Lennart Poettering
Lennart Poettering
> @poettering @tuxedo-wse I am not in favor of adding a property to tell systemd to not do its save/restore brightness thing just to work around the current _and temporary_...
So, with this change the PR_SET_PDEATHSIG logic will break (i.e. we make sure that the pam session process will be killed by the kernel whenever the main service process dies)....
> Hmm, but if we do track sd-pam process separately, then the order shouldn't matter and this can be left as-is? well, the only reason we do the weird exec-in-parent-rather-than-child-after-pam_open_session()...
i mean, the main service process really should be the a child of the service manager, so that it can correctly pick up exit code/status from it via waitid(). There...
i'd really like to see some switch btw, i.e. --mutable= or so (or --writable=), which parses as a boolean, but also takes additional values. i.e. if "yes" you get the...
looks pretty good, but I'd be really keen on having the switch right-away
btw, we should add atomic replacement of the overlayfs mount with this (using the new moutn beneath kernel thing), so that you can do `systemd-sysext --mutable=yes refresh` and `systemd-sysext --mutable=no`...
needs man page update
So I think we are getting close, but I am not happy about the choice of dirs for this. Let's come to an agreement oon the spec on this first.
oh, please squash commits. We want "perfect" commits, i.e. bisectable stuff, where each commit is a logical step, but not necessarily historical one. i.e. do not add commits to a...