Julianus Pfeuffer
Julianus Pfeuffer
Similar on the KNIME side. If we add another ifcase for this parameter type, interpret it as a file prefix and just collect every file under that path, it should...
It might indeed be backwards compatible if the new tag is added in addition only and KNIME ignores it and just interprets it as a prefix. But I am not...
Oh another question: do you plan to write subfolders with this or imagine that someone might want to? Not sure if we are currently collecting files recursively.
No, makes sense. They do need it for faster linalg. It's in the recipe since 7 months at least. Probably no one tested Intel Macs for signing.
Ah I just saw the error report. Yes, that is an error in the blas recipe then.
gcc15 is used on homebrew since May 25. By now recipes should have been rebuild. Please test again with nightly pkg.
As long as the CPP append is not wrapped, it should be fine. I still think autowrap support would have been more future proof, considering all the boilerplate but fine.
Idk something like #wrap-vector-member or #wrap-vector-interface with some reasonable defaults for Len, iter, getsetitem and append/extend.
We do have wrap-iter and wrap-size AFAIR but I have not seen any usage of non-defaults so far.
Sorry I mean wrap-upper-limit which is kind of different and could be used for any input param. But I think it is only used for vector-interface classes.. therefore maybe one...