I looked at r278501 and realized it’s doing more than just fixing the test case in PR28288. For example, when the following code was compiled,
vi8 = vi8 << vuc8; // vuc8: <8 x unsigned char>, vi8: <8 x int> vc8 = vc8 << vs4; // vc8: <8 x char>, vs4: <4 x short> prior to r278501, clang would reject the first statement but accept the second, while after r278501, it would accept the first and reject the second. gcc 6.2 rejects both. Was this change discussed thoroughly and is this really the behavior we want? > On Sep 7, 2016, at 12:53 AM, Vladimir Yakovlev <vbyakov...@gmail.com> wrote: > > I did another fix (attached). It fixes the error and chang-check has not any > new errors. Please take a look. > > Where can I see your fix? > > > <patch.p> _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
