Hi Dirk, On 22.04.2015 at 20:25, Dirk Eddelbuettel wrote: > [..] > Certainly, but can we please backtrack for a second? > > R has been in Debian since the laste 1990s. We never needed a TZ (nor should > having one break it, but let's leave that aside for a second), and given all > the releases and architectures we must by now have built a thousand > times. Without FTBFS.
Sure, I don't question that. And I think it's important to stress that R does _not_ FTBS on the buildds. I'm aware of that. > So who now sets TZ and why? Can we just unset before / while R is built? The reproducible team does to check for software quality and to ensure builds are reproducible in Debian. > R should work just fine with or without TZ. Most systems seem to have it; > Debian never did and I am not aware of side effects (but we sometimes do set > TZ in our (user) analysis scripts). Indeed, it does, see: $ TZ="GMT+1" Rscript -e "system('date')" Mi 22. Apr 19:57:33 GMT 2015 But R compiled with TZ set fails to install 'sysdata.rda' while building it's own package. And this should not happen. > So two things going forward: > i) find out who set TZ, make him stop it (should take care of the FTBFS now) The evil reproducible team does ;) I think the impact of this FTBFS is more theoretical, so we should not get nervous. > ii) contact upstream which I'll will and see what can be done about makeing > the behaviour more robust (as it used to be prior to R 3.1.3 as well, I > think) Yes, contacting upstream would be great, thanks. But I think prior to R 3.1.3 also FTBFS with TZ set, see: https://reproducible.debian.net/rbuild/testing/amd64/r-base_3.1.2-3.0~reproducible1.rbuild.log Best, Philip
signature.asc
Description: OpenPGP digital signature