Re: Partial version numbers in master

[Jean Abou Samra]

Le 25/03/2022 à 13:05, Michael Käppler a écrit :
> Am 22.03.2022 um 12:57 schrieb Jean Abou Samra:
> > Hi,
> >
> > With https://gitlab.com/lilypond/lilypond/-/merge_requests/1187,
> > I introduced a change that I didn't notice: something like
> >
> > \version "2.23"
> >
> > is now an error. It is accepted with current released versions.
> > On the other hand, convert-ly has always rejected such version
> > strings. Should we accept them or not, in LilyPond and in convert-ly?
> I see no benefit in accepting them. We do not have a version "2.23" and
> there would arise ambiguities. Does "2.23" mean:
> -works with 2.23.0 (as Dan suggested) or
> - works with every 2.23.x



I don't see one either, but there appear to be scores in the wild that
use them (I saw an example on the user list, which is what triggered
the original post). The question is whether we consider it fine to
start giving an error on them (it's not a fatal error, doesn't prevent
compilation of the score). I would agree that it's not good practice
for development releases. For stable releases though, "works with
2.22.0" is supposed to imply "works with every 2.22.x", so I wonder
if there is a case for accepting it.

If we consider this change good, it should be advertised in the
Changes document. If not, it should be undone, and preferably
we'd be consistent and convert-ly would start accepting such
partial version numbers. I'm still on the fence, but given
the imminent release, I need to have put up a merge request
very soon, so I wanted to see if there were opinions ...