pkgsrc-Users archive
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]
Re: python313 dependency mismatch?
And ... just to answer the specifics here as well:
>>4 packages to refresh:
>> libuuid-2.40.2 llvm-19.1.7 pkg_install-20250417 readline-8.3nb1
>>
>>4 packages to upgrade:
>> pkgin-25.10.0 python310-3.10.19 python312-3.12.12 sqlite3-3.51.1
>
> I'm assuming you are upgrading from a relatively recent pkgin, and so
> this might not make any difference, but just in case please can you
> upgrade just pkgin first.
Yes, this is an upgrade from a pkg installation which was 2025Q3
as of December 12, 2025.
Yes, I can upgrade pkgin first. Although, as I'm sure you agree,
manually doing that should not be needed. The thing which makes
the above fail is that llvm is included in the packages to be
upgraded / refreshed.
> Newer pkgin will now automatically upgrade itself and pkg_install
> first, just to ensure you are always running the latest prior to any
> upgrade.
Yes, I have seen the message "pkg tools upgraded, please re-do "pkgin
upgrade"" or something to that effect. However, llvm is not among the
packages which support pkgin, pkg_install, readline, sqlite3 or any of
the python packages, so why it's automatically included in this "just
upgrade pkgin and its dependencies" run I do not understand.
> I've not seen this failure, nor had any reports from my users of
> similar issues with my package sets, so I'd really like to understand
> what's going on here and reproduce.
Understood. I'm consistently observing this issue in my installed
base, so obviously there must be something peculiar to "my"
installations which triggers this problem.
> Is there a way to reproduce this using the sets available from
> ftp.netbsd.org, e.g. "install the 2025Q3 set, make sure some package
> using python is installed, then upgrade to the 2025Q4 set"?
I would have thought "yes", but have so far not created a "scratch
system" installed from scratch to reproduce the problem.
However, as you've probably observed, I'm lazy and try to reduce the
amount of "special customization" to a minimum, so I'm slightly
surprised that I appear to be alone in observing this issue.
Also see the message I just sent as a reply to Jarle Greipsland with
sqlite3 table output before and after "pkgin up" on one system which
appears to want to follow the same path to overlooking installing
python313 when refreshing llvm, which evidently happens in the pkgin
run which is perhaps attempting to just update "pkg_install and pkgin"
and associated dependencies.
Regards,
- Håvard
Home |
Main Index |
Thread Index |
Old Index