On 13 Oct 2010, at 03:55, Bruno Haible wrote: >> Originally, I added noinst_HEADERS to the module definition files, >> and then changed them all to nobase_nodist_include_HEADERS and >> EXTRA_HEADERS in gnulib-tool if the gnulib library is installable. >> However, I think it is much more flexible to add the >> 'nodist_include_HEADERS' or 'nobase_nodist_include_HEADERS' and >> 'EXTRA_HEADERS' as appropriate to the circumstances in each module >> file > > I agree that declaring them with ..._include_HEADERS is clearer than > with noinst_HEADERS. > > But why bother using nodist_include_HEADERS in some cases and > nobase_nodist_include_HEADERS in others? I would use > nobase_nodist_include_HEADERS everywhere, for consistency. It avoids a > pitfall when someone is writing a new module, say, 'sys_mman', based on > 'unistd'.
For pedantic correctness only; using nobase_nodist_include_HEADERS consistently makes more sense in light of what you say -- and actually reduces the number of lines added to gnulib-tool and generated Makefile.am too. > And what's the role or effect of EXTRA_HEADERS? Automake does not do > anything with .h files that it doesn't do with .png files. gnulib-tool > already synthesizes EXTRA_DIST augmentations for files that are to be > distributed. Automake complains if you add a computed (possibly empty) header name to xxx_include_HEADERS, without a matching EXTRA_HEADERS containing the static name. I added it purely to silence automake, but would be happy to remove it if that can be done without unnecessary automake invocation noise. Cheers, -- Gary V. Vaughan (g...@gnu.org)
PGP.sig
Description: This is a digitally signed message part
