On 29 May 2020 at 07:51, Dylan Aïssi wrote: | Hi, | | Le jeu. 28 mai 2020 à 18:58, Dirk Eddelbuettel <[email protected]> a écrit : | > | > Thanks everybody for the help with the transition: 4.0.0-3 is now in testing. | > | | \o/ | | Both transition trackers (r-api-4.0 and r-api-bioc-3.11) were not very | useful to determine the order to update Bioconductor packages. | Some Bioconductor packages were green in the first tracker but red in | the second one, because they were automatically rebuild without an | upgrade. | So, it was not possible to use the first tracker to follow the upgrade | order of Bioconductor packages. | And the second tracker did not consider the r-api-4.0 rebuild order, | so some packages in the first levels were not buildable until the | dependency chain was ready for r-api-4.0. | | Next time there is a transition with these two r-api virtual packages, | we should use a unique ben file for them, something like this: | | title = "r-api"; | is_affected = .depends ~ "r-api-3.5" | .depends ~ "r-api-4.0" | | .depends ~ "r-api-bioc-3.10" | .depends ~ "r-api-bioc-3.11"; | is_good = .depends ~ "r-api-4.0" | .depends ~ "r-api-bioc-3.11"; | is_bad = .depends ~ "r-api-3.5" | .depends ~ "r-api-bioc-3.10";
Good point. Probably worth trying if the next transition once again has BioC within a week (which is common). Nobody knows what will be in R 4.1.0 next year. With some luck we may get by without a transition (a la R 3.6.0). Cheers, Dirk -- http://dirk.eddelbuettel.com | @eddelbuettel | [email protected]
