Re: Is the dependency rule distribution-wise?
Santiago Vila wrote:
> > > If the dependency rule is to be considered "distribution-wise", ncurses3.4
> > > may not be made optional in a certain release until all the packages of
> > > standard or higher priorities have been recompiled in the same release.
> >
> > Which means that when uploading ncurses 3.4 the maintainer made a mistake
> > and used the wrong priority. [...]
>
> No. The priorities in the source control file are not "authoritative".
> The only priority that really counts is the one in the override file.
You're right, but the overrides file and the control file should be in
sync, otherwise it won't make sense to have the - not used anyway - priority
listed in the control file. Thus I assume that the ftp masters notice
when priorities mismatch and can react.
> If ncurses3.4 source has "optional" in the control file, is because it was
> expected that most packages would be recompiled with libncurses4 before
> the release.
>
> If this has not been the case, you can't blame the uploader (btw: me, in
> this case) for not "guessing the future".
Oh well, sure I can, we should have thought about this before the package
was re-uploaded and we should have ironed out that the priority of this
package will be changed even after the freeze to comply with policy.
> > Unfortunately slink is already frozen and
> > this bug is not important enough to change it.
>
> Since when we fix only important bugs during the freeze?
Oh, we fix a lot of bugs, but only important bugfixes make it into
the distribution.
> As the freeze is more solid, we of course restrict uploads because they
> require compiling, this has risks, but changing a single line in the
> override file does not have any risks.
I won't say that but the risk to cause damage by this action is very very
little.
Regards,
Joey
--
A mathematician is a machine for converting coffee into theorems.
Please always Cc to me when replying to me on the lists.
Reply to: