https://gcc.gnu.org/bugzilla/show_bug.cgi?id=4210
--- Comment #37 from Niels Möller <nisse at lysator dot liu.se> --- (In reply to Manuel López-Ibáñez from comment #35) > There is no such place. Dead code is identified in the middle-end and by > then, there is no parse tree, only GIMPLE and SSA: > https://gcc.gnu.org/onlinedocs/gccint/Passes.html#Passes So if emission of the warning is postponed to after the Tree SSA passes (https://gcc.gnu.org/onlinedocs/gccint/Tree-SSA-passes.html#Tree-SSA-passes). Could perhaps be inserted just after (or as part of) pass_remove_useless_stmts? > > 3. Alternatively, [...] construct a special tree object representing an > > expression with invalid/undefined behavior, and any meta data needed to emit > > a warning or error to describe it? Then emission of the warning could be > > postponed to later, say, close to the conversion to SSA form? > > That is still too early, since the dead code elimination happens after SSA. Then that special value needs to be passed through the conversions to gimple and SSA. I imagine it could be treated much like a compile time constant, but with an annotation saying it's invalid, and why, to be displayed in case the compiler thinks it needs to generate any code to evaluate it. (And if we go that route, we should probably do the same for most or all warnings on invalid operands detected by the frontend).
