Repository navigation
Conversation
For users who want to skip the `python-fastmcp-slim` dependency, which is optional for `setools` but required for some of the tests. Just skip the tests that require `python-fastmcp-slim`.
|
Hello, Thanks for your report and Pull Request. Actually I modified the package I therefore consider this PR as fixed, and these is no need to introduce |
|
This is great! Thanks a lot. |
For me, the new
python-fastmcp-slimdependency is a bummer.PKGBUILDs carefully (and indeed they should), adding a dozen more unfamiliarPKGBUILDs to review increases the probability of making a mistake.python-fastmcp-slimpackage itself is new, added to the AUR just this last June 9. The package was last updated on June 11, has since been flagged out-of-date, and has a pending request. So it is far from clear how well maintained this package will continue to be. I don't like this especially because I cannot find any solid documentation about what dependencies are actually required for FastMCP, so I can't verify thatpython-fastmcp-slim'sPKGBUILDis even listing the correct dependencies. (I suppose if I tried to install FastMCP withuvI could verify this – but I didn't bother because all of the other concerns are sufficient for me).python-fastmcp-slim, you still have to be in charge of building it and a dozen other AUR dependencies forever and ever, just to run the tests.This issue comes from how upstream chose to structure the tests, but I think it can be worked around pretty easily – for those who would prefer it – at the packaging level. All I did was modify the
setoolsPKGBUILDto not includepython-fastmcp-slimas a dependency, and to tellpytestto skip the two tests that require FastMCP. (I also think this falls squarely within the package-maintenance scope, since we're just making a decision about how to handle an optional dependency.) The question, to which I don't have a definitive answer, is how to make this sort of workaround available to people who want it, while also leaving open the option to build with the optionalpython-fastmcp-slimdependency.What I tentatively propose in this PR is to make a separate
setools-no-mcppackage. It is the same as thesetoolspackage, except thatpython-fastmcp-slimisn't a dependency and the tests that require FastMCP are skipped. The package provides and conflicts withsetools, sopacman -S --asdeps setools-no-mcpwill replacesetoolswithsetools-no-mcpwithout complaint.That's the cleanest proposal I can think of for now – it will at least be my set-up for the time being, but I propose it here as well because there might be other people in my position. Sorry again for the flurry of PRs, and thanks again for your work!