Hi all, Quoting Sebastian Ramacher (2023-05-24 21:47:19) > Adding Protected: yes to adduser has broken piuparts (and maybe other tools > that we do not know about). As the time is running out to experiment with > other solutions to this bug, please put back the dependency on adduser for > bookworm. This change restores the state from bullseye and was agreed on in > today's RT meeting. > > After the release, dropping the adduser from Depends and all other changes > can be revisisted. can we revisit this now? :) To make sure there are no regressions, I just installed all 11911 binary packages that come with a postrm or prerm script using the same script as last time. Out of those 11911 binary packages, 450 either failed to get purged or did something related to adduser. Out of these 450, 147 packages used adduser but succeeded to get removed. So 303 binary packages fail this script. Out of those 303 failures, 15 are in the essential-set and thus fail the script. Out of the 288 remaining failures, 139 failed to install in the first place. Out of the 149 remaining failures, 130 failed because adduser itself failed to remove because of the packages depending on it. 19 failures remain and I investigated the maintainer scripts by changing the test-script such that it would no longer create a dummy-adduser and thus no longer trigger an error in cases where the maintainer script first calls something like `which` to figure out if the tool like adduser/deluser is available. All of the 19 remaining packages succeed now except for one: matrix-sydent. And that I already found last time and filed #1035844 for it. Good news is, that matrix-sydent is not in testing. If anybody would like to see the installation logs to verify my findings, I can send them over. They are just half a MB xz-compressed. So, could apt drop its dependency on adduser (again) now? Thanks! cheers, josch
Attachment:
signature.asc
Description: signature