On 12 January 2016 at 05:21, Dirk Eddelbuettel wrote: | | On 12 January 2016 at 09:50, Sébastien Villemot wrote: | | Le dimanche 10 janvier 2016 à 22:03 -0600, Dirk Eddelbuettel a écrit : | | > Are you going to report the bug to gcc upstream? | | | | Yes I'll take care of the gcc bug report. | | +1 | | | Concerning the fix that we uploaded, it *almost* works. R compiled on | | mips and the R executables now runs correctly, which is a big | | improvement. | | | | However, one gets: | | | | (sid_mips-dchroot)sebastien@minkus:~$ R CMD config CC | | gcc-4.9 -std=gnu99 | | | | This is causing r-cran-rjags to FTBFS, and I guess many others as well, | | since gcc-4.9 is not in the build deps of those packages. | | | | Do you know how to fix this issue? I think one possibility is to simply | | add "gcc-4.9 [mips]" to the dependencies of r-base-dev. Another one | | I like that. I may have used that in the past too.
It is turning into a bit of a nightmare because r-base-dev is arch: all and therefore can't have arch-dep depends. Making it any turned on all types of side effects. Three builds in and nothing to show. I think I will put the 'Depends: gcc-4.9 [mips]' into r-base-core itself. Anything I am overlooking ? Dirk | There are one or two that parse the output and do weird things == I have | CC=ccache gcc and CXX=ccache g++ in many of my ~/.R/Makevars. But not a | FTBFS. | | | would be to have "R CMD config CC" return "gcc -std=gnu99" on mips | | (while still compiling R with gcc-4.9), but I don't know how to achieve | | that. | | | | What's your opinion? | | I'll try the former. I also got a lot of 'timestamped gz' warnings so I'll | set GZIPCMD='gzip -n' and see how that goes. | | Dirk | | -- | http://dirk.eddelbuettel.com | @eddelbuettel | e...@debian.org -- http://dirk.eddelbuettel.com | @eddelbuettel | e...@debian.org
