make error: multiple definitions

[radfordneal]

make error: multiple definitions

[radfordneal]

You could try it with --enable-R-shlib and CFLAGS="-g -O3 -fcommon -ffp-contract=off" as configure options.

I'll try to install gcc-10 soon and see where I'm relying of -fcommon, but I think the above may fix the problem for now.

[radfordneal]

These problems are now fixed in branch 93. There will be a new release with these fixes (and other changes) soon.

Thanks for reporting this!

[radfordneal]

Could you say what version of gcc this is? Just type "gcc --version" to find out.

I notice that the release notes for gcc-10 now say that -fno-common is the default (rather than -fcommon). It's possible that I accidently relied on the "common" behaviour.

Or maybe this is not the problem at all, if you're not using gcc-10. I haven't installed a gcc-10 compiler myself yet.

What options you gave to configure (eg, CFLAGS) would also help.

[danmrc]

I am using gcc 10.1.0 and the only flag was --enable-R-shlib

[danmrc]

Hi,

The -fcommon solve that problem.
It then threw a error at file /src/library/stats/src/lminfl.o:

 91 |         call dqrsl(x, ldx, n, k, qraux, sigma, sigma, dummy,
      |                                                      2       
......
  111 |                call dqrsl(x, ldx, n, k, qraux, sigma, sigma, dummy
      |                                                             1
Error: Type mismatch between actual argument at (1) and actual argument at (2) (REAL(4)/REAL(8)).
make[5]: *** [../../../../etc/Makeconf:145: lminfl.o] Erro 1

I looked around and using FFLAGS = "-fallow-argument-mismatch" compiles and allows me to run pqR. I don't know if it generates any side effects, but it seemed fine.

check fails at stat with a segfault

It works fine with RStudio. Looking forward to use R with auto diff.

Let me know if you want more info/a new issue to separate this problems.

Lots of thanks for the help!

[radfordneal]

Thanks for investigating some more! Looks like I need to investigate several issues with gcc-10.