On Friday, July 9, 2021 11:36:09 AM CEST Miro Hrončok wrote:
On 09. 07. 21 11:30, Pavel Raiskup wrote:
> On Friday, July 9, 2021 10:19:57 AM CEST Miro Hrončok wrote:
>> On 09. 07. 21 9:46, Pavel Raiskup wrote:
>>> This is pre-production experiment, right? Couldn't you just specify
>>> "Provides: python3dist(click) = 7.9"?
>>
>> It is, but it would not help me. By manually providing this, I would:
>>
>> - confuse %pyproject_buildrequires generator -- Python would think 'click
<
>> 8' is not satisfied, by RPM/dnf would insist it is.
>> - still not get a root.log resolution failure I want
>> - still not "satisfy" packages that BR "python3click"
instead of
>> "python3dist(click)"
>>
>> I wonder if adding Obsoletes: python3-click < %{version} helps, but I
don't
>> think so.
>>
>>> We have an RFEhttps://pagure.io/copr/copr/issue/1336 that could help you if
you
>>> could specify dnf.conf excludepkgs= option for the rawhide repo. But this
is
>>> not an easy one.
>>
>> Yes, this would work.
>
> Another idea...
> What if you had a syntetic package, say "click-blocker", with
'Conflicts:
> python3-click < XYZ'? That package could be added as always-installed
package
> into "Packages" field (chroot Edit form).
>
> Or maybe it could simply require 'python3-click >= XYZ', too?
Is "Packages" always-installed or initially-installed? What prevents dnf to
downgrade it?
I'll test this approach thou, seems reasonable.
You are right, it is always "initially installed".. It is definitely
interesting topic, please share the results of the experiment :-)
If DNF really removes the artificial package because it depends on a _newer_
click package (than the following transaction requires) -- we could probably
consider an upstream report..
Pavel