On 17.09.2009 08:10, William A. Rowe, Jr. wrote: > rj...@apache.org wrote: >> Author: rjung >> Date: Thu Sep 17 05:45:18 2009 >> New Revision: 816061 >> >> URL: http://svn.apache.org/viewvc?rev=816061&view=rev >> Log: >> Fix >> - implicit declaration of function 'fchmod' >> - 'S_I[RWX](USR|GRP|OTH)' undeclared > > Good catch, but... > >> +#ifdef HAVE_FCHMOD >> +#ifdef HAVE_SYS_STAT_H >> +#include <sys/stat.h> >> +#endif >> +#endif > > Grrr... this should *never* occur. The HAVE_tests should not be > layered. What is your question here? (And this goes for your > commit to ftp_util.c as well.) Is it FCHMOD or STAT_H? Is it WIN32 > or STAT_H? > > Including a system header should be harmless, so you should be > dropping the overarching conditions.
Aha, I see: you moved those includes in r782531 from those files into ftp_internal.h: http://svn.apache.org/viewvc?diff_format=h&view=rev&revision=782531 Now in ftp_internal.h the feature APR_HAVE_SYS_STAT_H is tested. And it seems that in contrast to APR_HAVE_SYS_TYPES_H, the define APR_HAVE_SYS_STAT_H never makes it into the apr header files. It's missing in apr.h.in. So I would say we need to fix APR_HAVE_SYS_STAT_H definition in apr and should use HAVE_SYS_STAT_H in ftp_internal.c instead. I will revert the include changes to the two C files. Anything I overlooked? Regards, Rainer
