> ISTR it was changed because at least with location info generated > by gas there's no way to have "no location" for a portion of code. > Instead the assigned location will be that of the previous .loc > directive which results in random and confusing results for the > pc range of the DECL_INGORED_P function.
Yes in the general case, but no if you put them at the beginning of the assembly file (what the Ada compiler precisely does), at least if you do not pass any -gdwarf-n switch now. This had worked for 2 decades at least... > I guess we should really revisit the decision to rely on gas > to produce line info. What's the advantage of doing so (apart > from "nice" annotated assembly)? Not a small advantage in my opinion, and I don't think that we want to change it because of a corner case in Ada in any case. I guess Bernd's patch is acceptable, modulo a small tweak for -gnatD. Let me experiment a little bit though. -- Eric Botcazou
