Hello Steven, all, (Let's keep the mailing list in Cc:, please. Thanks!)
* Steven G. Johnson wrote on Fri, Sep 01, 2006 at 01:42:43AM CEST: > On Wed, 30 Aug 2006, Ralf Wildenhues wrote: > >The incomplete patch below chooses the first answer for both questions. > >Based on feedback I can supply documentation and finish my tests. > > That choice seems reasonable to me. > > However, I would consider giving an error if the user has .f90 or .f95 > files and does *not* use AC_FC_SRCEXT, as the special flags are > (unfortunately) required for some compilers. This sounds like a good idea to me. However, the question about introducing AM_FC_SRCEXT was rather meant the other way round: Can there be uses where you would like to know for some reason the answer to AC_FC_SRCEXT([f]) but in your source tree, you have files with extension .f that should be compiled with $(F77) instead of $(FC)? Then, decoupling the override from the Autoconf macro would seem like a better choice to me, even if it means that current packages would need to change AC_FC_SRCEXT -> AM_FC_SRCEXT. They need to adjust anyway, so I don't think that would be a big issue. Was this understandable? I'll work on an updated patch. Cheers, Ralf
