I want to ensure that flatpak wasn’t updated since a specific date, so that I am able to file a report about its behaviour. [1] rpm -qi flatpak returns:
Install Date: Thu 15 May 2025 12:44:39 BST
However, I don’t know whether that means that it was last updated then, [2] or whether that’s even the correct method to ascertain this.
To see dnf5 transactions that affected flatpak:
dnf history list --contains-pkgs=flatpak
To see dnf4 transactions (which include anything you did from Discover, GNOME Software etc):
dnf4 history list flatpak
Thanks, @pg-tips, as ever. That worked:
RokeJulianLockhart@Beedell:~$ dnf history list --contains-pkgs=flatpak RokeJulianLockhart@Beedell:~$ dnf4 history list flatpak ID | Command line | Date and time | Action(s) | Altered ------------------------------------------------------------------------------------------------------ 36 | | 2025-06-09 20:13 | I, U | 569 < 32 | | 2025-04-18 15:27 | D, E, I, U | 2590 >< 20 | system-upgrade upgrade | 2025-01-16 13:34 | C, D, E, I, O, | 2557 >< 19 | upgrade --refresh -y | 2025-01-16 13:12 | C, E, I, O, U | 723 >< 2 | upgrade | 2024-10-02 22:20 | I, O, U | 1181 >< 1 | | 2024-04-14 23:56 | Install | 2153 >E
Worrisome that Discover still utilises dnf4, though. I’ve been unable to ascertain whether it’s even tracked anywhere (KDE BZ, pkcon’s GH issues, etcetera). I hope that doesn’t cause problems. If you’ve any information about it, although it’s tangential to this thread, I’ll be glad. [1]
See:
Moving to dnf5 requires upstream work on a Discover backend. The KDE bug tracker ticket is:
Yes, “dnf history list –contains-pkgs=flatpak” will work but to answer the other questions:
The “Install Date” in the “rpm -qi flatpak” is the actual time the rpm was upgraded or installed whichever was sooner. There is a time difference between the time shown in the "dnf history list ..” and the one in the rpm -qi for two reasons. That’s because the dnf transaction might have taken significant time. I think the time stamp in the dnf history list is when the dnf transaction started and the one in the rpm -qi is the actual time the package was installed or updated. The dnf history list time is reported in UTC but the rpm -qi “Install Date” time is in local time. So for a person the “Install Date” is probably easier to understand, while the UTC time is better for record keeping (no need to deal with DST).
To show this, I can show my last history entry and the “Install Date” from a package in that:
dnf history info last | grep -e Upgrade -e "Begin time" |head -2
Begin time : 2025-10-26 14:40:45
Upgrade b3sum-0:1.8.2-3.fc43.x86_64 User updates-testing
rpm -qi b3sum | grep Install
Install Date: Sun 26 Oct 2025 10:40:58 AM EDT
So you see the “+4” hours more on the history due to the UTC to EDT difference and a rpm time was about +13 seconds later due to the time dnf took to get to that package
Something I was playing with today.
Setup for dnf5
#!/usr/bin/env bash
set -o errexit
set -o nounset
set -o pipefail
if [[ -z $@ ]]; then
echo "Takes a space separated list of package names for input"
exit
fi
PACKAGES=$@
for PACK in ${PACKAGES}
do
FULL_PKG_NAME="$(dnf rq --installed "${PACK}" --qf '%{full_nevra}')"
if [[ -n "${FULL_PKG_NAME}" ]]; then
PKG_BUILD_TIME="$(dnf rq --installed "${PACK}" --qf '%{buildtime}' | xargs -i date -d @{})"
PKG_INSTALL_TIME="$(dnf rq --installed "${PACK}" --qf '%{installtime}' | xargs -i date -d @{})"
echo "PACKAGE NAME: ${FULL_PKG_NAME}"
echo "PACKAGE BUILD TIME: ${PKG_BUILD_TIME}"
echo "PACKAGE INSTALL TIME: ${PKG_INSTALL_TIME}"
else
echo "Package: "${PACK}" not found"
fi
echo ""
done
That’s a lot of code. This bash function does it in way less space:
lastupdated ()
{
if [[ -z "$1" ]]; then
echo "required package name missing"
return;
fi;
dnf history info 1..last --contains-pkgs="$1" | grep -e "Install\s*$1" -e "Upgrade\s*$1" -e "Begin time" | head -2
}
You could run:
lastupdated packagename
Like I was saying the time this shows would be a few seconds early (but who cares, you usually want the time a package was last installed/updated in a general way). The install/upgrade parts in the grep isn’t really necessary, but it’s there to ensure the time you see isn’t from a history entry for removing the package. Seeing just two time stamps in a row would mean the newest transaction on the package was a “remove”
@grumpey and @marko3209, what do these do?
It lacks some of the safety mechanisms of the first one, in case a user wants to modify it.
Does a repoquery based on a space separated list of package names.
Output looks like:
PACKAGE NAME: gnome-calculator-0:49.1-1.fc43.x86_64
PACKAGE BUILD TIME: Fri Oct 10 05:51:56 AM EDT 2025
PACKAGE INSTALL TIME: Sun Oct 26 02:11:15 PM EDT 2025
or
Package: gnomecalculator not found
My suggestion doesn’t do multiple packages whereas his does multiple ones, but I don’t see any “safety” problems (ie that it’s dangerous to run a “dnf history …” command). Maybe you mean “sanity tests”
@marko3209, I’m referring to baSH’s equivalent to preprocessor directives:
set -o errexit set -o nounset set -o pipefail
They’re similar to pwsh’s Set-StrictMode.
Ahh, okay, but my one liner didn’t need any of that so it doesn’t matter anyway
@marko3209, yes. They’re solely good form in case the script is retroactively modified.