On Fri, Sep 19, 2008 at 07:49:37PM +0400, Dmitry Kurochkin wrote: > Hi! > > My second attempt on resolving issue1039. This time we check for > common patchinfos instead of introducing repo-id. > > Besides --ignore-unrelated-repos flag is introduced to skip the check. > And test is there as well. > > To check for common patches results from get_common_and_uncommon > function are used. So there should be no performance hit. But there is > a problem with pull command for multiple repositories. In read_repos > we do union, intersection or complement. So after read_repos we can > not check for common patches in each repo. To solve this we need to do > get_common_and_uncommon for each repo in read_repos. I think this can > introduce performance regressions, so check for unrelated repo is done > for union of all pulled repos, their intersection, or just the first > repo (depending on parameters). This should be enough for most cases.
Applied, with a change to make small repositories always pass the test. This makes the test suite pass, and doesn't really hurt, since tiny repositories are generally easy to merge. Also, the refactor in pull_cmd was a minor regression, since it would break the type witnesses if they worked on that file, so I reverted it. -- David Roundy http://www.darcs.net _______________________________________________ darcs-users mailing list [email protected] http://lists.osuosl.org/mailman/listinfo/darcs-users
