On Thu, 9 May 2024 at 15:55, Jonathan Perkin <jperkin%pkgsrc.org@localhost> wrote:
* On 2024-05-09 at 15:43 BST, Benny Siegert wrote:
>Am 09.05.24 um 11:49 schrieb Jonathan Perkin:
>>* On 2024-05-09 at 10:40 BST, Benny Siegert wrote:
>>
>>># pkgin show-full-deps pkgin
>>>full dependency tree for pkgin-23.8.1nb3
>>> p5-File-BaseDir>=0.03
>>> p5-MIME-Charset>=1.006.2
>>> php56-pear-[0-9]*
>>> php>=5.6.40nb9<5.7
>>> php81-ldap>=5.2.12
>>> php81-pear-[0-9]*
>>
>>Is that after a "pkgin up"? If so please make a backup of
>>/var/db/pkgin/pkgin.db, then try a "pkgin -f up", and send me the
>>db.
>
>It turns out the last "pkgin up" was before the pkgin upgrade. I had
>done one more after upgrading but it did nothing, since the index file
>on netbsd.org hadn't changed. Running "pkgin -f up" did the trick, and
>now the upgrade is down to 68 packages, which is correct.
Hmm, there's still a bug there as pkgin should have detected that the
database schema was out of date and performed an update itself,
regardless of whether the pkg_summary had changed or not.
I'll need to see if I can reproduce that and add a test case.
Is the pkgin version recorded anywhere in /var/db/pkgin? If so, maybe
treat a change in version as a trigger to reprocess the data by
default? A milder variant of this would be to allow for known
"problematic" older versions and force the update if the old database
was last written by one of those. Of course that can still miss some
cases, but could catch most