On Sun, Mar 13, 2005 at 07:12:12PM +0200, Ville Skyttä wrote: > > FUD or not, I did not invent that. Quoting you from some 2 years ago, > rpm-4.[12] days (which is less than "always", I presume), both reports > CLOSED DEFERRED: > > https://bugzilla.redhat.com/89740 > - "Yup, rpm does not do erasure ordering." > - "Not implemented means exactly that." > > https://bugzilla.redhat.com/89500 > - "Ah yes, erase ordering has never been implemented in rpm." > >From context, one is supposed to infer "correctly.", as in rpm does not do erasure order sufficiently correctly. No matter what, there is an implementation, erasure ordering is the reverse of install ordering. That turns out not to be sufficiently correct. The final answer will be to incorporate the ordering from the smart package manager into rpm. Gustavo was wise enough to realize that a dependency graph should label nodes as "package-operation" rather than "package", thereby permitting explicit tsort relations like Order "package-install" before "package-erase". There are other ramifications too. Meanwhile, as we all know, no one has ever bothered to set up the necessary QA to insure that erasures, indeed, are well tracked by dependencies, mainly because install/upgrade, not erasure, is what is mainly of interest. Meanwhile, now thanks to Tim Waugh's careful maintenance of bash over the years, the syntactical sugar of Requires(post): et al can be automated, removing Yet Another error prone process adding dependencies to *.spec files. 73 de Jeff -- Jeff Johnson ARS N3NPQ jbj@xxxxxxxxxx (jbj@xxxxxxx) Chapel Hill, NC
