On 07/21/14 14:35, Dmitry Yemanov wrote: > 21.07.2014 13:13, Alex Peshkoff wrote: > >>> We were using system size_t until Nickolay had committed FB_SIZE_T a few >>> days ago in attempt to avoid warnings. The bad thing is that it wasnt't >>> discussed. >> Not too late to begin discussion now. >> I would sooner agree with the change and Nickolay's arguments to do it >> (present in comment in h-file) no matter of the fact that it would be >> good idea to check builds better after such changes. > Nickolay's comment is absolutely correct. However: > > 1) The root of the warnings issue was that size_t actively used in our > classes library was often compared and/or assigned to/from
Assignment from shorter var-s also causes problems? Strange for me. > shorter > SLONG/ULONG variables. And the compiler was correctly showing where our > code is not ready for big (>4GB) objects. Making both ULONG and > FB_SIZE_T the same [smaller] size is an obvious solution, but strictly > speaking it just hides the issue instead of fixing it. If one day we > want to support objects bigger than 4GB, one would need to enable > FB_USE_SIZE_T and start from scratch. > > This is not really an objection as long as FB_USE_SIZE_T is preserved, > rather just a comment of mine. > > 2) Also, it introduces more artificial datatypes than we had. I remember > we were dicussing an intention to minimize artificial datatypes in favor > of native compiler ones in all places that are not ODS/API related. New > API already does that. So I see the patch as a step backward in this > regard. And I really hate seeing/writing typedefs in such trivial places > like counter-based loops. Agreed. Even use of SLONG/ULONG in such loops looks ugly. I.e. we have to choose best from 2 bad variants? :) I think we should wait for Nickolay's mind before doing some big changes. > 3) If size_t was used in any public interface, then IMHO it was a > mistake. Looking at the new API, unsigned int should be used instead. > IMO, FB_SIZE_T must be a private typedef and thus should be moved from > types_pub.h to fb_types.h. The only such place is in circularAlloc()., and it can and should be changed to unsigned. Will do. ------------------------------------------------------------------------------ Want fast and easy access to all the code in your enterprise? Index and search up to 200,000 lines of code with a free copy of Black Duck Code Sight - the same software that powers the world's largest code search on Ohloh, the Black Duck Open Hub! Try it now. http://p.sf.net/sfu/bds Firebird-Devel mailing list, web interface at https://lists.sourceforge.net/lists/listinfo/firebird-devel
