reconverse-support branch of charm fails to build at configuring hwloc

[trquinn]

Running ./build charm++ netlrts-linux-x86_64 -g -O2 --force -j2 on a fedora 42 laptop gives the following. The same error also occurs when building on TACC stampede3.

--` Configuring done (52.0s)
-- Generating done (0.2s)
-- Build files have been written to: /home/trq/nchil/src/charm/netlrts-linux-x86_64
[  0%] Generating pup_f.f90
[  0%] Creating directories for 'hwloc'
[  0%] Performing download step (DIR copy) for 'hwloc'
[  0%] No update step for 'hwloc'
[  0%] No patch step for 'hwloc'
[  0%] Performing configure step for 'hwloc'
configure: error: source directory already configured; run "make distclean" there first
make[2]: *** [CMakeFiles/hwloc.dir/build.make:92: hwloc-prefix/src/hwloc-stamp/hwloc-configure] Error 1
make[1]: *** [CMakeFiles/Makefile2:836: CMakeFiles/hwloc.dir/all] Error 2
make[1]: *** Waiting for unfinished jobs....
[  0%] Building Fortran object CMakeFiles/conv-utilf.dir/pup_f.f90.o
[  1%] Linking Fortran static library lib/libconv-utilf.a
ar: creating lib/libconv-utilf.a
[  1%] Built target conv-utilf
make: *** [Makefile:136: all] Error 2

[ritvikrao]

Try using multicore instead of netlrts, we haven't tested with netlrts. But I don't think that will fix this error, I am just curious what will happen

[trquinn]

I did get it to build if I installed hwloc on the system via dnf.

[ritvikrao]

Good to hear! We have set up the reconverse-level cmake to build if hwloc isn't detected, but this implies we need to change the charm cmake to do so as well if we build on reconverse. One thing that we've changed is that when running charm on reconverse, we are using the system's hwloc, not the Charm-specific implementation in the contribute folder.

[JiakunYan]

I would always use multicore when building charm with reconverse (for both single-node and multi-node runs. Reconverse will handle both cases automatically). There is no reason to enable the netlrts machine layer alongside reconverse, and I am not sure whether it results in any strange behavior during compilation/runtime.