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
Cython 3.0.0 has recently been released (July 17, 2023), and has likely backward incompatibilities with the Cython 0.2* series. Since X-PSI relies on Cython to be build afaik, this may cause build problems.
Ultimately, if there are build errors with Cython 3, it should be fixed here in the code that uses and builds Cython, so that X-PSI is forward-compatible and future-proof.
A temporary(!) work-around is to set stricter requirements for the Cython dependency in setup.py. So this line
setup_requires = ['cython'],
may become
setup_requires = ['cython ~= 0.23'],
(or whatever minimal Cython version works. 0.29 might be fine.)
To clarify, the compatible release specifier, ~= 0.23, is similar to >= 0.23, == 0.*. Which you could also use if you find that clearer.
Aside, something similar should be the case for the NumPy dependency. There should be a minimum version requirement, and a maximum version requirement, for any dependency.
The text was updated successfully, but these errors were encountered:
evertrol
changed the title
Newly released Cython may be backward incompatible for an X-PSI
Newly released Cython may be backward incompatible for an X-PSI installation
Aug 3, 2023
Cython 3.0.0 has recently been released (July 17, 2023), and has likely backward incompatibilities with the Cython 0.2* series. Since X-PSI relies on Cython to be build afaik, this may cause build problems.
Ultimately, if there are build errors with Cython 3, it should be fixed here in the code that uses and builds Cython, so that X-PSI is forward-compatible and future-proof.
A temporary(!) work-around is to set stricter requirements for the Cython dependency in
setup.py
. So this linemay become
(or whatever minimal Cython version works. 0.29 might be fine.)
To clarify, the compatible release specifier,
~= 0.23
, is similar to>= 0.23, == 0.*
. Which you could also use if you find that clearer.Aside, something similar should be the case for the NumPy dependency. There should be a minimum version requirement, and a maximum version requirement, for any dependency.
The text was updated successfully, but these errors were encountered: