I have discovered an issue with this release when installing/compiling packages with R in that Makeconf in /usr/lib64/R/etc has missing compiler references, specifically c++11. using devtoolset-7 , you get "C++11 standard requested but CXX11 is not defined".
In Makeconf all references with CXX11 are either blank or incorrect. This was not an issue in previous builds? I have manually set these values and it now works.
If it helps i used:
CXX11 = g++ was blank
CXX11FLAGS = -O2 -g $(LTO)
CXX11PICFLAGS = -fPIC
CXX11STD = -std=c++11 was blank
Bodhi is disabling automatic push to stable due to negative karma. The maintainer may push manually if they determine that the issue is not severe.
quantumcheese: We have never built R with the devtoolset toolchain, but rather, the stock EL toolchain, which I do not believe has full support for C++11. This leads to the Makeconf not having CXX11 references. If we build R for EL with devtoolset, it gets tricky, because then R package building will break for any users not using devtoolset.
I think that most users will want C++11 support (which many modern R packages depend on), so I'll just adjust the package to build with it and Requires it.
This update has reached 14 days in testing and can be pushed to stable now if the maintainer wishes
(apologies for delay, and your quick response)
Yes, i appreciate you don't compile R with a devtoolset, however many of the packages we need to install into R here req it, to pick one at random: biocLite("rtracklayer") from the bioconductor suite. I'm used to having to hop in and out of the devtoolset to install various R packages, but this time it wouldn't work until i made the change above. FYI our previous R build was 3.5.2 and this didn't suffer the issue, nor other previous versions. OS: CentOS 6.10
quantumcheese: I totally understand. Right now, the issue is that the epel6 buildroot does not include devtoolset, so I can't build R with it and properly generate Makeconf with CXX11 values. (Yes, I could do what you did and monkey patch them in, but then, I'll start getting build failure reports from random R modules when people try to build them with the system toolchain. This didn't fail like this before because R 3.6.0 is doing stricter checking on compiler functionality.
Also, it's not clear if EPEL6 will add devtoolset to the repo, since the EPEL6 repo is going to be end-of-lifed in 6 months. If they do, I can build it (the logic is added to the RPM now), but... do you have a CentOS 6 to 7 migration plan? :D
This update has been submitted for testing by spot.
This update test gating status has been changed to 'waiting'.
This update has obsoleted R-3.5.3-1.el6, and has inherited its bugs and notes.
This update test gating status has been changed to 'ignored'.
This update has been pushed to testing.
I have discovered an issue with this release when installing/compiling packages with R in that Makeconf in /usr/lib64/R/etc has missing compiler references, specifically c++11. using devtoolset-7 , you get "C++11 standard requested but CXX11 is not defined". In Makeconf all references with CXX11 are either blank or incorrect. This was not an issue in previous builds? I have manually set these values and it now works. If it helps i used: CXX11 = g++ was blank CXX11FLAGS = -O2 -g $(LTO) CXX11PICFLAGS = -fPIC CXX11STD = -std=c++11 was blank
Bodhi is disabling automatic push to stable due to negative karma. The maintainer may push manually if they determine that the issue is not severe.
quantumcheese: We have never built R with the devtoolset toolchain, but rather, the stock EL toolchain, which I do not believe has full support for C++11. This leads to the Makeconf not having CXX11 references. If we build R for EL with devtoolset, it gets tricky, because then R package building will break for any users not using devtoolset.
I think that most users will want C++11 support (which many modern R packages depend on), so I'll just adjust the package to build with it and Requires it.
This update has reached 14 days in testing and can be pushed to stable now if the maintainer wishes
(apologies for delay, and your quick response) Yes, i appreciate you don't compile R with a devtoolset, however many of the packages we need to install into R here req it, to pick one at random: biocLite("rtracklayer") from the bioconductor suite. I'm used to having to hop in and out of the devtoolset to install various R packages, but this time it wouldn't work until i made the change above. FYI our previous R build was 3.5.2 and this didn't suffer the issue, nor other previous versions. OS: CentOS 6.10
quantumcheese: I totally understand. Right now, the issue is that the epel6 buildroot does not include devtoolset, so I can't build R with it and properly generate Makeconf with CXX11 values. (Yes, I could do what you did and monkey patch them in, but then, I'll start getting build failure reports from random R modules when people try to build them with the system toolchain. This didn't fail like this before because R 3.6.0 is doing stricter checking on compiler functionality.
Also, it's not clear if EPEL6 will add devtoolset to the repo, since the EPEL6 repo is going to be end-of-lifed in 6 months. If they do, I can build it (the logic is added to the RPM now), but... do you have a CentOS 6 to 7 migration plan? :D
"do you have a CentOS 6 to 7 migration plan? :D" .... Its complicated! All the best. Carl
This update's test gating status has been changed to 'greenwave_failed'.
This update's test gating status has been changed to 'ignored'.
This update's test gating status has been changed to 'greenwave_failed'.
This update's test gating status has been changed to 'ignored'.
This update's test gating status has been changed to 'greenwave_failed'.
This update's test gating status has been changed to 'ignored'.
This update's test gating status has been changed to 'greenwave_failed'.
This update's test gating status has been changed to 'ignored'.
This update is marked obsolete because the EL-6 release is archived.