http://code.toofishes.net/gitweb.cgi?p=pacman.git;a=commitdiff;h=4bef7544e11... This stuff still requires fixing, right? I see these functions are now declared in db.h, but we probably need generic functions here. So the names are wrong. Maybe we could just rename them with : s/cache/list or something. Actually, it seems like we mostly need get_pkglist and free_pkglist. And same for grouplist. I don't know why we need to handle add_pkgincache and remove_pkgfromcache manually. These are called just after db_write and db_remove respectively. So the implementation of these two db functions in the cache backend could handle the cache themselves. Hm well, just after writing this, I see there is a problem with the db_update code : /* remove the old dir */ _alpm_log(PM_LOG_DEBUG, "flushing database %s\n", db->path); for(lp = _alpm_db_get_pkgcache(db); lp; lp = lp->next) { pmpkg_t *pkg = lp->data; if(pkg && _alpm_db_remove(db, pkg) == -1) { _alpm_log(PM_LOG_ERROR, _("could not remove database entry %s%s\n"), db->treename, alpm_pkg_get_name(pkg)); RET_ERR(PM_ERR_DB_REMOVE, -1); } } /* Cache needs to be rebuilt */ _alpm_db_free_pkgcache(db); I remember there was the same problem with db_unregister / db_unregister_all. We have to walk through the list and remove its items one by one. I am still not sure how to do that correctly in C. But it reminds me of the java Iterator interface where this is possible: http://java.sun.com/j2se/1.4.2/docs/api/java/util/Iterator.html while(it.hasnext()) { it.next(); it.remove(); } Well, I don't know, but imo it would be better if the rest of libalpm code didn't have to deal with this cache stuff at all.