Dne 03. 11. 20 v 17:55 Miro Hrončok napsal(a):
On 11/3/20 5:10 PM, clime wrote:
> On Tue, 3 Nov 2020 at 15:40, Vít Ondruch <vondruch(a)redhat.com> wrote:
>>
>>
>> Dne 03. 11. 20 v 11:54 Petr Pisar napsal(a):
>>
>> On Mon, Nov 02, 2020 at 12:03:12PM +0100, Miro Hrončok wrote:
>>
>> git (amahdal, besser82, chrisw, pcahyna, pstodulk, skisela, tmz)
>> git.src requires cvs
>> git-cvs.noarch requires cvs
>>
>> <everything else> ( <everybody else>)
>> <package> requires git
>>
>> Note that if xinetd indeed goes away, your package will most likely
>> not be
>> affected, unless you actually need git-cvs.
>>
>> Nontheless the packagers should review their packages whether they
>> indeed have
>> to depend on the full-blown git which pulls in Perl and Python. Very
>> probably
>> they could switch to a tinier git-core:
>>
>>
>> Or switch to depend on `%{_bindir}/git`?
>
> If we do it like this, we will never be able drop repo download times
> for Fedora users.
>
> I personally think packages should depend on package names only even
> though filelists.xml lazy loading is not yet implemented
>
(
https://lists.fedoraproject.org/archives/list/devel@lists.fedoraproject.o...).
>
Actually, several locations end up in primary metadata.
https://docs.fedoraproject.org/en-US/packaging-guidelines/#_file_and_dire...
And as a side note, it is not recommended to use %{_bindir} in
buildrequires. The package actually needs /usr/bin/git even if you
redefine %{_bindir} (e.g. in a flatpak build). However I cannot find
the discussion about this.
I could find at least discussion in one of package review I did
recently, but I still disagree. Even for flatpak, it would be better to
use some prefix such as %{root_bindir} if it was supported.
Anyway, this is OT here. But if somebody decides to propose this into
guidelines, I'd like to be part of the discussion.
Vít
[1]