i've seen too many reports of this to ignore... it seems to be unrelated to bug 93159, but we can't be sure
5.18 is p.mask'd due to this issue
xlockmore-5.17 works, so that is ~arch'd in the tree... still working on 5.18
Created attachment 61810 [details, diff] xlockmore-5.18-passwd.diff I am going to try to narrow this problem down, but this worked for me.
Alright, basically commenting out the line "pam_error = pam_acct_mgmt(pamh, 0);" from passwd.c seems to fix the problem. Now, if you really want to be amused, the following fixed the problem for me too. Add the line "fprintf("debugz0r: %s/n", &pam_error);" bellow "pam_error = pam_acct_mgmt(pamh, 0);" and it may actually work as well. Don't ask why.
this sed magic went into 5.18-r2 (and a call for testing on -dev ML to ensure that it works for everyone) sed -e '/pam_error = pam_acct_mgmt(pamh, 0);/a\ fprintf("stupid pam fix %s/n", &pam_error);' \ -i passwd.c
fprintf() takes a FILE pointer as first agrument, so: fprintf(stderr, ... or printf(...
using printf causes the bug to reappear
bah, i think i'm just going to wait for 5.19 to come out (upstream dev says it should be pretty soon) for now, 5.17 works just dandy
marking LATER