On 11 May 2015 at 17:54, Marek Polacek <pola...@redhat.com> wrote:
> The problem here isn't in the -Wshift-negative-value warning itself; the
> problem is with marking -1 << 0 as a non-constant: later on, we warn in
> a context where a constant expression is needed ("initializer element is
> not a constant expression"), and for e.g. int foo = -1 << 0 | 9; there's
> an error ("initializer element is not constant").Yes, I understand. What I'm saying is that if folding was done at the moment that the constant is requested, this would be a case of pedwarn-if-pedantic, which is exactly what Joseph is discussing here for case (b): https://gcc.gnu.org/ml/gcc/2013-11/msg00253.html It doesn't matter that the error "initializer element is not a constant expression" is not given as long as an error (e.g., "left shift of negative value is undefined according to ISO C99") is given with -pedantic-errors. I'm sorry to sound so "pedantic" but we have faced this same issue in the past (https://gcc.gnu.org/bugzilla/show_bug.cgi?id=19976#c7) and the solution was to delay folding (https://gcc.gnu.org/ml/gcc-patches/2008-10/msg01061.html). > My change means that we wouldn't complain unless -pedantic (to not upset > too many users). I'm not particularly fond of it, but it seems like the > simplest solution. -Wpedantic (which is the same as -pedantic) should not give errors. Otherwise, it would mean that someone who uses -Wpedantic- -Wno-pedantic will silence the error, but someone that uses -Werror=pedantic -Wno-error=pedantic will still see an error (!). There is also no way to tell that error is related to -Wpedantic. But if this is the accepted solution, please at least test global_dc->pedantic_errors instead and put a big /* ???? This is a hack to reject non-conforming programs with -pedantic-errors and accept them otherwise. See https://gcc.gnu.org/ml/gcc-patches/2008-10/msg01061.html and https://gcc.gnu.org/ml/gcc/2013-11/msg00253.html about to fix this properly. */ Cheers, Manuel.
