Le 26/05/2019 à 00:04, Lukas Fleischer a écrit :
> On Sat, 25 May 2019 at 13:46:45, Bruno Pagani wrote:
>> Le 25/05/2019 à 19:07, Lukas Fleischer a écrit :
>>> + This is a VCS package. Please do <strong>not</strong> flag this package out-of-date when the package version in the AUR does not match the most recent commit.
>> Maybe we should also hint when it is appropriated to do so? E.g. missing
>> new dependency or old one to be removed, moved upstream…
> There already is a message stating that package flagging should not be
> used to report such bugs. Is it not obvious/precise enough?
Well I think it should instead be used for such bugs in the case of VCS
packages, because those are the only cases where they can be OOD, and in
contrary to normal packages those are valid OOD reasons. And that is the
case for the linked package in the FS ticket. But I acknowledge this is
not what we say currently, though I would use the opportunity of that
addition to change the guidelines regarding this. And then I’m in favour
of saying so in the message:
“This is a VCS package. Please do not flag it out-of-date if the package
version in the AUR does not match the most recent commit. Flagging this
package should only be done if the sources moved or changes in the
PKGBUILD are required because of recent upstream changes.”
>>> +/**
>>> + * Determine whether a package base is (or contains a) VCS package
>>> + *
>>> + * @param int $base_id The ID of the package base
>>> + *
>>> + * @return bool True if the package base is/contains a VCS package
>>> + */
>>> +function pkgbase_is_vcs($base_id) {
>>> + $suffixes = array("-cvs", "-svn", "-git", "-hg", "-bzr", "-darcs");
>> What about false positives and missing items like -nightly’s? I think it
>> would be a good time to implement FS#56602, auto-seed the value
>> depending on your above list and let maintainers override this.
> Yes, there are false positives and false negatives. That is why we only
> display a warning and do not automatically disable the feature for VCS
> packages. Read the comments in FS#62733 for details.
All I’ve read was the same thing as before regarding the impossibility
to correctly detect all VCS packages and just them, but I did not see
why manual override wouldn’t be an option. ;) Regarding false positives,
without override possibility they will be misleading to users, so I
don’t agree on “it’s OK because we are not plainly disabling the
feature”. Also for me the strongest reason to not disable the feature
for VCS packages is rather because it is still useful even for those, as
stated by Eli. :)
Regards,
Bruno