Closed vsnever closed 1 year ago
I think it would be good to push out a Cherab release against 0.7.1 first, and then incorporate the changes required for 0.8 into a subsequent release. It's possible that the API breakage in 0.7.1 may break some users' downstream workflows if they use raysect features directly as well as Cherab.
Yes, that's probably right. I'll leave this PR open for the future though.
Now that Cherab 1.4.0 is out I think we can look at updating the development branch for raysect 0.8. Tests all pass: any objections to the merge?
not from my side, thanks
Just one thing, can we temporarily specify that we need matplotlib<=3.5.3
, because this set_window_title()
issue was fixed only in Raysect 0.8.1, which is not yet released?
Not too keen on working around raysect's bugs, but in this case setting a limit on the matplotlib version does seem like a pragmatic solution. I'll make the change.
I'll release Raysect v0.8.1 this weekend, if I get time.
Thanks @CnlPepper. In that case I propose we wait until 0.8.1 is out and then just support that going forwards. I can't see anything in the 0.8.1 changelog that would affect Cherab: there are no instances of extract_translation
or extract_rotation
in Cherab's source code.
@jacklovell I've released Raysect 0v.8.1. It fixes a number of niggling issues, such as the deprecated matplotlib window title call. If there are any further issues, let me know.
@jacklovell, should we add the -O3
compilation flag like Raysect 0.8.1 does?
I'd recommend you do. Anaconda and other python distributions seem to have changed their defaults to -o2, which results in a performance drop.
Interesting. Manylinux still uses -O3
which is what I build the PyPI wheels on, so forcing the use of -O3
in setup.py shouldn't make a difference for most users. I'll make this change and then merge.
Raysect 0.8 has been released, so since the development branch should build against Raysect's master, the
raysect-0.8-compat
branch is ready to be merged into development.