - pkgconf-2.2.0 < pkgconf-2.2.0nb1
fonts/harfbuzz - harfbuzz-8.4.0 < harfbuzz-8.5.0
mail/aerc - aerc-0.17.0nb1 < aerc-0.17.0nb4
multimedia/libass - libass-0.17.1nb1 < libass-0.17.2
multimedia/ffmpeg6 - ffmpeg6-6.1.1nb4 < ffmpeg6-6.1.1nb5
net/yt-dlp - yt-dlp-2024.4.9 < yt-dlp-2024.5.27
pkgtools/pkgin - pkgin-23.8.1nb3 < pkgin-23.8.1nb4
security/gnutls - gnutls-3.8.5nb1 < gnutls-3.8.5nb2
security/gnupg2 - gnupg2-2.4.5 < gnupg2-2.4.5nb1
print/libcups - libcups-2.4.6nb4 < libcups-2.4.8
x11/libxkbcommon - libxkbcommon-1.7.0 < libxkbcommon-1.7.0nb1
x11/gtk3 - gtk3+-3.24.41nb2 < gtk3+-3.24.42
www/firefox - firefox-125.0.3 < firefox-126.0
Rechecking packages after deletions
devel/glib2 - glib2-2.80.2 < glib2-2.80.2nb2
devel/pkgconf - pkgconf-2.2.0 < pkgconf-2.2.0nb1
fonts/harfbuzz - harfbuzz-8.4.0 < harfbuzz-8.5.0
mail/aerc - aerc-0.17.0nb1 < aerc-0.17.0nb4
multimedia/libass - libass-0.17.1nb1 < libass-0.17.2
multimedia/ffmpeg6 - ffmpeg6-6.1.1nb4 < ffmpeg6-6.1.1nb5
net/yt-dlp - yt-dlp-2024.4.9 < yt-dlp-2024.5.27
pkgtools/pkgin - pkgin-23.8.1nb3 < pkgin-23.8.1nb4
security/gnutls - gnutls-3.8.5nb1 < gnutls-3.8.5nb2
security/gnupg2 - gnupg2-2.4.5 < gnupg2-2.4.5nb1
print/libcups - libcups-2.4.6nb4 < libcups-2.4.8
x11/libxkbcommon - libxkbcommon-1.7.0 < libxkbcommon-1.7.0nb1
x11/gtk3 - gtk3+-3.24.41nb2 < gtk3+-3.24.42
www/firefox - firefox-125.0.3 < firefox-126.0

is this avoidable, or should I include some shell arithmetic to somehow subtract one line (the "Rechecking packages after deletions" line) and also divide by two to get the true number of packages?

pfr What exactly do you mean?
I cannot really test this cause, I have no outdated packages but, I don't get the "Rechecking packages after deletions" part on my output.

pfr

pin yeah, I don't really know to be honest... 🤷

pfr

rvp Try this:
xx-massage.awk

The fact that @pin doesn't get the "Rechecking packages after deletions" stuff makes me wonder why I am. Before I use the awk script to tidy up the output I thought I'd simply ask the developer of pkgchkxx to see what happening

rvp I don't know why you want to compromise security like that

Do you mean the permit nopass :wheel ? yeah, a bit loose I know, but I'll take my chances for now. Honestly, that is simply so that I can run the fzf-pkgsrc script together with invoking doas in place of su in my mk.conf and I wont be prompted for a password. Perhaps I should just login as root 😆

rvp Just set any env. vars. you want in Bob's ~/.profile or ~/.shrc

Fair call. The only reason I had keepenv in the first-place is so that when editing files as root using vim I would retain my ~/.vimrc.. but I suppose I should probably just copy it over to /root and also /home/bob and source it.

Question: can I (or more specifically should I) source /home/dave/.vimrc from /root/.profile or /home/bob/.profile?? Or should I just copy the vimrc into each users home directory and source it from there?

My ~/.vimrc changes fairly regularly so it would just be a pain to keep 3 versions of it up to date... Unless I could symlink it across users?? 🤔

rvp

pfr Do you mean the permit nopass :wheel ?

No, the keepenv bit.

pfr My ~/.vimrc changes fairly regularly so it would just be a pain to keep 3 versions of it up to date... Unless I could symlink it across users?

Keep a master copy in /usr/local/etc/vimrc.local, then source it in ~/.vimrc

pfr

rvp
Cool, will do.

Back to my script, any idea why it executes update_pkgsrc then exits before executing update_wip ?

rvp

pfr Back to my script, any idea why it executes update_pkgsrc then exits before executing update_wip ?

I think A.I. goofed (the technical term is, I believe, hallucinated) in the re-write (relevant lines only):

set -e
PKGDIR="/usr/pkgsrc"
doas tar -xzf pkgsrc.tar.gz -C $PKGDIR

It's unpacking the tarball into the wrong directory; and on subsequent runs, cvs update probably fails when it encounters that extra dir.

(Forget Co-pilot mate; read any one of these instead.)

@pfr I'm getting the same recheck as you now ... back to pkg_chk for me.

pfr

pin I'll be sure to let you know if there is a solution to this.

rvp Your link is not working. I could edit it but, others have been pissed-off about such edits, so I'll leave it to you 😉

pfr

rvp Good spotting, but that's not it. my $PKGDIR is /usr/pkgsrc I haven't run this script yet without a current pkgsrc tree in /usr. Also, cvs up -dP succeeds, it's just that the script wont move on to pull from pkgsrc/wip and instead it exits

rvp

pin Oh, go ahead (just don't put in bad poetry--dirty limericks are fine.)

rvp

pfr Great... more A.I. screw-ups for RVP to fix: Your original script had both git clone and git config running as bob. A.I. has decided, for some reason, to run only the first as bob; the other is now executed as root.

 31         echo "Cloning pkgsrc-wip to $WIPDIR..."
 32         doas -u bob git clone git://wip.pkgsrc.org/pkgsrc-wip.git $WIPDIR
 33         doas git config --global --add safe.directory $WIPDIR

However, this may not be it either. Some other command may be exiting with a non-zero status--which you may have to ignore. Run the script like sh -vx whatever args ... That'll show what causes the premature exit.

One other thing: I don't like this script much (I didn't like the original one either...). It's running as 3 different users with, possibly, 3 different enviroments (the keepenv setting makes this worse): most as dave, some as root and others bob. Don't do this! The point of creating bob (with rootly powers), is to do all this build-related stuff as bob. dave should just su -l bob or doas -s -u bob about once a fortnight or something to keep the trees up to date.