On Thu, Aug 21, 2003 at 08:11:21PM -0400, Joey Hess wrote: > Steve Langasek wrote: > > Given that all the files involved were conffiles prior to this > > transition, I think no additional work is needed to correctly support > > systems that are being upgraded. Can you confirm whether > > /etc/pam.d/other already contains 'md5' on the machine above? If not, > > I'll add that to the top of the TODO list. > Yes it would contain md5 on the password line. Ok, thanks. > > A decision still needs to be made about which package should be asking > > the question, I think. Policy would normally require that > > libpam-runtime is the only package editing any of these files; and > > indeed, since the passwd package is not essential, it is conceivable to > > have a system with libpam-runtime installed but not passwd, and the > > question would still be relevant for other services that allow password > > changes. Are there problems with using debconf from a package that's in > > the dependency chain of login (an essential package)? If so, this would > > also pose a problem with trying to poll debconf values to fix > > /etc/pam.d/other on upgrade. > debconf is itself "only" important (and it will never be essential). > However there are ways to use debconf that don't require a dependency on > it. Would you recommend this "best effort" approach for the present case? Will that be effective for d-i, or do we have to worry about libpam-runtime's postinst completing before debconf is available? -- Steve Langasek postmodern programmer
Attachment:
pgpbVizOCfyEM.pgp
Description: PGP signature