Opened 2 years ago
Last modified 22 months ago
#17759 new bug
Updating after a previous working state is restored made the system unbootable.
| Reported by: | Nexus-6 | Owned by: | bonefish |
|---|---|---|---|
| Priority: | normal | Milestone: | Unscheduled |
| Component: | Servers/package_daemon | Version: | R1/Development |
| Keywords: | Cc: | ||
| Blocked By: | Blocking: | ||
| Platform: | All |
Description
The description of all the steps, symptoms and finally my personal workaround are described here:
https://discuss.haiku-os.org/t/solved-hrev56131-no-boot/12217/9?u=nexus-6
I am not able to provide a syslog because the system never hit the bootloader after the reboot, moreover I did not captured the syslog before rebooting the system. Nevertheless, I don't believe that nothing would have been written to it as this a userland process.
My personal summary is that there are two scenarios which seem related:
- updating a system either from SoftwareUpdater or pkgman temporarily puts the system in an inconsistent state until reboot. This seems to be the same process as restoring the system to a previous state from Haiku boot menu.
- Updating from a restored state is not well supported and potentially leads to a non bootable system (like occurred to me).
I hope this details would be sufficient for a triage.
Change History (2)
comment:1 by , 2 years ago
| Component: | System/runtime_loader → Servers/package_daemon |
|---|---|
| Keywords: | restore update failure unbootable removed |
| Owner: | changed from to |
comment:2 by , 22 months ago
| Milestone: | R1/beta4 → Unscheduled |
|---|
Note:
See TracTickets
for help on using tickets.
The first scenario I have encountered and is definitely reproducible. The second I have never encountered, and it sounds like something else went wrong: if packages are left in a "transaction" directory, this means the upgrade was not actually completed; so you probably rebooted while your system was in an inconsistent state and thus broke it. Probably we can and should implement safeguards to prevent this from happening, but first we would have to determine how to trigger it, or at least guess.