You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Hi,
thank you for the excellent pointer.
On HackerNews a user volunteered to take care of this: I will wait a
couple of weeks to see if they actually manage to produce a conda
recipe. If not, I will investigate this myself.
For future reference, I have the following two questions/doubts.
Q1. If I understand correctly, librosa does not have Python C/C++
extensions, however what guides the conda build is basically this line:
https://github.com/conda-forge/librosa-feedstock/blob/master/recipe/meta.yaml#L16
So, in my case, I just need to make sure the C extensions are compiled
correctly, and, if so, it should be good to go. Note, however, that the
current setup.py requires numpy to be installed BEFORE aeneas, because
to find the numpy headers for the C extensions, the package numpy should
be importable. This aspect might be checked. (A workaround consists in
suggesting to conda install numpy first, as done for pip.)
(Of course a better solution would be solving the numpy headers issue in
the first place.)
Q2. aeneas also needs ffmpeg and espeak to be present in the
environment. The first (ffmpeg) is already available in conda-forge, and
it is needed only at run time. The second (espeak) is not on conda-forge
--- which might imply that its own recipe should be created first. Also,
espeak must be present at setup time, since the cew Extension needs to
link against libespeak.
See https://conda-forge.github.io/
I have no good intuition on how to specify ffmpeg and eSpeak(-ng) as dependencies.
The text was updated successfully, but these errors were encountered: