On Wed, Jan 6, 2021 at 1:55 PM Marc Glisse <marc.gli...@inria.fr> wrote: > > On Wed, 6 Jan 2021, David Edelsohn via Gcc-patches wrote: > > > Currently the type of streamoff is determined at libstdc++ configure > > time, chosen by the definitions of _GLIBCXX_HAVE_INT64_T_LONG and > > _GLIBCXX_HAVE_INT64_T_LONG_LONG. For a multilib configuration, the > > difference is encoded in the different multilib header file paths. > > For "FAT" library targets that package 32 bit and 64 bit libraries > > together, G++ also expects a single header file directory hierarchy, > > causing an incorrect value for streamoff in some situations. > > Shouldn't we change that? I don't see why using the same directory for > linking should imply using the same directory for includes.
First, the search path assumption is rather strongly embedded in the GCC driver in somewhat delicate code. There already is a lot of fragile, complicated processing for multilibs and search paths and exception. It is more risky, both in the potential new search logic and in possible breakage, to perform surgery on the multilib support code. Second, it's confusing and error-prone to have different search behavior for different parts of the compiler until we can completely remove multilib headers. Third, there is no inherent reason that the header files should be multilib-dependent. I'm not trying to boil the ocean and remove all multilib differences in headers. I'm trying to solve specific issues caused by the differences in header files that also happen to move GCC in a direction of less multilib differences encoded in header files, which I think is a GOOD THING[tm]. Thanks, David
