Bug#946046: not fixed
On 2020-08-07 23:57:02 +0200, Stéphane Glondu wrote:
> The unison in Debian 10 is compiled with OCaml 4.05.0, which is long
> gone. The only way to synchronize with a buster machine is to use unison
> from buster.
This is what I've done until now, with the unison package put on hold.
But when I wanted to upgrade, based on the the new packaging structure
(which claimed to fix the bug), this version got upgraded too!!!
> I'm sorry I haven't been sufficiently clear on that.
>
> Since the version of OCaml changes at each Debian release, there will
> not be /in the archive/ a package (or metapackage) that will be
> compatible with two successive Debian releases (or with stable and
> testing/unstable). My goal with the new packaging structure is that once
> a version of Unison is installed, it may be kept installed even after an
> upgrade of Unison (meta)packages. So, for example, if we transition to a
> new version of OCaml (say 4.11.0), I will ask for the removal of
> unison-2.48 from the archive and upload a new unison-2.48+4.11.0. In
> this hypothetic future, unison-2.48 can be kept installed on systems (or
> installed from snapshots) where its compatibility set is needed.
> Moreover, unison-2.48 and unison-2.48+4.11.0 would be co-installable,
> easing migration from one version to the other.
unison-2.48 should have never been created: buster has a unison
package with version number 2.48.4-1, thus one has the impression
that this is compatible with unison-2.48, while this is not the
case. The package should have been named unison-2.48+4.08.1 right
now to avoid this confusion.
Moreover, unison-2.48 provides /usr/bin/unison-2.48; this is not fine
since buster also has /usr/bin/unison-2.48, which is not compatible.
As the client will start a server with the same version number on
the remote machine (by using the recommended "addversionno = true"),
this means that one will likely get an archive corruption (which
happened to me). The OCaml version must be part of the version and
attached to the name of the unison executable.
--
Vincent Lefèvre <vincent@vinc17.net> - Web: <https://www.vinc17.net/>
100% accessible validated (X)HTML - Blog: <https://www.vinc17.net/blog/>
Work: CR INRIA - computer arithmetic / AriC project (LIP, ENS-Lyon)
Reply to: