Honnix
Honnix
> I don't think removing the type altogether is the right solution. Failing early with a definite traceback is better, in my opinion. It's not that early I guess but...
I tried early validation in #2040 but I honestly feel it is overshooting.
If I understand it correctly, this command line option is more or less transparently passed down to pip internal, so it doesn't help much applying an additional layer of validation.
@chrysle Haha, I never tried extending click type, so it was a fun exercise. > Until then, I'd like to hear other opinions. Yeah that makes sense.
> Honestly, this strikes me as dangerous and it removes the immutability/reproducibility property of the constraint files. This makes it possible to influence the resolved deptree by modifying an externally...
> I think that's exactly the reason for doing it by default. Not sure I understand this. What do you mean as "by default"? I think the original PR adding...
> I'm not sure pip-tools should support everything that pip does the same way. It's purpose is to aid reproducibility and constraints that are mutable at random points in time...
In terms of reproducibility, I think it boils down to how the remote constraints file is served and digested. A local file does not ensure reproducibility either.
I'm not exactly sure why `CI / Linters / linkcheck-docs (pull_request)` failed but it doesn't seem to be related to the change.
Please let me know how I can move this forward together with #2040. Thank you.