On Mon, Jan 07, 2002 at 01:20:45AM +0100, Wichert Akkerman wrote: > Previously David N. Welton wrote: > > Right, so how do we fix this? It is our problem, in that we need to > > make the software we distribute work together. But are you also > > saying that upstream shouldn't be setting that bit in their header > > file? > > As long as the API (and ABI) never exports things like struct stat and > offset_t (ie things that are affected by enabling LFS) it should not > matter if you link things that are compiled with LFS with things that > are not. > > However that define should never be in a header file, especially > not one that other applications can use. Doing that should warrant > a hefty bugreport.
And if they do require the LFS for their interface, then I'd go with: #ifndef _FILE_OFFSET_BITS #error _FILE_OFFSET_BITS must be defined for foo #endif Which leaves the app to explicitely enabling it, either at the top of the .c files, or with CFLAGS. -- Adam Olsen, aka Rhamphoryncus
