%pyproject_patch_dependency behavior wrt. multiple packages in one spec

[Miro Hrončok]

Hello,

So I've used %pyproject_patch_dependency in a specfile that packages two Python 
distribution packages.

It filtered the dependencies from both. Is this expected? Should it gain some 
finer tuning (--dist-name argument perhaps)?

[Maxwell G]

On 8/24/26 6:59 AM, Miro Hrončok via python-devel wrote:
> So I've used %pyproject_patch_dependency in a specfile that packages two Python
> distribution packages.
>
> It filtered the dependencies from both. Is this expected? Should it gain some
> finer tuning (--dist-name argument perhaps)?

I think providing options to filter it to only one dist makes sense, but 
the bigger question is what the default behavior in the case of multiple 
dists in a single package should be --- and I don't have a good answer 
for that yet :).

[Miro Hrončok]

On 28. 08. 26 18:00, Maxwell G via python-devel wrote:
> On 8/24/26 6:59 AM, Miro Hrončok via python-devel wrote:
> > So I've used %pyproject_patch_dependency in a specfile that packages two Python
> > distribution packages.
> >
> > It filtered the dependencies from both. Is this expected? Should it gain some
> > finer tuning (--dist-name argument perhaps)?
>
> I think providing options to filter it to only one dist makes sense, but the 
> bigger question is what the default behavior in the case of multiple dists in a 
> single package should be --- and I don't have a good answer for that yet :).

It could either fail or filter it from everything (currently it does that).

I think the current behavior makes sense (e.g. for filtering pytest-cov etc.) 
if documented properly.

[Maxwell G]

On 8/28/26 4:06 PM, Miro Hrončok via python-devel wrote:
> It could either fail or filter it from everything (currently it does that).
>
> I think the current behavior makes sense (e.g. for filtering pytest-cov etc.) 
> if documented properly.

It would technically be a behavioral divergence from the other macros if 
the macro doesn't error out without explicit --dist-name when there are 
multiple dists present, but I think this is fine as long as it's 
documented. The current behavior is sensible in most cases (like 
removing coverage and lint deps, as you mentioned), and I don't think we 
would easily be able to make %pyproject_patch_dependency directly error 
out in the presence of multiple dists, since it's used in %prep.