On Tuesday 17 January 2012 01:55:45 Case Van Horsen wrote: > Hi, > > I've compiled Brian's mpir-exp branch with the VS2008 SDK and a > modified configure.bat/make.bat. I changed the warning level to /W3 > and examined some of the warnings. The first three warnings I examined > are: > > 1) mpn_addsub_n is defined to return int but the returned value 'ret' > is defined as mp_limb_t. Should ret be defined as int and the > intermediate results from mpn_add/mpn_sub be cast to int? >
the addsub can return -1,0 or 1 so a signed mp_limb_t would best , perhaps we should change it to a signed long (win64 signed long long) . We could have it unsigned but you would need the usual if(mpn_addsub(....) <0 )..... to be writen as if(mpn_addsub(....) == -1 )..... and perhaps a cast to mp_limb_t on the -1 As the next version will export addsub we should come to a decision > 2) In tdiv_qr_3by2, the fragment "_mask = - (mp_limb_t) ((r1) >= > _q0)" generates a warning. On two's complement hardware, I think that > is equivalent to "_mask = (mp_limb_t)( -((r1) >= _q0))" which doesn't > generate a warning. > > 3) Is compat.c needed? Specifically, the duplicate definition > divexact_by_3 causes a warning. > The division functions are a bit of a pain , there are some others that the cl script also had to ignore. We will have to do something at some stage though > GCC especially is getting more strict in regards to undefined > behavior. For example, see > http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50066 > > Is it worth continuing to analyze these warnings? > > Is it a goal for MPIR to compile without warnings? And if so, at what level? > > Regards, > Case > > -- You received this message because you are subscribed to the Google Groups "mpir-devel" group. To post to this group, send email to mpir-devel@googlegroups.com. To unsubscribe from this group, send email to mpir-devel+unsubscr...@googlegroups.com. For more options, visit this group at http://groups.google.com/group/mpir-devel?hl=en.
