parser not building under C++11
35 views
Skip to first unread message
Ben Goodrich
Jun 23, 2016, 3:04:10 PM6/23/16
to stan development mailing list
I am getting a build error from the parser when C++11 is explicitly or implicitly (in the case of g++-6) used. I think it was working before, but I can't think of anything I would have done that could affect it. Log attached.
Ben
Ben
Daniel Lee
Jun 23, 2016, 3:05:26 PM6/23/16
to stan-dev mailing list
Is there a way I can reproduce this locally? Or possibly log into a machine that has it set up?
DanielOn Thu, Jun 23, 2016 at 3:04 PM, Ben Goodrich <goodri...@gmail.com> wrote:
I am getting a build error from the parser when C++11 is explicitly or implicitly (in the case of g++-6) used. I think it was working before, but I can't think of anything I would have done that could affect it. Log attached.
Ben
--
You received this message because you are subscribed to the Google Groups "stan development mailing list" group.
To unsubscribe from this group and stop receiving emails from it, send an email to stan-dev+u...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.
Ben Goodrich
Jun 23, 2016, 5:51:45 PM6/23/16
to stan development mailing list
On Thursday, June 23, 2016 at 3:05:26 PM UTC-4, Daniel Lee wrote:
Is there a way I can reproduce this locally? Or possibly log into a machine that has it set up?
You can pull all of the rstan repo, including the submodules in StanHeaders/inst/include and then do
R CMD build StanHeaders
R CMD INSTALL StanHeaders_2.10.0-2.tar.gz
R CMD build rstan/rstan
R CMD INSTALL rstan_2.10.1.tar.gz
I fixed it by changing rstan/rstan/src/Makevars to define BOOST_RESULT_OF_USE_TR1_WITH_DECLTYPE_FALLBACK rather than -DBOOST_RESULT_OF_USE_TR1 -DBOOST_NO_DECLTYPE but that doesn't work under the pre-C++11 standard, so now it is using C++11 on everything but Windows and leaving -DBOOST_RESULT_OF_USE_TR1 -DBOOST_NO_DECLTYPE instead of -DBOOST_RESULT_OF_USE_TR1_WITH_DECLTYPE_FALLBACK on Windows.
Ben
Daniel Lee
Jun 23, 2016, 5:52:38 PM6/23/16
to stan-dev mailing list
Sorry -- I don't have that compiler.
Do you have it on a machine that I can remote into?
DanielDo you have it on a machine that I can remote into?
Ben Goodrich
Jun 23, 2016, 5:54:14 PM6/23/16
to stan development mailing list
On Thursday, June 23, 2016 at 5:52:38 PM UTC-4, Daniel Lee wrote:
Sorry -- I don't have that compiler.
It is the same with clang++. It is just a matter of whether you use --std=c++11 or not.
Daniel Lee
Jun 23, 2016, 5:55:59 PM6/23/16
to stan-dev mailing list
Ok. I'll try it now.
--
Daniel Lee
Jun 23, 2016, 6:04:02 PM6/23/16
to stan-dev mailing list
Ben, I'm gonna need more help.
What branch are you on? On that branch, are the flags set up so it'll fail?Ben Goodrich
Jun 23, 2016, 6:08:24 PM6/23/16
to stan development mailing list
On Thursday, June 23, 2016 at 6:04:02 PM UTC-4, Daniel Lee wrote:
Ben, I'm gonna need more help.What branch are you on? On that branch, are the flags set up so it'll fail?
develop, and the flags are set up so that it doesn't fail. To get it to fail, you would first have to do something like
cp rstan/rstan/src/Makevars.win rstan/rstan/src/Makevars.
Ben
Daniel Lee
Jun 23, 2016, 6:27:08 PM6/23/16
to stan-dev mailing list
I did that and I'm ok. There are a lot of warnings, but it compiled.
The output is attached. Let me know how you want to proceed. Here is what I did:
- I cloned rstan
- I updated rstan to the develop branch: 48a50a94c
- I went to StanHeaders/inst/include/mathlib and updated to develop: 91cb184a9
- I went to StanHeaders/inst/include/upstream and updated to develop: 0c492ab81
- I did this: cp rstan/rstan/src/Makevars.win rstan/rstan/src/Makevars
- then I did:
R CMD build StanHeaders
R CMD INSTALL StanHeaders_2.10.0-2.tar.gz
R CMD build rstan/rstan
R CMD INSTALL rstan_2.10.1.tar.gz
The output is from the last line.
Can you try with a different compiler with -std=c++11?
Daniel
Ben Goodrich
Jun 23, 2016, 6:43:26 PM6/23/16
to stan development mailing list
On Thursday, June 23, 2016 at 6:27:08 PM UTC-4, Daniel Lee wrote:
I did that and I'm ok. There are a lot of warnings, but it compiled.The output is attached. Let me know how you want to proceed. Here is what I did:- I cloned rstan- I updated rstan to the develop branch: 48a50a94c- I went to StanHeaders/inst/include/mathlib and updated to develop: 91cb184a9- I went to StanHeaders/inst/include/upstream and updated to develop: 0c492ab81- I did this: cp rstan/rstan/src/Makevars.win rstan/rstan/src/Makevars
The cp accidentally overwrote the instructions to use C++11, in which case it should work. You can force C++11 by first defining
export USE_CXX1X=1and then doing
R CMD INSTALL rstan_2.10.1.tar.gz
again. I can trigger the error with either g++ or clang++ using C++11 and -DBOOST_RESULT_OF_USE_TR1 -DBOOST_NO_DECLTYPE but g++ has a less readable error log than the clang++ one I first attached.
Ben
0 new messages