Re: comparable PYTHON_VERSION

[Thomas Klausner <wiz%NetBSD.org@localhost>]

On Mon, Mar 27, 2023 at 06:45:43PM -0400, Greg Troxel wrote:
> Thomas Klausner <wiz%NetBSD.org@localhost> writes:
> 
> > The attached patch introduces a PYTHON_VERSION that solves all of
> > these. It's set to "major_version + 100 * minor_version".
> 
> I think you mean
> 
> > these. It's set to "major_version * 100 + minor_version".
> 
> but ok.

Yes, sorry, that's what I meant.

> > a) it's 0 if there is no overlap (but I'd like to hear arguments if
> > 999 would be a better value)
> 
> 0 feels more like no python that 9.99.  But, I think the real reason
> there is even 0 is to just avoid "PYTHON_VERSION > 309" being an error,
> and that if PYTHON_VERSION is actually 0, then we are going to get the
> "no valid versions" error, and the comparison does not actually matter.
> If that's accurate, I would like to see a comment before setting
> PYTHON_VERSION that says that while comparsion results are UB, having no
> valid versions leads to a failure and it's just to avoid a type error
> masking the intended failure.  But said more coherently!

I tried to add a comment in that direction and committed it, thanks!
 Thomas