On 13 May 2010 at 10:01, Don Armstrong wrote: | On Thu, 13 May 2010, Dirk Eddelbuettel wrote: | > On 12 May 2010 at 22:40, Don Armstrong wrote: | > | However, the whole idea of having source package which do not | > | produce at least a binary of the same name is fundamentally | > | broken, and I have spoken repeatedly against doing it. | > | > Well narrowly spoken I then still do the right thing as my 'source | > pkg: sm' with 'binary pkg: r-cran-sm' does provide the 'command' | > | > library(sm) | > | > inside R. | | Right, the logic behind choosing sm as a source package name for | r-cran-sm is clear. My main problem with it is just the ramifications | of doing that in Debian. | | If I was paying more attention when R packages started coming into the | archive, I would have lobbied harder for them to follow the lib*-perl
If memory serves, I started with RODBC and tseries. Thanks to the corresponding debian/changelog entries, that can be pinned to March 2003. Seven years ago. AFAICR we had no consistent Perl Policy (TM) at the time. On 30 Dec 2003 I posted (to debian-devel and r-devel) a proposed 'Debian R Policy' but never followed in a formal manner. *Informally* we now have this consensus: new r-cran-$FOO packages of CRAN sources $FOO comes with source and binary named r-cran-$FOO. So the question now is whether we need to clean up the several dozen packages that do not correspond. | model, where the source package and the binary package have the same | name... so the binary package r-cran-sm would be built by the | r-cran-sm source package, and if you wanted the R library sm, you'd | just look for r-*-sm. [Since users don't generally see the source | package name, it tends not to matter what it's called.] Yes. That is what we do. We even once had r-bioc-* for a BioC package (that was later withdrawn, the BioC people changed their packaging layoyt) and we have r-other-$AUTHOR-$PACKAGE for things that comes from neither CRAN nor BioC. | > There can't be another 'binary command' for the R language, but | > every R users refers to this as 'sm'. | | Well, there could be, but hopefully people wouldn't at least conflict | with the CRAN repository. | | > But I know what you really imply. I do get the 'letter' but not the | > 'spirit' of this suggestion. | | The spirit here is really just to avoid confusion where you have a | binary and source package which share the same name, but are built | from different source packages. I've made changes to the BTS to handle | this issue, so that the BTS deals with it properly, but (as evidenced | by this bug) it's confusing for the maintainers and users who haven't | been exposed to this problem and its solution. | | For the record, if you mean the source package sm, you need to file | bugs against: | | Source: sm | | or reassign them using | | reassign nnn src:sm Ack. | Source packages will always have 'src:' prefixed, which works ok, | because ';' isn't a valid character for a package name. As a nod to | backwards compatibility, bugs filed against a package 'foo' which | isn't a binary package, but is a source package, will behave as if | they were assigned to 'src:foo'. Otherwise, it's assumed that you mean | the binary package, not the source package. | | At least, that's how it's supposed to work. [If it didn't, or doesn't | in the future, please let me know.] :) Dirk | | | Don Armstrong | | -- | When I was a kid I used to pray every night for a new bicycle. Then I | realized that the Lord doesn't work that way so I stole one and asked | Him to forgive me. | -- Emo Philips. | | http://www.donarmstrong.com http://rzlab.ucr.edu -- Regards, Dirk -- To UNSUBSCRIBE, email to debian-bugs-rc-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
