On Wed, Oct 05, 2022 at 03:17:01PM +0200, picca wrote: > So what is the right way to fix this issue if a library need this -N option > to be efficient on architecture which support it ?
Nearly every instance of this problem is with a testsuite, and it's easier to remove -N across the board and deal with a neglible slowdown on the buildds on architectures that do support it than to try to solve it any other way. Ideally we would have porters care about maximizing GHC features on their architectures but it generally doesn't happen.
