Eric Blake <[EMAIL PROTECTED]> wrote: > According to Eric Blake on 12/27/2006 6:18 AM: >> >> I'm checking this in to fix this issue on platforms that do not need a >> replacement <wctype.h>. It is not a true reversion, since it rearranges >> the include order to let <wctype.h> pulls in prerequisites required for >> <wchar.h> on platforms where <wchar.h> is not self-standing. >> >> 2006-12-27 Eric Blake <[EMAIL PROTECTED]> >> >> * lib/fnmatch.c: Reinstate inclusion of <wchar.h>, since this >> module uses more than what <wctype.h> is required to provide. > > I'm supposing the alternative would be to make the wctype module smarter, > and implement the CX shading in POSIX that says that <wctype.h> is allowed > to pull in all of <wchar.h>. Then gnulib modules should only ever need to > include <wctype.h> and not worry about <wchar.h>, regardless of whether > the platform's wctype.h opted not to pull in wchar.h.
Thanks for helping out. FYI, I've just noticed that Bob's buildbot is showing compile failures for all HPUX builds: http://proulx.com:9000/ I won't have time to investigate today. In file included from fnmatch.c:172: fnmatch_loop.c: In function `internal_fnmatch': fnmatch_loop.c:245: error: `wctype_t' undeclared (first use in this function) fnmatch_loop.c:245: error: (Each undeclared identifier is reported only once fnmatch_loop.c:245: error: for each function it appears in.) fnmatch_loop.c:245: error: parse error before "wt" fnmatch_loop.c:275: error: `wt' undeclared (first use in this function)
