Tom Eulenfeld

See also #2758

@megies I am temporarily reopening this issue to discuss how it is best fixed for the conda-forge feedstock. Pinning libffi to ~~`

Of course, sorry for the confusion.

Interesting. Can you please also post the values of `reftime` and `move` variables? The different behaviour is probably not intended. Stream.trim and Stream.slice should give the same result. The following...

Minimal reproducing example: ```py from obspy import read st = read() utc = st[0].stats.starttime dt = 1.045 print(utc+dt) print(st.copy().trim(None, utc+dt)[0]) print(st.slice(None, utc+dt)[0]) print(len(st.copy().trim(None, utc+dt)[0]) == len(st.slice(None, utc+dt)[0])) ``` outputs ```...

This issue should be fixed in the linked PR.

I think its not a big deal to support 3.8? Maybe we need to mark this one stubborn test case as expected fail for linux py3.8.

I realize the last commit should go into maintenance. I can cherry-pick it after merging.

I dropped commit [06e9e1c](https://github.com/obspy/obspy/pull/3087/commits/06e9e1cb7307764778ee8fe154f80f2159ba4bd6), it is already implemented.

Not sure if it helps, but you could use asdf format for writing waveforms. An alternative is obspyh5 plugin allowing to write/read HDF5 files directly with ObsPy. Metadata should be...