Hello all: After have a test, "gcc version 6.0.0 20151023 (experimental) (GCC)" has no this issue. And bug63510 can be closed. :-)
So for me, we need not spend additional time resources on it. I shall continue to other issues in gcc or qemu. Now, I guess, my 1st priority is to rewrite tilegx qemu floating point insns within 2015-10-31. Welcome additional ideas, suggestions, and completions. Thanks. On 10/24/15 08:15, Chen Gang wrote: > > On 10/23/15 16:56, Richard Biener wrote: >> On Thu, Oct 22, 2015 at 9:46 PM, Jeff Law <l...@redhat.com> wrote: >>> >>> Note that the call to fold_binary from tree-ssa-sccvn.c has been removed. >>> So that hunk either needs to be removed or the change applied elsewhere. >>> > > Oh, really, it uses gimple_simplify instead of. > >>> I think passing around the location through fold-const.c is OK. >>> > > OK, thanks. > >>> I'd like to see a testcase in a form ready for inclusion into the testsuite. > > OK, thanks, I shall try. > >> >> As an additional remark - I'd like to see us not use input_location >> but always loc, > > For me, it sounds reasonable. > >> even if UNKNOWN_LOCATION. The diagnostic machinery should handle this >> correctly(?). That is, if bootstrap/testign doesn't show testsuite >> regressions because >> of this. >> > > I will try. > > > Hope I can finish trying above all within 2 days (2015-10-25). > > > Thanks. > -- Chen Gang (陈刚) Open, share, and attitude like air, water, and life which God blessed