Fix the clock being hours off after booting Windows
Checked both cited wiki pages as raw wikitext and man timedatectl from systemd 261.2-1 on this workstation. The direction and the command are current and exact: the System_time page says configuring Windows for UTC is recommended rather than switching Linux to localtime, and the record's reg add "HKEY_LOCAL_MACHINE\System\CurrentControlSet\Control\TimeZoneInformation" /v RealTimeIsUniversal /d 1 /t REG_DWORD /f is byte-identical to the wiki's line. The cause's systemd 216 claim is supported word for word: since version 216, when the RTC is configured to local time systemd will never synchronize back to it. The danger holds on both halves, from the source and locally. The wiki gives DST over-correction across multiple operating systems and system time going backwards during the boot sequence for a localtime RTC, and man timedatectl here says maintaining the RTC in the local timezone is not fully supported, will create problems with time zone changes and daylight saving adjustments, and to keep the RTC in UTC if at all possible. The danger stays unchanged, as do severity medium, frequency very-common, the symptom and the verify. The timesyncd drop-in path /etc/systemd/timesyncd.conf.d/local.conf, the [Time] key, the four *.arch.pool.ntp.org hostnames and timedatectl show-timesync --all all match the Systemd-timesyncd page, and the three lines the verify field asks for appear verbatim in this machine's live timedatectl status output. One Omarchy 4 fact changes the fix. Confirmed on this machine: /etc/adjtime does not exist and is owned by no package, and it is absent from upstream's own fresh-install manifest in omacom/omarchy-iso. The wiki states that when /etc/adjtime is absent systemd already assumes the hardware clock is UTC, so sudo rm /etc/adjtime fails with 'No such file or directory' for most readers. Also confirmed here: timedatectl status already reports RTC in local TZ: no and NTP service: active, and the fresh-install manifest carries the sysinit.target.wants/systemd-timesyncd.service symlink, so timesyncd is enabled out of the box. The Linux half of the record therefore verifies state rather than fixing it, and a reader who runs it and sees nothing change may wrongly conclude the fix failed. The rewrite says that plainly, switches to rm -f, and moves set-ntp true ahead of any RTC write so set-local-rtc 0, which the man page notes also synchronises the RTC from the system clock, cannot push a still-wrong time into the RTC. Overlap checked against the corpus: pgp-signature-invalid-wrong-system-clock reaches the same registry key from the pacman signature symptom and windows-update-takes-over-uefi-boot-order is a different problem, firmware boot order rather than time, so this record stays the general clock one and the boundary holds. Not exercised: I have no sudo, so I ran no timedatectl set-*, no hwclock and no timesyncd restart, and I have no Windows install to test the registry value against.
timedatectl set-local-rtc 1 is the wrong direction and is explicitly discouraged, because it causes over-correction across DST changes and can make the system clock go backwards during boot. Only remove /etc/adjtime if you then immediately reset the hardware clock, or the next boot may come up with a wildly wrong time (which breaks TLS and pacman signature checks).
Symptom
Every time I boot into Windows and come back to Arch/Omarchy, the clock is off by exactly my UTC offset. Fixing it in one OS breaks it in the other. Sometimes HTTPS sites then fail with certificate date errors.
Cause
Both systems read the same hardware (RTC) clock but interpret it differently: Linux treats it as UTC, Windows treats it as local time. Each one "corrects" it on boot and they fight. Since systemd 216, if the RTC is set to local time, systemd will never write back to it, which makes the drift worse.
Fix
The remedy is on the Windows side. Omarchy 4 already keeps the RTC in UTC with NTP on, so the Linux commands below mostly confirm that state rather than change it.
Check what Linux currently believes:
timedatectl status
cat /etc/adjtime 2>/dev/null || echo 'no /etc/adjtime, so systemd assumes the RTC is UTC'
On a stock Omarchy 4 install this already reports RTC in local TZ: no and NTP service: active, and /etc/adjtime does not exist. Linux is not the side that is wrong. The clock keeps moving because Windows writes local time into the same RTC.
Make Windows use UTC. In an Administrator Command Prompt on Windows:
reg add "HKEY_LOCAL_MACHINE\System\CurrentControlSet\Control\TimeZoneInformation" /v RealTimeIsUniversal /d 1 /t REG_DWORD /f
Should Windows later offer to update the clock for a DST change, let it. It leaves the RTC in UTC and only corrects the displayed time.
Back on Linux, set the time zone and get NTP synchronised **before** anything writes the RTC, so a correct system clock is what gets pushed out:
sudo timedatectl set-timezone Europe/London # your zone
sudo timedatectl set-ntp true
timedatectl status # wait for 'System clock synchronized: yes'
Only if timedatectl status still reports RTC in local TZ: yes does the RTC standard need changing:
sudo timedatectl set-local-rtc 0
If systemd-timesyncd is not syncing, point it at the Arch pool via /etc/systemd/timesyncd.conf.d/local.conf:
[Time]
NTP=0.arch.pool.ntp.org 1.arch.pool.ntp.org 2.arch.pool.ntp.org 3.arch.pool.ntp.org
FallbackNTP=0.pool.ntp.org 1.pool.ntp.org
then:
sudo systemctl restart systemd-timesyncd.service
timedatectl show-timesync --all
If a stale drift value in /etc/adjtime is making the hardware clock jump, clear it and write the corrected time straight back. Use rm -f, because on Omarchy 4 the file is normally absent and a bare rm just errors:
sudo rm -f /etc/adjtime
sudo hwclock --systohc --utc
If both machines run a time client, disable synchronisation in Windows so the two do not each estimate RTC drift without knowing about the other.
Verify
timedatectl status shows RTC in local TZ: no, System clock synchronized: yes, NTP service: active. Reboot into Windows and back. The time is still correct.