keep any automake discussion public please >>> "Michael" == Michael R Nolta <[EMAIL PROTECTED]> writes:
[...] Michael> Not all files produce modules; only files containing Michael> "module x ... end module" blocks do. And there can be Michael> multiple module definitions per file. >> I hope that does not mean many *.mod file can be produced by >> a single compilation. Michael> Yes, it does. :( I don't know how that can be handled. I assume any of these modules can be included (if that's the word) by another .f90 file during the build. All the .f90 files must therefore be compiled in the correct order, so that .mod are available to others as needed. One way to ensure this is with proper Makefile dependencies. However the only way to know which .mod files a .f90 will produce is to poke into that .f90, isn't it? That would therefore had to be done at automake-time, but this is a very unpleasant thing to do. The Texinfo handling code already does something similar (it search @setfilename into the .texi files) and we have hit a problem because of this: the file must exist when automake is run. This hinders support of generated .texi files. Back to F90, that would mean that if automake has to scan the .f90 files, any .f90 file which is actually produced at make time (and couldn't be scanned by automake) is not allowed to output modules. [...] Michael> Yes, you're right. Should I say "nodist_EXTRA_HEADERS = Michael> ..." too? >> You can. It even looks more sensible. But actually it's >> superfluous: EXTRA_* is really useful when there are extra >> files to distribute or extra compile rule to generate. Michael> But if I omit EXTRA_HEADERS (or if I use Michael> nodist_EXTRA_HEADERS), then I get the following error Michael> message: Michael> f90/Makefile.am:11: `nodist_include_HEADERS' contains configure Michael> substitution, Michael> f90/Makefile.am:11: but `EXTRA_HEADERS' not defined Well, so it is required then... Sorry for the confusion. -- Alexandre Duret-Lutz
