> On Fri, 26 Jan 2018, Jan Hubicka wrote: > > > > On Thu, 25 Jan 2018, Jan Hubicka wrote: > > > > > > > Hi, > > > > the testcase triggers invalid warning on type mismatch because array > > > > of pointers to complete type has different alias set from array of > > > > pointers > > > > to incomplete type. This is valid, because incoplete pointer has alias > > > > set > > > > of void_ptr which alias all pointers and arrays alias with their > > > > members. > > > > > > But isn't that a problem? Not if the pointer to incomlete type have > > > alias set zero but IIRC they don't, right? > > > > pointers to incomplete type are same as alias set of ptr_type_node (1 in > > this case) > > and pointer to complete types have different alias set but that set has > > is_pointer flag. When checking for conflict we make 1 to alias with > > everything that > > has is_pointer flag in it, so it will return true for these two array types. > > Ah, ok. The patch is ok then. I suppose in theory complex pointers > or vector pointers would have the same issue if we'd ever generate > those (GIMPLE at least doens't like them ;)).
Yep, VECTOR_TYPE is only remaining case where get_alias_set recurse, but I would wait for those to be actually used. We will not only generate them but produce static variables of them for waring to triger. Honza
