Thanks for the patch! Unfortunately, the tests are still failing with it... Same error locally on my machine.
It seems they use wtmp on Debian as well (previous changelog disabling utmpdump was on Debian) and the syslogd test passes without issues on Debian, so that must be "utmp-compatible" or something indeed. I will try to see if I can find more data on wtmp, and I also asked for more help on the Ubuntu side, especially for insights from people knowledgeable around utmp and what we might do differently on Ubuntu. I will keep you updated. Thanks, Valentin On Thu, 13 Nov 2025 at 12:01, Erik Auerswald <[email protected]> wrote: > > Hi Valentin, > > On Thu, Nov 13, 2025 at 11:11:19AM +0100, Valentin Haudiquet wrote: > > > > It seems that utmpdump is not built on Debian / Ubuntu since December 2024. > > > > > util-linux (2.40.2-12) unstable; urgency=medium > > > > > > * Stop building utmpdump, should not be necessary on wtmpdb systems > > > * Stop marking mkfds-socketpair test as failing > > > * Skip tmpfs-sensitive tests if fstype cannot be determined > > > * Amend NEWS entry to hint updating /root/.profile (Closes: #1088009) > > > > > > -- Chris Hofstaedtler <[email protected]> Fri, 06 Dec 2024 13:35:11 +0100 > > > > I can try to compile that from source, but I wonder if the output will > > be relevant then, given that they decided to stop building it. I'm not > > familiar with "wtmp" as well, so this gives me something more to read > > on ! > > wtmpdb is new for me, too. Anyway, I'd hope that the new "wtmpdb" > provides compatibility for applications using utmp, such as syslogd. > But "utmpdump" reads a file in utmp format, not in wtmpdb format, so it > seems plausible that "utmpdump" won't work with a wtmpdb system. > > Perhaps you could test the attached patch that should make syslogd skip > all utmp entries that are not of type USER_PROCESS. It does not change > anything for me, but if the non-TTY entries were of a different type, > e.g., LOGIN_PROCESS ('Session leader process for user login'), then this > should skip them. This might fix the syslogd test failures you see. > > Depending on which code is used for reading utmp entries, it could already > apply such a filter, but it seems to me as if the code path selected on > my system does not do that. > > I have no idea if this patch breaks compatibility with non-GNU/Linux > systems. I consider it more of a debug patch than an attempt to fix > the problem. > > Thanks, > Erik
