On 28 October 2011 at 22:50, Julian Taylor wrote:
| On 10/28/2011 10:33 PM, Dirk Eddelbuettel wrote:
| >
| > The "Depends: r-base-core (>= minVersion), r-base-core (<< nextVersion) had
| > been tried in the past and rejected a few years ago:
| >
| >
| > rpy (1.0.0-1) unstable; urgency=low
| >
| > * New upstream release announced today
| >
| > * debian/control: Update Depends: on r-base-core to (>= 2.6.0) but
remove
| > the (<< 2.7.0) so that rpy should not prevent R from entering
testing
| >
| > -- Dirk Eddelbuettel <e...@debian.org> Mon, 19 Nov 2007 16:51:59 -0600
| >
| >
| > What else can we do?
| >
| > Dirk
| >
| >
| breaking rpy is probably a valid reason to deny R to migrate to testing
| or for rpy to be removed from testing.
| But I think you will have to get in contact with the release team to
| hint the packages into testing together when ready each time.
I think it was this need for manual intervention that made me reject the
formal Breaks: / Depends: (<< ...) model.
| How does rpy2 handle this?
It uses a different scheme in setup.py and does not encode the version number
of the R version in the .so loaded by Python -- which causes the breakage you
see.
rpy2 is simply smarter, as is my own 'littler' (cmdline frontend to R,
similar build issue, survives R upgrades just fine). They just have an
unversioned .so and that just works. [ In the ~ 10 years I have maintained R
it has had an ABI/API change just once. ]
| can't you just relax the version requirements in rpy and match up
| migrations with when rpy2 needs them?
Yes, that is what I have done. Look at the changelog -- I usually builds once
every six months when changes,
NOW: R will switch to a annual cycle with the next release so it will be less
of an issue.
In sum: I lean against encoding this as a "hard" Depends line. I currently
have
Breaks: r-base-core (>= 2.15.0)
and
* debian/control: Added 'Breaks: r-base-core (>= 2.15.0)'
(Closes: #646969)
but may remove this. Unless you should real loud real soon :)
| Or does the coupling go beyond api/abi compatibility?
Not at all. It is a (bad, outdated) setup.py issue local to rpy.
Dirk
|
| xapplication/pgp-signature [Click mouse-2 to save to a file]
--
"Outside of a dog, a book is a man's best friend. Inside of a dog, it is too
dark to read." -- Groucho Marx
--
To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
