[BUG] Failed compilation with Cuda and c++ 11.1.0

[lorenzo-rovigatti]

[BUG] Failed compilation with Cuda and c++ 11.1.0

[mmguar]

Thank you for the pointer.
update-alternatives is unfortunately unavailable in arch, but the solution was to create a simlink in /usr/local/bin for gcc, g++ and cc (see here):
sudo ln -s $(which gcc-5) /usr/local/bin/gcc sudo ln -s $(which g++-5) /usr/local/bin/g++
Unfortunately that is not enough as the command that gives the error specifies -ccbin /usr/bin/cc, so it was necessary to temporary replace /usr/bin/cc:
sudo mv /usr/bin/cc /usr/bin/cc.new sudo cp /usr/bin/cc-10 /usr/bin/cc
Compilation is fine then.
If anyone was to stumble upon this again, please remember to revert to your standard gcc with:
sudo rm /usr/bin/cc sudo mv /usr/bin/cc.new /usr/bin/cc sudo rm /usr/local/bin/gcc /usr/local/bin/g++ hash -r
Thank you again Lorenzo.

[lorenzo-rovigatti]

Hi! What do you mean by "c++ 11.1.0"? Is that your g++ version? Or the CUDA version?

Edit:
Can you copy the output of these two commands?

nvcc --version
g++ --version

[mmguar]

Hi Lorenzo. Apologies for being unclear.
g++ and gcc in the path end up being the cuda versions

[matteo@matteOs4 build]$ g++ --version
g++ (Arch Linux 10.2.0-3) 10.2.0
Copyright (C) 2020 Free Software Foundation, Inc.
This is free software; see the source for copying conditions.  There is NO
warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.

[matteo@matteOs4 build]$ type g++
g++ is hashed (/opt/cuda/bin/g++)

aliasing them to /usr/bin/gcc shows version 11.1.0, which should be correct, but does not solve compilation issues.
As for nvcc

[matteo@matteOs4 build]$ nvcc --version
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2021 NVIDIA Corporation
Built on Sun_Mar_21_19:15:46_PDT_2021
Cuda compilation tools, release 11.3, V11.3.58
Build cuda_11.3.r11.3/compiler.29745058_0

[lorenzo-rovigatti]

OK, it seems there is a mismatch between the compiler and the libs it uses to compile. If you look at the errors, they refer to the path /usr/include/c++/11.1.0, which seems to indicate that the system compiler is g++ 11.1.0, which is not the one used to compile the code. I'm not sure how to solve this, but it doesn't seem strictly related to oxDNA. Is this a cluster node? Is there anyone who can help you with this?

[mmguar]

Hm, I understand that this is not an oxDNA problem most likely and I appreciate the support.
I removed the wrong bindings to gcc and g++, now they all show version 11.1.0, which is the one installed system-wide. The previous ones were the one shipped with cuda, which overrode the system ones through a poor PATH specification.
So now I believe I am compiling oxDNA with gcc 11.1.0 (--version confirms it), but I do still get the same errors.

Unfortunately it is my own computer and thus I have to debug it myself. I am more so looking for any pointers towards the right direction. It might also just be that arch is too "nightly" and cuda will not compile correctly before an update.

[lorenzo-rovigatti]

As far as I know CUDA 11.3 is incompatible with g++ > 10. You'll have to install an older version and set it as the default (if you using ubuntu there is the handy update-alternatives command that can help you switch between versions).