> > On 06/10/12 11:04 AM, Lu, Hongjiu wrote: > >> On 06/10/12 09:22 AM, Lu, Hongjiu wrote: > >>> We are proposing to extend x86-64 psABI to support > >>> ILP32 model: > >>> > >>> http://gcc.gnu.org/ml/gcc/2012-05/msg00258.html > >>> > >>> If a software package needs update to support ILP32 > >>> in x86-64, __ILP32__ can used to distinguish LP64 model > >>> from ILP32 model. If there are no differences > >>> between ILP32 and LP64 models, no changes are needed. If > >>> an x86-64 compiler doesn't support ILP32 model, checking > >>> __ILP32__ will make sure that it will works. > >> > >> If the software didn't care about the ABI, they wouldn't check for > >> __amd64__ in the first place. This is a compatibility breaking > >> change that should not be allowed. Define a new __x32__ symbol, > >> don't break __amd64__. > >> > > > > If a package doesn't support x32, it won't work on x32 > > anyway. Please show me one example of adding __ILP32__ check > > breaks anything. > > This patch is proof that redefining __amd64__ breaks existing software, > forcing us to go around adding new changes to avoid damage. It's not > adding the __ILP32__ check that breaks things, it's forcing all the > existing software to add it that is the breakage. >
That is false. You still don't show adding __ILP32__ check will break anything. H.J. _______________________________________________ [email protected]: X.Org development Archives: http://lists.x.org/archives/xorg-devel Info: http://lists.x.org/mailman/listinfo/xorg-devel
