Score:0

Is compiling/testing/using a systemd-resolved not aligned with the version of the rest of systemd present on the system supported?

in flag

Is it possible to compile only systemd-resolved from the large systemd sources (without all the rest of systemd) to attempt identify a fix to backport? Can a new systemd-resolved coexist with the rest of a system that is based on an oldish systemd?

Score:1
cn flag

No, only building and upgrading one binary of systemd is not supportable. As in, people you might ask for help will have difficulty reproducing what you are doing.

resolved, like the dozes of binaries that compose this thing, links to some systemd shared code. This is not likely to remain binary compatible across an arbitrary number of releases. It might work, but personally I'm not willing to untangle systemd's internal dependencies to prove it.

Instead, try upgrading all of systemd. For a start, try reproducing the problem on a distro with a relatively new version. In Red Hat land, as of 2021 this could be Fedora 35 or RHEL 9. Once there is evidence the newer distro improves things, then start isolating relevant changes. Or start an OS upgrade project.

callegar avatar
in flag
Thanks, that clarifies. Unfortunately, replacing the whole of the iniit system risking breaking init or creating subtle inconsistencies with how the services are set up on the distro in order to fix a bug in resolved seems too risky. Easier to eliminate the local DNS caching altogether or - as it is being tested right now - to modify the DNS server on the LAN to avoid triggering bugs in the systemd-resolved version shipped with ubuntu focal (at the cost of slightly deviating from DNS standards no big deal because this will remain local).
John Mahowald avatar
cn flag
Your proposed partial upgrade seems slightly riskier than upgrading all of systemd, but do what you want. If the actual desired beahvior is not use resolved's stub resolver, that is possible with configuration.
callegar avatar
in flag
I understand your point, but I try to evaluate the risk that can get more expensive. I understand that if I break a single package/binary it will be possible for me to boot and to revert easily possibly using the package manager. But if I break the system boot, cleaning up is not going to be that easy, because I'll not get to use the package manager or the usual tools (is this correct? also this will need to be done on multiple machines). Temporarily giving up the stub resolver or working with a workaround in the local DNS server seems a lower price to pay... Thank you anyway for your answers!
mangohost

Post an answer

Most people don’t grasp that asking a lot of questions unlocks learning and improves interpersonal bonding. In Alison’s studies, for example, though people could accurately recall how many questions had been asked in their conversations, they didn’t intuit the link between questions and liking. Across four studies, in which participants were engaged in conversations themselves or read transcripts of others’ conversations, people tended not to realize that question asking would influence—or had influenced—the level of amity between the conversationalists.