At 18:19 Uhr -0400 16.04.2002, Kyle Moffett wrote: >On Tuesday, April 16, 2002, at 06:00 PM, Max Horn wrote: > >>At 17:54 Uhr -0400 16.04.2002, Kyle Moffett wrote: >>>On Tuesday, April 16, 2002, at 09:46 AM, Max Horn wrote: >>>>At 6:42 Uhr -0400 16.04.2002, Kyle Moffett wrote: >>>>>I feel kind of stupid now, but maybe something where Fink could >>>>>update itself to the latest cvs would be helpful >>>> >>>>"fink selfupdate-cvs" >>>> >>>>It's already there. All essential packages (including Fink) are >>>>automaticall updated if you run it >>>> >>>>Also, binary distro doesn't involve the Fink package manager at >>>>all, hence such a field as you propose wouldn't do anything with >>>>regards to the bindist. And there is no need for it either, since >>>>the "new binaries" work fine with the "old" package manager. >>> >>>I think you misinterpreted my idea, let me try to rephrase it: >>> >>>Let's say that we need a new tag 'Foo:' for some packages. The >>>tag is made, added to pm cvs and everything is OK. The problems >>>come when somebody using unstable cvs packages tries to install a >>>package using that tag. At the moment fink will fail with some >>>weird error (Depending on the tag and its purpose). >> >>Well, if they do it the right way (using fink selfupdate-cvs), then >>Fink will already be updated! Fink is usually no more than a couple >>of days in unstable only. And in any case, if you mess with >>unstable, you bear the risks. > >Yes, but it is sometimes a week or two when new features are in cvs >but not available as a package. fink selfupdate-cvs only updates >from the /fink/packages tree in cvs. The /fink/fink tree is ignored.
Uhm - yes and that has good reasons. Features of a not yet released Package manager are not to be used in any package in the package tree! They are only to be used for experiemental packages. > >>> I was suggesting that instead, all packages needing features not >>>present in the latest released package manager could be identified >>>by 'Depends: fink-cvs (>= #date#)' or 'NeedsFinkCVS: #date#'. >> >>We do this already, though with BuildDepends. > >Yes, but if it depends on features only available in cvs, the >versioning fails. Some packages might need a newer version of fink, >but if the version isn't available as a tarball, it cannot be >obtained except by the cvs tree /fink/fink. Wrong. This doesn't happen, and if it does, it's by mistake of the package maintainer who then should immediatly revoke the faulty package. > >>> Then Fink could by default ignore packages in CVS (stable or >>>unstable) that have a non-cvs version or a cvs version older than >>>#date#. Then an optional tag could be set in fink.conf for >>>developers and beta testers. This tag would tell Fink to ask the >>>developer if they want to update the fink pkg manager from cvs on >>>the detection of such a package. >> >>Well, why? >> >>a) developers should also use fink selfupdate-cvs > >Yes, but selfupdate-cvs does not upgrade the package manager version >that is still in cvs. True. If you have to use that (because you are developing a package based on new features). It's a three line procedure to update the PM manually. > This is not a very long time, but occasionally I see mails asking >why a package errors out with new tags that should work or some new >feature doesn't seem to exist (Because the developer/beta tester >didn't update the pkg manager from cvs.) If you encounter that *with packages from the official package tree* in the future, please immediatly report these buggy packages so that they can be removef rom CVS ASAP. Max -- ----------------------------------------------- Max Horn Software Developer email: <mailto:[EMAIL PROTECTED]> phone: (+49) 6151-494890 _______________________________________________ Fink-devel mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/fink-devel
