Le Wed, 07 Apr 2010 22:57:19 +0200, Ingo Weinhold a écrit : > On 2010-04-07 at 22:57:39 [+0200], Axel Dörfler <axeld@xxxxxxxxxxxxxxxx> > > wrote: > > ingo_weinhold@xxxxxx wrote: > > > Log: > > > Define B_GENERAL_ERROR_BASE to INT_MIN instead of LONG_MIN. On 32 > > > bit > > > platforms it's the same value, but we actually want type int, not > > > long. > > > > Do we? At least status_t is int32 in BeOS, and this is defined to > > long, > > for whatever reason. > > Yeah, and that will be rectified when we break compatibility (or even > earlier in the gcc4 build, if anyone is motivated enough to fix all > the > resulting warnings). Besides, errno is int and on 64 bit platforms > long > would be 64 bit. Furthermore int promotes cleanly to long (unlike the > other > way around). Hmm int can be 64bit too on 64bit platforms IIRC, it depends on the data model, and there are many of them: http://www.viva64.com/terminology/Data_model.html François.