I think there was an #include “compute.h” missing in thr_omp.cpp from the USER-OMP package? Compiling the original version gave me an “invalid use of incomplete class”-error, adding the include to the cpp made it go away. See attached patch. This is again for the 29 Aug 2015 version.
thr_omp.cpp.patch (254 Bytes)
I think there was an #include "compute.h" missing in thr_omp.cpp from the
USER-OMP package? Compiling the original version gave me an "invalid use of
incomplete class"-error, adding the include to the cpp made it go away. See
attached patch. This is again for the 29 Aug 2015 version.
this should also be fixed in the git / svn repo after the nightly sync and
some more small issues caused by the refactoring included in the last two
also, as of a few weeks ago, we are using the static code analysis from
Coverity Scan ( https://scan.coverity.com/ ) to more systematically check
for the little oversights where code works correctly except for obscure
corner cases or cut-n-paste bugs and similar hard to find issues (it also
finds a large number of false positives that are due to programming style
and logic that is not caught by the heuristics used in static code analysis.
overall the goal is to have a more reliable, correct and predictable code
base, but - as it often happens - when fixing problems or replacing code
with some cleaner and safer alternatives, it may expose other issues.
LAMMPS has grown so much that it would help immensely if people that have
the time would keep a git/svn checkout and compile the latest development
version and run quick tests on typical systems and report any problems here
or file an issue at https://github.com/lammps/lammps/issues
Ok, sure, I will do that. I like being up-to-date anyway, so I will probably run into these things when they happen.
If you stay behind the latest code version in the git repo by a day or so you will have less problems… In this case the problem was already fixed, but the changes hadn’t propagated to the git repo yet (which is updated only once a day).