Closed ZuseZ4 closed 6 months ago
I agree with the sentiment, though it'll be common enough for people to define ENZYME_DIR
or ENZYME_LIB
for their build systems.
I am not sure how many people will actually develop both rust and enzyme and set those as env vars? Usually you pass them directly to cmake as -D something, right? But I als0 wouldn't mind much adding them to the list of whitelisted flags.
Here's an example doing that. The user sets ENZYME_LIB=/path/to/ClangEnzyme-17.so
to activate Enzyme, which is otherwise optional. When I merge Rust support to this repository, it will still be able to coexist with the C/Enzyme code.
https://gitlab.com/micromorph/ratel/-/blob/main/Makefile?ref_type=heads#L57
https://gitlab.com/micromorph/ratel/-/blob/main/Makefile?ref_type=heads#L204
https://xkcd.com/1172/ ? But fair, I just didn't thought about pure makefiles.
There's an argument that it was a mistake that make variables set on the command line to be exported by default, but that bird has flown.
closing since both t-lang and t-compiler recommended to move this to flags, which IIRC already handle this.
Given the amount of ENZYME dbg vars, we should loop through all env vars and check if any unknown ones start with ENZYME, and abort compilation because it probably implies a typo.