Aleksej Saushev <asau%inbox.ru@localhost> writes: > Things are these: all packages that can work with F77 were converted to > "USE_LANGUAGES" containing "fortran77", "fortran" is now reserved for > more modern Fortran. Currently, there're no packages that utilise this > separation, and in fact, "fortran" and "fortran77" are synonymous for now. > But I have heard wishes to import packages that require F90 at least, > I have changes in my local tree that switch to using "g95" for packages > that use "fortran" rather than old "fortran77", and I propose to commit > these changes. Not knowing the details, it sounds like we should leave fortran77 in USE_LANGUAGES to mean Fortran 77, and add fortran90 and fortran95, and not use bare 'fortran' because it's likely to be misinterpreted. I would certainly expect that to mean F7. Then making g95 be the default compiler for fortran95 sounds like a good plan. Beyond that, it seems like there should be some way for the user to specify the preferred toolchain for a language, from the set that's know to mostly work. > From user point of view, noone is writing in F77 nowadays, there're more > and more software packages that break compatibility with F77 and convert > either to F90 (at least) or F95. I'm not sure we need to assume who is writing what, but it seems clear that significant software exists in each language spec and that we should have some way to support them all. I appreciate you working on this - some time I should try out the NGA geoid model - so I don't mean to sound too cranky.
Attachment:
pgpnMglTX1JV9.pgp
Description: PGP signature