From: Arjen M. <arj...@wl...> - 2004-01-08 07:20:28
|
"Alan W. Irwin" wrote: > > On 2004-01-07 08:46+0100 Arjen Markus wrote: > > > Given the current set-up (Fortran programs call stubs to the actual C > > routines) > > it should be simple to solve this: > > > > for these compilers: > > #define FNAMEU(x,y) y##__ > > > > otherwise: > > #define FNAMEU(x,y) y##_ > > > > This introduces an extra STUB_LINKAGE case, but it keeps the Fortran > > names > > identical to the C ones. > > This solution is not correct because the actual behaviour of the fortran > compilers is more complicated: if there is an embedded underscore then > append two underscores, otherwise append one. With this convention, > "hello_world" maps to "hello_world__" while "helloworld" maps to > "helloworld_". The above solution would map every name to a double > underscore suffix. > I am sorry, I was unclear: I meant the introduction of an additional macro, for names with underscores: In the ordinary case: #define FNAME(x,y) y##_ #define FNAMEU(x,y) y##_ and if embedded underscores introduce two underscores at the end: #define FNAME(x,y) y##_ #define FNAMEU(x,y) y##__ > > Arjen, as discussed elsewhere, please send your changes for sys/win32/msdev > to me as a patch (or patches) which you would like me to apply. > Yes, I will do that (one last thing to check is the issue of tmpfile()). Sending the patches in the format you asked should be no problem. Regards, Arjen |