Bug#861333: API transition of R packages
On Thu, 27 Apr 2017, Dirk Eddelbuettel wrote: > I feel like we have had the substitution of R (>= 'currentBuildVersion') for > a decade. I didn't realize that it was the current build version; I just assumed it was updated manually. -- Don Armstrong https://www.donarmstrong.com Those who begin coercive elimination of dissent soon find themselves exterminating dissenters. Compulsory unification of opinion achieves only the unanimity of the graveyard. -- Justice Roberts in 319 U.S. 624 (1943)
Bug#861333: API transition of R packages
On 27 April 2017 at 11:37, Don Armstrong wrote: | On Thu, 27 Apr 2017, Dirk Eddelbuettel wrote: | > I don't know, and I tend not to run dated r-base-core packages. | | I'll try to check this out later. Thanks! | > Is that what debian/control ensures? | | Cool; I didn't check to see whether the substitution variable had been | updated. I feel like we have had the substitution of R (>= 'currentBuildVersion') for a decade. | > I uploaded one beta build to experimental. Approximately nobody uses | > those. | | Yeah, that's always a problem. | | > The 'blocking' mechanism really works. R 3.4.0 will not seep into | > testing. | | The problem isn't that R won't enter testing, but that any package which | builds against R has to be rebuilt using the R in testing and uploaded | to testing-proposed-updates. Yes, I hear you on that one. | That's a pretty painful thing to have to do. [Luckily, R is leaf enough | that there aren't too many RC bugs in R packages, so we should be OK.] Fingers crossed :) Dirk -- http://dirk.eddelbuettel.com | @eddelbuettel | e...@debian.org
Bug#861333: API transition of R packages
On Thu, 27 Apr 2017, Dirk Eddelbuettel wrote: > I don't know, and I tend not to run dated r-base-core packages. I'll try to check this out later. > Is that what debian/control ensures? Cool; I didn't check to see whether the substitution variable had been updated. > I uploaded one beta build to experimental. Approximately nobody uses > those. Yeah, that's always a problem. > The 'blocking' mechanism really works. R 3.4.0 will not seep into > testing. The problem isn't that R won't enter testing, but that any package which builds against R has to be rebuilt using the R in testing and uploaded to testing-proposed-updates. That's a pretty painful thing to have to do. [Luckily, R is leaf enough that there aren't too many RC bugs in R packages, so we should be OK.] -- Don Armstrong https://www.donarmstrong.com Mozart tells us what it's like to be human, Beethoven tells us what it's like to be Beethoven, and Bach tells us what it's like to be the universe. -- Douglas Adams
Bug#861333: API transition of R packages
On 27 April 2017 at 12:49, Don Armstrong wrote: | Control: severity -1 serious | | Do we know if this issue may also mean that any packages built with this | new version are incompatible with older R versions? [I'm thinking so, | but my ABI-fu is not super strong.] I don't know, and I tend not to run dated r-base-core packages. | If so, we'll need to make sure that they depend on at least this R | version. Is that what debian/control ensures? Ie from one of my most recent uploads: edd@max:~$ dpkg -f /var/cache/pbuilder/result/r-cran-foreign_0.8.68-1_amd64.deb | grep Depends Depends: libc6 (>= 2.14), r-base-core (>= 3.4.0-1), r-api-3 edd@max:~$ Pretty much ensure you cannot use this with R 3.3.* or older. | We also may need to populate a breaks with all of those packages which | have the older version. | | As a side note, it's really important not to start transitions like this | when we're in a freeze; this upload of R 3.4 should have been made to | experimental, not unstable. [Not that I can really point too many | fingers; I accidentally uploaded a new release of scowl to unstable | which I meant to target at experimental the other day...] I uploaded one beta build to experimental. Approximately nobody uses those. The 'blocking' mechanism really works. R 3.4.0 will not seep into testing. Dirk -- http://dirk.eddelbuettel.com | @eddelbuettel | e...@debian.org
Bug#861333: API transition of R packages
Control: severity -1 serious Do we know if this issue may also mean that any packages built with this new version are incompatible with older R versions? [I'm thinking so, but my ABI-fu is not super strong.] If so, we'll need to make sure that they depend on at least this R version. We also may need to populate a breaks with all of those packages which have the older version. As a side note, it's really important not to start transitions like this when we're in a freeze; this upload of R 3.4 should have been made to experimental, not unstable. [Not that I can really point too many fingers; I accidentally uploaded a new release of scowl to unstable which I meant to target at experimental the other day...] -- Don Armstrong https://www.donarmstrong.com a friend will help you move a best friend will help you move bodies but if you have to move your best friend's body you're on your own -- a softer world #242 http://www.asofterworld.com/index.php?id=242
