Branko Čibej wrote: > On 20.03.2013 16:16, Julian Foad wrote: >> C. Michael Pilato wrote: >> >>> Since svn_hash.h includes svn_types.h, won't this be more like > replacing the >>> inclusion of the latter with the inclusion of the former? >> I'm not sure exactly what you mean, but if we do decide to leave the > definitions in svn_hash.h and add '#include <svn_hash.h>' to each > C file, that would be functionally equivalent to replacing one #include > directive with the other, because of the include-guards. I would oppose > actually removing '#include <svn_types.h>' from the source files, > as a matter of style. > > Actually, since we have svn_hash.h, and that (I hope) includes > apr_hash.h, and presumably files that refer to APR hash functions also > include apr_hash.h ... we're looking at > > s/apr_hash.h/svn_hash.h/ > > in every .c file. So /if/ we decide to accept the code churn, it seems > to me this #include change makes the most sense.
Good point. +1. There was a flaw in my argument: svn__apr_hash_index_key etc. are in svn_types.h because they are intended to be temporary names for functions that APR should eventually provide, whereas the newer 'svn_hash_gets' etc. are not. - Julian
