Guillaume Fraux

Thanks a lot, this helped! Unfortunately, it looks like the issue come from the Julia bindings. I can't reproduce this with just C++ code; and adding a `Base.GC.gc()` call seems...

From reading around, it seems that the best solution here would be to make Julia's GC aware of the allocations done in C++ by chemfiles, and the recommended way to...

> In my real use case I do not open and close the trajectory multiple times. I am just running the script once, so there should not be multiple opening/closing....

Thanks for letting us know @axsk. I tried to debug what was going on for a bit, but could not find the core origin of the issue. I'll try to...

Some more debugging notes: - `finalize` does not help because it does not actually call the finalizer immediately. I don't know why, but the workaround can be to manually call...

More debugging notes: the memory increase in C++ comes from the `mmap`ed file loading its data. I'll check if there is a way to tell to the OS that it...

Yes, I spoke too fast, all binary format would have the same issue & grow the memory usage until they either read the whole file or use all the available...

Ok, so there where two separate things going on here, with separate fixes. First, there is an issue with creating a new frame at each step in Julia because the...

> I think the issue with the above examples is that, if the memory grows to 95%, there is not much room for anything else to happen in the system,...

ok, the mmap code has been merged and backported to the 0.10 release. You should be able to give it a try by compiling chemfiles yourself with ```bash git clone...