Second test, unicode002, I still can't wrap my mind around it. Basically, it generates a huge table with the result of 7 character tests for first 6554 codepoints. The way this test is mainted, I guess, is just replace the whole reference table with every Unicode standard update (like this was done, e.g., in this commit).
My current guess that T10412.hs should be deleted or changed to what we have now, and unicode002 should be just updated.
Isn't libraries/base/cbits/WCsubst.c generated by libraries/base/cbits/ubconfc though? That would make te latter the one to update I think. (As I recall libraries/base/cbits/WCsubst.c is regenerated when you build GHC but I'd have to check)
@Azel WCsubst.c is generated by ubconfc indeed, but I don't think this is done on rebuilding. Git history says WCsubst.c was updated 4 years ago last time. And the manual update using the current Unicode spec (v.12) fixes several bugs (Trac #5518, Trac #15525), which is the aim of this revision. We just need to decide what to do with T10412. /cc @bgamari