Module search is not predictable about zz

[aep]

Thanks for reporting. This is indeed not correct, but fortunately easy to fix

[aep]

@sternenseemann as for your actual issue with zz referencing the build dir, i believe this also needs to be addressed.

How about we just add a super long string to the default list which you can sed replace?
Odd that there's no tool for replacing strings in binaries properly.

Or could you patch the source if we added all the distro specific stuff to a single file?

[sternenseemann]

Since we need to generate a shell wrapper anyways for z3, patching out the CARGO_MANIFEST_DIR thing is actually enough. Also /nix/store/modules remaining in the search path is not pretty, but no problem at all since that directory will never exist.

I'm not sure which search path entry you are referring to exactly with your proposal, but I'd prefer it to patch the source rather than the binary. It doesn't need to be sed-able, I can also just vendor in a patch, but it'd be nice if that wouldn't require updating at every version bump.

[aep]

specifically would it help if we moved this

zz/src/lib.rs

Line 120 in 9b4bfa0

searchpaths.push(std::path::Path::new(env!("CARGO_MANIFEST_DIR")).join("modules"));

to a separate file that only contains something like

const vendor_modules = vec![std::path::Path::new(env!("CARGO_MANIFEST_DIR")).join("modules"))];

so you can replace it here

since the file will just contain vendor paths, patches will probably apply for a long time

[sternenseemann]

Yes that would work!

Also I'm wondering if we could make this bit a configuration option or something while we're at it:

zz/src/lib.rs

Lines 107 to 119 in 9b4bfa0

	searchpaths.push(
	std::env::current_exe()
	.expect("self path")
	.canonicalize()
	.expect("self path")
	.parent()
	.expect("self path")
	.parent()
	.expect("self path")
	.parent()
	.expect("self path")
	.join("modules"),
	);