Hi Greg, I can take a look at removing the mutex and testing.
It looks like there's already `testMultiThread` in `External/INCHI-API/test.cpp` which covers most of the functions with the exception of `MolBlockToInchi`. Do you think it is sufficient to add coverage for `MolBlockToInchi` and enable `testMultiThread` (s/#if 0/#if 1 in main())?. Also, how can we run the tests with asan/msan/tsan and friends? Best, Jin On Mon, 5 Oct 2020 at 20:53, Greg Landrum <greg.land...@gmail.com> wrote: > Hi Jin Pan, > > That's a great question. It sounds like we probably can remove the mutex, > and it's definitely something worth trying (and then testing). > I won't have time to do this in the immediate future, but removing the > mutex and adding some tests that are analogous to the other multithreaded > tests is something others who are comfortable with C++ could do as well. :-) > > -greg > > > On Tue, Oct 6, 2020 at 4:04 AM Jin Pan <jin....@postera.ai> wrote: > >> Hello rdkit-discuss, >> >> Commit 6cfd34 >> <https://github.com/rdkit/rdkit/commit/6cfd347eafc713a66ba66051d2fc21046e60be58> >> from >> 2012 introduced a mutex to make the InChi code threadsafe. InChI v1.05 was >> released in 2017, with release notes announcing better thread safety: >> >> As global and static objects may cause problems in a multi-threaded >>> environment due to a possible race condition arising when more than one >>> thread tries to access, they had to be removed from the InChI code. >>> Corresponding changes were implemented. >>> In particular, global variables associated with sorting and timing were >>> replaced by objects which are passed to all functions that need access to >>> the related data. Also, several functions which uses static variables that >>> were initialized upon first entry into the function have been modified in >>> order to avoid a race condition. >>> Tests were performed for various applications using InChI Software >>> Library and running up to 32 threads concurrently, and for datasets up to >>> the whole Pubchem Compound set. No problems were encountered. >>> >> >> Can we safely remove the mutex with the v1.05 InChI release, or do we >> still consider that library to be thread unsafe? >> >> Best, >> Jin Pan >> _______________________________________________ >> Rdkit-discuss mailing list >> Rdkit-discuss@lists.sourceforge.net >> https://lists.sourceforge.net/lists/listinfo/rdkit-discuss >> >
_______________________________________________ Rdkit-discuss mailing list Rdkit-discuss@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/rdkit-discuss
