Bug#961709: lintian: Warn if R binary packages don't depend on virtual r-api-* package

2020-05-28 Thread Dylan Aïssi
Package: lintian
Version: 2.77.1
Severity: wishlist
X-Debbugs-CC: debia...@lists.debian.org

Hi,

I just saw a R binary package (r-cran-isospec) with wrong dependencies
(bug not yet opened). Lintian doesn't warn about a problem in its
dependencies, so it would be cool to add a new warning (maybe
"missing-dependency-on-r-api") to warn if any R binary packages (
r-{cran|bioc|other}-* ) don't depend at least to a virtual r-api-*
package.

Thanks

Best,
Dylan



Bug#961709: lintian: Warn if R binary packages don't depend on virtual r-api-* package

2020-05-28 Thread Dylan Aïssi
Hi,

Le jeu. 28 mai 2020 à 10:15, Dylan Aïssi  a écrit :
>
> I just saw a R binary package (r-cran-isospec) with wrong dependencies
>

Some days ago, I found another package with similar bug (r-bioc-mofa).
Already fixed in unstable but the version in testing has wrong
dependencies. This tag will help to identify these packages. With the
recent r-api-4.0 transition, it would be quite bad to leave behind
these packages.

Best,
Dylan



Bug#961709: lintian: Warn if R binary packages don't depend on virtual r-api-* package

2020-05-28 Thread Andreas Tille
Hi Dylan,

On Thu, May 28, 2020 at 10:10:54AM +0200, Dylan Aïssi wrote:
> Package: lintian
> Version: 2.77.1
> Severity: wishlist
> X-Debbugs-CC: debia...@lists.debian.org
> 
> Hi,
> 
> I just saw a R binary package (r-cran-isospec) with wrong dependencies
> (bug not yet opened). Lintian doesn't warn about a problem in its
> dependencies, so it would be cool to add a new warning (maybe
> "missing-dependency-on-r-api") to warn if any R binary packages (
> r-{cran|bioc|other}-* ) don't depend at least to a virtual r-api-*
> package.

Please go for it.

Regarding the actual package:  I'd strongly suggest to move it to
R pkg team as we did with all those packages (and when doing so also
change the source package name from isospec to r-cran-isospec).

Kind regards

Andreas. 

-- 
http://fam-tille.de



Bug#961709: lintian: Warn if R binary packages don't depend on virtual r-api-* package

2020-05-28 Thread Dirk Eddelbuettel


On 28 May 2020 at 10:10, Dylan Aïssi wrote:
| Package: lintian
| Version: 2.77.1
| Severity: wishlist
| X-Debbugs-CC: debia...@lists.debian.org
| 
| Hi,
| 
| I just saw a R binary package (r-cran-isospec) with wrong dependencies
| (bug not yet opened). Lintian doesn't warn about a problem in its
| dependencies, so it would be cool to add a new warning (maybe
| "missing-dependency-on-r-api") to warn if any R binary packages (
| r-{cran|bioc|other}-* ) don't depend at least to a virtual r-api-*
| package.

Not a bad idea.

OTOH the way we implement the tag doesn't it get added automagically by the
r-base-core package when constructing an r-{cran,bioc,...}-* package?

Dirk

-- 
http://dirk.eddelbuettel.com | @eddelbuettel | e...@debian.org



Bug#961709: lintian: Warn if R binary packages don't depend on virtual r-api-* package

2020-05-28 Thread Dylan Aïssi
Hi Dirk,

Le jeu. 28 mai 2020 à 18:03, Dirk Eddelbuettel  a écrit :
>
> OTOH the way we implement the tag doesn't it get added automagically by the
> r-base-core package when constructing an r-{cran,bioc,...}-* package?
>

Yes, it is.

But, in one week I found by chance two packages (r-cran-isospec and
r-bioc-mofa) with similar problem.These packages are not simple, the
Debian source package generates more than one binary package. In
addition to the binary R package, isospec generates libraries packages
and a python package and r-bioc-mofa generates a python package. They
have a non standard d/rules file that induce a bug.

I would say if I have found this bug twice in one week, it will happen
again :-(.

Best,
Dylan



Bug#961709: lintian: Warn if R binary packages don't depend on virtual r-api-* package

2020-05-29 Thread Dirk Eddelbuettel


Dylan,

On 29 May 2020 at 08:03, Dylan Aïssi wrote:
| Hi Dirk,
| 
| Le jeu. 28 mai 2020 à 18:03, Dirk Eddelbuettel  a écrit :
| >
| > OTOH the way we implement the tag doesn't it get added automagically by the
| > r-base-core package when constructing an r-{cran,bioc,...}-* package?
| >
| 
| Yes, it is.
| 
| But, in one week I found by chance two packages (r-cran-isospec and
| r-bioc-mofa) with similar problem.These packages are not simple, the
| Debian source package generates more than one binary package. In
| addition to the binary R package, isospec generates libraries packages
| and a python package and r-bioc-mofa generates a python package. They
| have a non standard d/rules file that induce a bug.

Fair point. Multi-language repos / project exists (was just talking to
upstream with an existing C++/Python layer about adding an R package). Some
(most?) still use the standard default but there can always be extra tests.
 
| I would say if I have found this bug twice in one week, it will happen
| again :-(.

Sure. And it is a fairly cost (for a new piece of lintian checking code) with
a large potential benefit.

Dirk

-- 
http://dirk.eddelbuettel.com | @eddelbuettel | e...@debian.org