https://gcc.gnu.org/bugzilla/show_bug.cgi?id=121238
--- Comment #3 from Rohan Lean <rohan at rohanlean dot de> --- (In reply to Patrick Palka from comment #2) > My few attempts at reducing this testcase further via cvise resulted in > non-deterministically buggy code (where in non-modules mode the testcase > would unreliably crash)... maybe someone else will have better luck. I had feared that it would be a difficult case for automatic reduction, because the correct workings depend on large amounts of the formatting library being included. My intuition is that the best way forward would be either stepping through with a debugger or inspecting the (unfortunately large) outputs, ideally with an inkling of what to look for. It seems like some kind of linking error to me, in a very general sense. Since the module should not actually affect the output at all, perhaps one can diff with the non-module case?
