Re: Reconsider defaults for use-package-vc-prefer-newest

[Suhail Singh]

Philip Kaludercic <philipk@posteo.net> writes:

>> Would it be possible to describe a concrete example and state
>> concretely what sort of problem this would aim to detect?
>
> Suhail gave the example of julia-mode.

Specifically, for the 0.4 version of julia-mode the release on NonGNU
ELPA differed from the 0.4 release that was served on MELPA Stable.

>> How do those problems arise in practice?
>
> Because of differences in how releases are marked.  ELPA checks if a
> commit touches the "Version" header of an Elisp file, while MELPA
> requires a Git tag to annotate the commit used to create the tarball.

And in the case of julia-mode, for the 0.4 release, the commit that was
tagged was (meaningfully) distinct from the commit that updated the
"Version" header.

Users who had enabled the melpa-stable archive and installed julia-mode
version 0.4 experienced differing behavior based on whether they
installed it from melpa-stable or nongnu, due to the mistake by the
maintainer.  Specifically, the release on nongnu was buggy since it
preceded the commit where the regression was fixed.

-- 
Suhail